You are on page 1of 323

Flask Documentation

Release 0.12
Jan 25, 2017

CONTENTS

I User’s Guide 1
1 Foreword 3

2 Foreword for Experienced Programmers 5

3 Installation 7

4 Quickstart 11

5 Tutorial 29

6 Templates 45

7 Testing Flask Applications 49

8 Application Errors 57

9 Debugging Application Errors 63

10 Configuration Handling 65

11 Signals 73

12 Pluggable Views 77

13 The Application Context 83

14 The Request Context 87

15 Modular Applications with Blueprints 93

16 Flask Extensions 99

17 Command Line Interface 101

18 Development Server 107

19 Working with the Shell 109

i

20 Patterns for Flask 111

21 Deployment Options 169

22 Becoming Big 183

II API Reference 187
23 API 189

III Additional Notes 255
24 Design Decisions in Flask 257

25 HTML/XHTML FAQ 261

26 Security Considerations 267

27 Unicode in Flask 269

28 Flask Extension Development 273

29 Pocoo Styleguide 281

30 Python 3 Support 285

31 Upgrading to Newer Releases 287

32 Flask Changelog 297

33 License 311

ii

Part I
USER’S GUIDE

This part of the documentation, which is mostly prose, begins with some background
information about Flask, then focuses on step-by-step instructions for web develop-
ment with Flask.

1

2

so that Flask can be everything you need and nothing you don’t. 1. form validation or any- thing else where different libraries already exist that can handle that. Flask does not include a database abstraction layer. but it’s ready for production use on a variety of needs. templates and static files are stored in subdi- rectories within the application’s Python source tree. 1. While this can be changed. Instead. with the names templates and static respectively. This hopefully answers some questions about the purpose and goals of the project. 3 . such as what database to use. with sensible defaults. you usually don’t have to. Flask won’t make many decisions for you. and a few conventions when getting started. upload handling. various open authentication technologies. The “micro” in microframework means Flask aims to keep the core simple but exten- sible. form val- idation. Flask supports extensions to add such functionality to your application as if it was imple- mented in Flask itself. and when you should or should not be using it. especially when getting started. Numerous extensions provide database integration. Those decisions that it does make.2 Configuration and Conventions Flask has many configuration values. and more. Everything else is up to you. such as what templating engine to use. Flask may be “micro”. By convention. are easy to change. By default. nor does it mean that Flask is lacking in functionality. CHAPTER ONE FOREWORD Read this before you get started with Flask.1 What does “micro” mean? “Micro” does not mean that your whole web application has to fit into a single Python file (although it certainly can).

The Flask core team reviews extensions and ensures approved extensions do not break with future releases. Flask will continue to provide a very simple glue layer to the best that Python has to offer.3 Growing with Flask Once you have Flask up and running. If you are curious about the Flask design principles. you’ll find a variety of extensions available in the community to integrate your project for production. the Python web interface. Should you need more cus- tomization. You can implement advanced patterns in SQLAlchemy or an- other database tool. Continue to Installation.1. As your codebase grows. the Quickstart. or the Foreword for Experienced Programmers. Flask includes many hooks to customize its behavior. you are free to make the design decisions appropriate for your project. 4 . introduce non-relational data persistence as appropriate. head over to the section about Design Decisions in Flask. check out the Becoming Big chapter. the Flask class is built for subclassing. If you are interested in that. and take advantage of framework-agnostic tools built for WSGI.

And even if you are the only user that might leave data in your application. 5 . Some of these security concerns are far more complex than one might think. there are many ways the security of a web application can be com- promised.2 Develop for the Web with Caution Always keep security in mind when building web applications. you still want that data to be stored securely. they should not take a lot of code and yet they should not limit you. CHAPTER TWO FOREWORD FOR EXPERIENCED PROGRAMMERS 2. Flask protects you against one of the most common security problems of modern web applications: cross-site scripting (XSS). you are probably allowing users to register and leave their data on your server. The Flask project is honest about thread-locals.until a clever attacker figures out a way to exploit our applications. The users are entrusting you with data. Because of that. Flask uses thread-local objects internally so that you don’t have to pass ob- jects around from function to function within a request in order to stay threadsafe. But there are many more ways to cause security problems. Unfortunately. For example. and calls out in the code and documentation where they are used. does not hide them. And don’t think that your application is not important enough to attract an attacker.1 Thread-Locals in Flask One of the design decisions in Flask was that simple tasks should be simple. and we all sometimes underestimate the likelihood that a vulnerability will be exploited . Flask has a few design choices that some people might find surprising or unorthodox. 2. but requires a valid request context for dependency in- jection or when attempting to reuse code which uses a value pegged to the request. This approach is convenient. Unless you deliberately mark in- secure HTML as secure. Flask and the underlying Jinja2 template engine have you covered. The documentation will warn you about aspects of web development that require at- tention to security. If you write a web application.

Flask is no different from any other framework in that you the developer must build with caution. 6 . If you want to use Flask with Python 3 have a look at the Python 3 Support page. watching for exploits when building to your requirements. chances are that automated bots are probing for ways to fill your database with spam. and most Flask extensions all support Python 3. its dependencies. Continue to Installation or the Quickstart. 2.3 Python 3 Support in Flask Flask.Depending on the kind of attack. and the like. links to malicious software.

Werkzeug is a toolkit for WSGI. It doesn’t actually install separate copies of Python. 3.x installation. too. So how do you get all that on your computer quickly? There are many ways you could do that. Jinja2 renders templates. you’ll probably want to use it there. try: 7 . chances are that the following command will work for you: $ sudo pip install virtualenv It will probably install virtualenv on your system. but it does provide a clever way to keep different project environments isolated. Let’s see how virtualenv works.6 or newer to get started. For using Flask with Python 3 have a look at Python 3 Support. If you use Ubuntu. the more likely it is that you will be working with different versions of Python itself. and it’s unlikely that any serious application will have zero dependencies. So what do you do if two or more of your projects have conflicting dependencies? Virtualenv to the rescue! Virtualenv enables multiple side-by-side installations of Python. or at least different versions of Python libraries. Let’s face it: quite often libraries break backwards compatibility. like Werkzeug and Jinja2. one for each project. What problem does virtualenv solve? If you like Python as much as I do. so be sure to have an up-to-date Python 2. but the most kick-ass method is virtualenv. But the more projects you have. and if you have shell access to your production machines. Maybe it’s even in your package manager. the standard Python interface between web applications and a vari- ety of servers for both development and deployment. If you are on Mac OS X or Linux. so let’s have a look at that first.1 virtualenv Virtualenv is probably what you want to use during development. chances are you want to use it for other projects besides Flask-based web applications. You will need Python 2. CHAPTER THREE INSTALLATION Flask depends on some external libraries.

pip.$ sudo apt-get install python-virtualenv If you are on Windows and don’t have the easy_install command.. Once you have it installed.. Just run pip with root privi- leges: $ sudo pip install Flask (On Windows systems. you should now be using your virtualenv (notice how the prompt of your shell has changed to show the active environment).. though I do not recommend it... you must install it first. I usually create a project folder and a venv folder within: $ mkdir myproject $ cd myproject $ virtualenv venv New python executable in venv/bin/python Installing setuptools. the prompt of your shell should be as familiar as before. do the following: $ . whenever you want to work on a project.. use the following command: $ deactivate After doing this. let’s move on.. And if you want to go back to the real world. just fire up a shell and create your own environ- ment.. On OS X and Linux.2 System-Wide Installation This is possible as well.) 8 . and leave out sudo.. run the same commands as above. Enter the following command to get Flask activated in your vir- tualenv: $ pip install Flask A few seconds later and you are good to go.. you only have to activate the corre- sponding environment. Once you have virtualenv installed. the following command is for you: $ venv\Scripts\activate Either way. venv/bin/activate If you are a Windows user. Check the pip and setuptools on Windows section for more information about how to do that.done. run it in a command-prompt window with administrator priv- ileges.. 3. but without the sudo prefix. Now. Now.

Fortunately there is a “bootstrap script” you can run to install.. Get the git checkout in a new virtualenv and run in development mode: $ git clone http://github.py file and run that. Finally. To fix this. then find the win_add2path. virtualenv is recommended... then go to Tools.. get-pip. but nothing very hard. Then all you have to do is run git pull origin to update to the latest version. once you pull up a command prompt you want to be able to type pip and python which will run those things.. If you already have pip. The crucial package you will need is pip .this will let you install anything else (like virtualenv).. 3.3 Living on the Edge If you want to work with the latest version of Flask.g C:Python27).4 pip and setuptools on Windows Sometimes getting the standard “Python packaging tools” like pip.. to install virtualenv. because it doesn’t know where those executables are (give either a try!)... Finished processing dependencies for Flask This will pull in the dependencies and activate the git head as the current version inside the virtualenv. or you can tell it to operate on a git checkout. setuptools and virtualenv can be a little trickier.git/ $ cd flask $ virtualenv venv New python executable in venv/bin/python Installing setuptools.done.git Initialized empty Git repository in ~/dev/flask/. you can simply run: 9 . If you don’t currently have pip. $ . Either way.py will install it for you. but this might not automatically happen on Windows..3. venv/bin/activate $ python setup. you can upgrade them by running: > pip install --upgrade pip setuptools Most often.. Open a new Command Prompt and check that you can now just type python to bring up the interpreter. there are two ways: you can either let pip pull in the development version.com/pallets/flask. then Scripts. then get-pip. pip.py develop ...py It should be double-clickable once you download it. you should be able to navigate to your Python install directory (e.

> pip install virtualenv Then you can be off on your way following the installation instructions above. 10 .

This is needed so that Flask knows where to look for templates. and so on. The function is given a name which is also used to generate URLs for that partic- ular function. The first argument is the name of the application’s module or package. Make sure to not call your application flask. First we imported the Flask class.route('/') def hello_world(): return 'Hello. An instance of this class will be our WSGI application. CHAPTER FOUR QUICKSTART Eager to get started? This page gives a good introduction to Flask. 2.1 A Minimal Application A minimal Flask application looks something like this: from flask import Flask app = Flask(__name__) @app.py or something similar. Next we create an instance of this class. We then use the route() decorator to tell Flask what URL should trigger our function. Just save it as hello. It assumes you already have Flask installed. If you do not. and returns the message we want to display in the user’s browser. 11 . static files. head over to the Installation section. you should use __name__ because depending on if it’s started as ap- plication or imported as module the name will be different ('__main__' versus the actual import name). If you are using a single module (as in this example).py because this would conflict with Flask itself. World!' So what did that code do? 1. 3. 4. 4. For more information have a look at the Flask documentation.

11 use to have different ways to start the application.py $ python -m flask run * Running on http://127. Now head over to http://127.0.1:5000/.0. and neither did python -m flask.1:5000/ If you are on Windows you need to use set instead of export. First of all you need to look at the error message.1 Old Version of Flask Versions of Flask older than 0. Alternatively you can use python -m flask: $ export FLASK_APP=hello.0.2 What to do if the Server does not Start In case the python -m flask fails or flask does not exist. 4.0. In short. Before you can do that you need to tell your terminal the application to work with by exporting the FLASK_APP environment variable: $ export FLASK_APP=hello. you can make the server publicly available simply by adding --host=0.2. there are multiple reasons this might be the case.0 to the command line: flask run --host=0.0 This tells your operating system to listen on all public IPs. not from any other in the network. For deployment options see Deployment Options. 4. the flask command did not exist.To run the application you can either use the flask command or python’s -m switch with Flask.0.0. which is good enough for testing but prob- ably not what you want to use in production. and you should see your hello world greet- ing.0.0.0. If you have the debugger disabled or trust the users on your network.1:5000/ This launches a very simple builtin server. Externally Visible Server If you run the server you will notice that the server is only accessible from your own computer.py $ flask run * Running on http://127.0. This is the default because in debugging mode a user of the application can execute arbitrary Python code on your computer. 12 . In that case you have two options: either upgrade to newer Flask versions or have a look at the Development Server docs to see the alternative method for running a server.

This does the following things: 1.2. To enable debug mode you can export the FLASK_DEBUG environment variable before running the server: $ export FLASK_DEBUG=1 $ flask run (On Windows you need to use set instead of export). it activates the automatic reloader 3. and it will also provide you with a helpful debugger if things go wrong.3 Debug Mode (Want to just log errors and stack traces? See Application Errors) The flask script is nice to start a local development server. it enables the debug mode on the Flask application. it still allows the execution of arbitrary code. 4. If you enable debug support the server will reload itself on code changes.4.2 Invalid Import Name The FLASK_APP environment variable is the name of the module to import at flask run. it activates the debugger 2. Screenshot of the debugger in action: 13 . This makes it a major security risk and therefore it must never be used on production machines. There are more parameters that are explained in the Development Server docs. The most common reason is a typo or because you did not actually create an app object. Attention Even though the interactive debugger does not work in forking environments (which makes it nearly impossible to use on production servers). That is not very nice and Flask can do better. In case that module is incorrectly named you will get an import error upon start (or if debug is enabled when you navigate to the application). It will tell you what it tried to import and why it failed. but you would have to restart it manually after each change to your code.

As you have seen above.Have another debugger in mind? See Working with Debuggers. 4. This helps people remember the URLs. Here are some basic examples: @app. World' 14 . the route() decorator is used to bind a function to a URL.route('/') def index(): return 'Index Page' @app. which is especially handy for applications that are used from mobile devices with slower network connections.4 Routing Modern web applications have beautiful URLs.route('/hello') def hello(): return 'Hello. If the user can directly go to the desired page with- out having to hit the index page it is more likely they will like the page and come back next time.

Optionally a converter can be used by specifying a rule with <converter:variable_name>. The idea behind that module is to ensure beautiful and unique URLs based on precedents laid down by Apache and earlier HTTP servers. the canonical URL for the projects endpoint has a trailing slash. it is similar to a folder on a filesystem. Such a part is then passed as a keyword argument to your function. the id is an integer return 'Post %d' % post_id The following converters exist: string accepts any text without a slash (the default) int accepts integers float like int but for floating point values path like the default but also accepts slashes any matches one of the items provided uuid accepts UUID strings Unique URLs / Redirection Behavior Flask’s URL rules are based on Werkzeug’s routing module. Take these two rules: @app. Accessing it without a trailing slash will cause Flask to redirect to the canonical URL with the trailing slash. 15 . In that sense.route('/projects/') def projects(): return 'The project page' @app. Here are some nice examples: @app. they differ in their use of the trailing slash in the URL definition.route('/about') def about(): return 'The about page' Though they look rather similar. In the first case.1 Variable Rules To add variable parts to a URL you can mark these special sections as <variable_name>. 4.But there is more to it! You can make certain parts of the URL dynamic and attach multiple rules to a function.4.route('/user/<username>') def show_user_profile(username): # show the user profile for that user return 'User %s' % username @app.route('/post/<int:post_id>') def show_post(post_id): # show the post with the given id.

the URL is defined without a trailing slash.. >>> @app.. however.test_request_context(): ... username='John Doe') . 4.In the second case. def index(): pass . Context Locals). To build a URL to a specific function you can use the url_for() function.route('/user/<username>') . print url_for('login') ... def login(): pass ... explained below.. the URLs will stay unique. can Flask also generate them? Of course it can. without having to remember to change URLs all over the place.. Reversing is often more descriptive than hard-coding the URLs.. it allows you to change URLs in one go. each corresponding to the variable part of the URL rule.. even though we are interacting with it through a Python shell.. print url_for('profile'. 16 . >>> with app.route('/login') . url_for >>> app = Flask(__name__) >>> @app. Have a look at the explanation below. Unknown variable parts are appended to the URL as query parameters.. which helps search engines avoid indexing the same page twice. print url_for('login'.. next='/') . Also......route('/') . Accessing the URL with a trailing slash will produce a 404 “Not Found” error.4.. It accepts the name of the function as first argument and a number of keyword arguments. It tells Flask to behave as though it is handling a request. def profile(username): pass ..2 URL Building If it can match URLs. More impor- tantly. Here are some examples: >>> from flask import Flask. print url_for('index') . / /login /login?next=/ /user/John%20Doe (This also uses the test_request_context() method. Why would you want to build URLs using the URL reversing function url_for() in- stead of hard-coding them into your templates? There are three good reasons for this: 1. consistent with how Apache and other servers work. >>> @app. This behavior allows relative URLs to continue working even if the trailing slash is omitted. rather like the pathname of a file on UNIX-like systems.

say.route('/login'. An application is supposed to handle that as if a GET request was received but to not deliver the actual content. Likewise. In Flask you don’t have to deal with that at all. 'POST']) def login(): if request. Now you might be asking why this is useful. the underlying Werkzeug library handles that for you. as of Flask 0. but it is only interested in the headers. The following methods are very common: GET The browser tells the server to just get the information stored on that page and send it. 2. methods=['GET'. 3. 4. This is probably the most common method. By default. but that can be changed by providing the methods argument to the route() decorator. so you don’t have to deal with them. It will also make sure that HEAD requests are handled as the HTTP RFC (the document describing the HTTP protocol) demands. If your application is placed outside the URL root . You don’t have to deal with that. HEAD The browser tells the server to get the information. but there are some good reasons to do it this way. OPTIONS is implemented for you automatically as well. URL building will handle escaping of special characters and Unicode data trans- parently for you. in /myapplication in- stead of / .3 HTTP Methods HTTP (the protocol web applications are speaking) knows different methods for ac- cessing URLs.method == 'POST': do_the_login() else: show_the_login_form() If GET is present. Here are some examples: from flask import request @app. POST The browser tells the server that it wants to post some new information to that URL and that the server must ensure the data is stored and only stored once. not the content of the page. This is how HTML forms usually transmit data to the server. PUT Similar to POST but the server might trigger the store procedure multiple times by overwriting the old values more than once. Consider that the connection is lost during transmission: in this situation a system between the browser and the server might receive the request safely a second time without 17 .url_for() will handle that properly for you. a route only answers to GET requests. here is a quick introduction to HTTP methods and why they matter: The HTTP method (also often called “the verb”) tells the server what the client wants to do with the requested page.4. HEAD will be added automatically for you. You have no idea what an HTTP method is? Worry not.6. so you can completely ignore that part of the HTTP specification.

name=name) Flask will look for templates in the templates folder. and actually pretty cumbersome be- cause you have to do the HTML escaping on your own to keep the application secure. Here’s a simple example of how to render a template: from flask import render_template @app. 4. use the special 'static' endpoint name: url_for('static'.5 Static Files Dynamic web applications also need static files.route('/hello/<name>') def hello(name=None): return render_template('hello. But with JavaScript and future HTML standards you can use the other methods as well.css. the only methods a form can submit to the server are GET and POST. Because of that Flask configures the Jinja2 template engine for you automatically. Now the interesting part is that in HTML4 and XHTML1.6. Furthermore HTTP has become quite popular lately and browsers are no longer the only clients that are using HTTP. many revision control systems use it. 4. To generate URLs for static files. if it’s a package it’s actually inside your package: Case 1: a module: 18 . To render a template you can use the render_template() method.html'. That’s usually where the CSS and JavaScript files are coming from.css') The file has to be stored on the filesystem as static/style. OPTIONS Provides a quick way for a client to figure out which methods are supported by this URL. For instance. breaking things. filename='style. DELETE Remove the information at the given location. With POST that would not be possible because it must only be triggered once. but during development Flask can do that as well.6 Rendering Templates Generating HTML from within Python is not fun.route('/hello/') @app. Just create a folder called static in your package or next to your module and it will be available at /static on the application. All you have to do is provide the name of the template and the variables you want to pass to the template engine as keyword arguments. Ideally your web server is configured to serve them for you. this folder is next to that module. So if your application is a module. Starting with Flask 0. this is implemented for you automatically.

Automatic escaping is enabled.html For templates you can use the full power of Jinja2 templates.py /templates /hello.hacker&lt./application. head over to the Template Inheritance pattern documentation. session and g1 objects as well as the get_flashed_messages() function. Head over to the official Jinja2 Template Documentation for more information. HTML'). If you want to know how that works./blink&gt. check the documentation of that object (g) and the Using SQLite 3 with Flask for more information. so if name contains HTML it will be escaped automati- cally.') >>> Markup('<em>Marked up</em> &raquo.escape('<blink>hacker</blink>') Markup(u'&lt. navigation and footer)./blink&gt. If you can trust a variable and you know that it will be safe HTML (for example because it came from a module that converts wiki markup to HTML) you can mark it as safe by using the Markup class or by using the |safe filter in the template.blink&gt. 19 . Templates are especially useful if inheritance is used. Basically tem- plate inheritance makes it possible to keep certain elements on each page (like header. Head over to the Jinja 2 documentation for more examples.hacker&lt. Here is an example template: <!doctype html> <title>Hello from Flask</title> {% if name %} <h1>Hello {{ name }}!</h1> {% else %} <h1>Hello. World!</h1> {% endif %} Inside templates you also have access to the request.blink&gt. The 1Unsure what that g object is? It’s something in which you can store information for your own needs.html Case 2: a package: /application /__init__. Here is a basic introduction to how the Markup class works: >>> from flask import Markup >>> Markup('<strong>Hello %s!</strong>') % '<blink>hacker</blink>' Markup(u'<strong>Hello &lt.striptags() u'Marked up \xbb HTML' Changed in version 0.!</strong>') >>> Markup.py /templates /hello.5: Autoescaping is no longer enabled for all templates.

Certain objects in Flask are global objects.method == 'POST' 20 . In combination with the with statement it will bind a test request so that you can interact with it. Templates loaded from a string will have autoescaping disabled.xhtml. read this section. But that is actually quite easy to understand.htm. the underlying object is capable of dealing with concurrency systems other than threads). Imagine the context being the handling thread.test_request_context('/hello'. otherwise just skip it. The easiest solution for unit testing is to use the test_request_context() context manager. such as basic assertions: assert request. The solution is creating a request object yourself and binding it to the context. method='POST'): # now you can do something with the request until the # end of the with block. So what does this mean to you? Basically you can completely ignore that this is the case unless you are doing something like unit testing. If you have some experience with Python you might be wondering how that object can be global and how Flask manages to still be threadsafe. In Flask this information is provided by the global request object. These objects are actually proxies to objects that are local to a specific context. It does that in an intelligent way so that one application can invoke another application without breaking. 4. . The answer is context locals: 4. You will notice that code which depends on a request object will suddenly break because there is no request object. . When Flask starts its inter- nal request handling it figures out that the current thread is the active context and binds the current application and the WSGI environments to that context (thread).following extensions for templates trigger autoescaping: . but not of the usual kind.1 Context Locals Insider Information If you want to understand how that works and how you can implement tests with context locals. . What a mouthful. Here is an example: from flask import request with app. A request comes in and the web server decides to spawn a new thread (or something else.html.7.7 Accessing Request Data For web applications it’s crucial to react to the data a client sends to the server.xml.path == '/hello' assert request.

args.2 The Request Object The request object is documented in the API section and we will not cover it here in detail (see request). 21 . To access parameters submitted in the URL (?key=value) you can use the args at- tribute: searchword = request.The other possibility is passing a whole WSGI environment to the request_context() method: from flask import request with app.get('key'.form['password']): return log_the_user_in(request. '') We recommend accessing URL parameters with get or by catching the KeyError be- cause users might change the URL and presenting them a 400 bad request page in that case is not user friendly. a HTTP 400 Bad Request error page is shown instead.route('/login'. request. Here is a broad overview of some of the most common operations. Here is a full example of the two attributes mentioned above: @app. So for many situations you don’t have to deal with that problem.7. For a full list of methods and attributes of the request object.method == 'POST': if valid_login(request. 'GET']) def login(): error = None if request. error=error) What happens if the key does not exist in the form attribute? In that case a special KeyError is raised.html'.form['username']. First of all you have to import it from the flask module: from flask import request The current request method is available by using the method attribute. methods=['POST'.method == 'POST' 4. You can catch it like a standard KeyError but if you don’t do that. head over to the request documentation.form['username']) else: error = 'Invalid username/password' # the code below is executed if the request method # was GET or the credentials were invalid return render_template('login.request_context(environ): assert request. To access form data (data transmitted in a POST or PUT request) you can use the form attribute.

.files['the_file'] f. Each uploaded file is stored in that dictionary. checkout the Uploading Files pattern. It behaves just like a standard Python file object..method == 'POST': f = request. 'POST']) def upload_file(): if request. To set cookies you can use the set_cookie method of response objects. Just make sure not to forget to set the enctype="multipart/form-data" attribute on your HTML form. You can access those files by looking at the files attribute on the request object. you can access the filename attribute. 4. For some better examples. but it also has a save() method that allows you to store that file on the filesys- tem of the server. However please keep in mind that this value can be forged so never ever trust that value.save('/var/www/uploads/' + secure_filename(f. Here is a simple example showing how that works: from flask import request @app. do not use the cookies directly but instead use the Sessions in Flask that add some security on top of cookies for you.txt') . 'POST']) def upload_file(): if request. If you want to know how the file was named on the client before it was uploaded to your application. Uploaded files are stored in memory or at a temporary location on the filesystem. Reading cookies: 22 . methods=['GET'.method == 'POST': f = request.7.7. If you want to use sessions.utils import secure_filename @app.3 File Uploads You can handle uploaded files with Flask easily.save('/var/www/uploads/uploaded_file. otherwise the browser will not transmit your files at all..route('/upload'.4 Cookies To access cookies you can use the cookies attribute.route('/upload'.4.filename)) . The cookies attribute of request objects is a dictionary with all the cookies the client transmits. methods=['GET'.files['the_file'] f. If you want to use the file- name of the client to store the file on the server. pass it through the secure_filename() function that Werkzeug provides for you: from flask import request from werkzeug..

get('username') # use cookies.route('/') def index(): username = request. If you want to customize the error page. By default a black and white error page is shown for each error code.cookies..)) resp.. 4.route('/') def index(): resp = make_response(render_template(. you can use the errorhandler() decorator: 23 .route('/login') def login(): abort(401) this_is_never_executed() This is a rather pointless example because a user will be redirected from the index to a page they cannot access (401 means access denied) but it shows how that works.get(key) instead of cookies[key] to not get a # KeyError if the cookie is missing. This is possible by utilizing the Deferred Request Callbacks pattern. If you explicitly want to do that you can use the make_response() function and then modify it. url_for @app. 'the username') return resp Note that cookies are set on response objects. Sometimes you might want to set a cookie at a point where the response object does not exist yet. For this also see About Responses. Storing cookies: from flask import make_response @app. Since you normally just return strings from the view functions Flask will convert them into response objects for you. use the redirect() function. redirect.8 Redirects and Errors To redirect a user to another endpoint.from flask import request @app. to abort a request early with an error code. use the abort() function: from flask import abort.set_cookie('username'.route('/') def index(): return redirect(url_for('login')) @app.

a 200 OK status code and a text/html mimetype. 2. Such tuples have to be in the form (response. headers) or (response.errorhandler(404) def page_not_found(error): return render_template('page_not_found. The status value will override the status code and headers can be a list or dictionary of additional header values. 404) 24 . status.html'). 4. If the return value is a string it’s converted into a response object with the string as response body.errorhandler(404) def not_found(error): resp = make_response(render_template('error. If you want to get hold of the resulting response object inside the view you can use the make_response() function. Flask will assume the return value is a valid WSGI appli- cation and convert that into a response object. If a tuple is returned the items in the tuple can provide extra information. 3. See Error handlers for more details.errorhandler(404) def not_found(error): return render_template('error. By default 200 is assumed which translates to: all went well. 404 You just need to wrap the return expression with make_response() and get the re- sponse object to modify it.9 About Responses The return value from a view function is automatically converted into a response ob- ject for you. If it’s a string. The logic that Flask applies to converting return values into response objects is as follows: 1. Imagine you have a view like this: @app. a response object is created with that data and the default param- eters. 4. If none of that works. then return it: @app.html'). headers) where at least one item has to be in the tuple. This tells Flask that the status code of that page should be 404 which means not found.from flask import render_template @app. If a response object of the correct type is returned it’s directly returned from the view. 404 Note the 404 after the render_template() call.html').

resp. 'POST']) def login(): if request.10 Sessions In addition to the request object there is also a second object called session which allows you to store information specific to a user from one request to the next.?RT' The escape() mentioned here does escaping for you if you are not using the template engine (as in this example).route('/login'.route('/logout') def logout(): # remove the username from the session if it's there session. In order to use sessions you have to set a secret key. url_for.form['username'] return redirect(url_for('index')) return ''' <form method="post"> <p><input type=text name=username> <p><input type=submit value=Login> </form> ''' @app.route('/') def index(): if 'username' in session: return 'Logged in as %s' % escape(session['username']) return 'You are not logged in' @app. This is implemented on top of cookies for you and signs the cookies cryptographically. What this means is that the user could look at the contents of your cookie but not modify it.headers['X-Something'] = 'A value' return resp 4.method == 'POST': session['username'] = request. session. None) return redirect(url_for('index')) # set the secret key. Here is how sessions work: from flask import Flask. keep this really secret: app. How to generate good secret keys 25 . redirect. request app = Flask(__name__) @app. escape.secret_key = 'A0Zr98j/3yX R~XHH!jmN]LWX/.pop('username'. unless they know the secret key used for signing. methods=['GET'.

And a secret key should be as random as possible. to get hold of the messages you can use get_flashed_messages() which is also available in the templates. To flash a message use the flash() method. This might be caused by a user tampering with the data. Flask provides a really simple way to give feedback to a user with the flashing system.5\xd1\x01O<!\xd5\xa2\xa0\x9fR"\xa1\xa8' Just take that thing and copy/paste it into your code and you're done. This is where loggers come in handy. For example you may have some client-side code that sends an HTTP request to the server but it’s obviously malformed.The problem with random is that it’s hard to judge what is truly random. there are several Flask extensions that support this. 26 . 4. check the size of the cookie in your page responses compared to the size supported by web browsers. but actually is not. if you want to handle sessions on the server-side instead. cookies are indeed enabled. and you are not getting a clear error message. 4.12 Logging New in version 0. Your operating system has ways to generate pretty random stuff based on a cryptographic random generator which can be used to get such a key: >>> import os >>> os. If the user does not get enough feedback they will probably end up hating the application. Sometimes you might be in a situation where you deal with data that should be correct. Check out the Mes- sage Flashing for a full example.3. As of Flask 0. but sometimes that won’t do and the code has to continue working. Most of the time it’s okay to reply with 400 Bad Request in that situation. You may still want to log that something fishy happened.urandom(24) '\xfd{H\xe5<\x95\xf9\xe3\x96. or the client code failing. Besides the default client-side based sessions.3 a logger is preconfigured for you to use. The flashing system basically makes it possible to record a message at the end of a request and access it on the next (and only the next) request. This is usually combined with a layout template to expose the message.11 Message Flashing Good applications and user interfaces are all about feedback. If you are finding some values do not per- sist across requests. A note on cookie-based sessions: Flask will take the values you put into the session object and serialize them into a cookie.

wsgi_app = LighttpdCGIRootFix(app. so head over to the official logging documentation for more information. For more on Flask extensions.15 Deploying to a Web Server Ready to deploy your new Flask app? Go to Deployment Options. For example if you want to use one of the middlewares from the Werkzeug package to work around bugs in lighttpd. have a look at Flask Extensions.contrib. Read more on Application Errors. Flask- SQLAlchemy provides SQLAlchemy support that makes it simple and easy to use with Flask. 27 . 42) app.13 Hooking in WSGI Middlewares If you want to add a WSGI middleware to your application you can wrap the internal WSGI application. you can do it like this: from werkzeug.error('An error occurred') The attached logger is a standard logging Logger.fixers import LighttpdCGIRootFix app.logger.Here are some example log calls: app.logger. 4. For example. 4.wsgi_app) 4.14 Using Flask Extensions Extensions are packages that help you accomplish common tasks.logger.debug('A value for debugging') app.warning('A warning occurred (%d apples)'.

28 .

Let the user sign in and out with credentials specified in the configuration. SQLite3 will be used directly for this application because it’s good enough for an ap- plication of this size. If you want the full source code in advance or for comparison. Here a screenshot of the final application: 29 . but feel free to choose your own less Web-2.0-ish name . allowing you to target different relational databases at once and more. they can add new entries to the page consisting of a text-only title and some HTML for the text. You might also want to consider one of the popular NoSQL databases if your data is more suited for those. CHAPTER FIVE TUTORIAL You want to develop an application with Python and Flask? Here you have the chance to learn by example. 5. For larger applications. it makes a lot of sense to use SQLAlchemy. check out the example source. It only supports one user that can create text-only entries and there are no feeds or comments. This HTML is not sanitized because we trust the user here. 2. however. we will create a simple microblogging application. as it handles database connections in a more intelligent way. The index page shows all entries so far in reverse chronological order (newest on top) and the user can add new ones from there if logged in. 3. We will use Flask and SQLite as a database (which comes out of the box with Python) so there is nothing else you need. it will do the following things: 1.1 Introducing Flaskr This tutorial will demonstrate a blogging application named Flaskr. but it still features everything you need to get started. In this tutorial.) Essentially. Only one user is supported. When the user is logged in.

2 Step 0: Creating The Folders Before getting started. Flask will look for Jinja2 templates. Inside the templates folder. You will see exactly how to run flaskr later on in this tutorial.Continue with Step 0: Creating The Folders. 30 . This is the recommended way to install and run Flask applications. As a quick side note. In the next few steps you will be creating the database schema as well as the main module. This is the place where CSS and JavaScript files go. 5. You will see examples of this later on. the files inside of the static folder are available to users of the application via HTTP. For now go ahead and create the applications directory structure. you will need to create the folders needed for this application: /flaskr /flaskr /static /templates The application will be installed and run as Python package.

and a text. redirect. After that there will be a few lines of configuration code. Continue with Step 2: Application Setup Code. The id is an automatically incrementing integer and a primary key. For small applications like flaskr. 5.join(app. it is possible to drop the configuration directly into the module.config.py file. 31 . session.4 Step 2: Application Setup Code Now that the schema is in place.py): # all the imports import os import sqlite3 from flask import Flask. g.root_path. the other two are strings that must not be null. a cleaner solution is to create a separate .sql in the flaskr/flaskr folder: drop table if exists entries. load that. Each row in this table has an id. url_for. This schema consists of a single table called entries.ini or .3 Step 1: Database Schema In this step.For now you should continue with Step 1: Database Schema. This file should be placed inside of the flaskr/flaskr folder.db'). you will create the database schema. 'flaskr.update(dict( DATABASE=os. 5.py # Load default config and override config from an environment variable app. \ render_template.py: app = Flask(__name__) # create the application instance :) app. flaskr. title text not null. you can create the application module. flaskr. abort. request. Only a single table is needed for this application and it will only support SQLite. Here are the import statements (in flaskr. The first several lines of code in the application module are the needed import statements. a title. However. All you need to do is put the following contents into a file named schema. 'text' text not null ). and import the values from there. flash The next couple lines will create the actual application instance and initialize it with the config from the same file in flaskr. create table entries ( id integer primary key autoincrement.path.from_object(__name__) # load config from this file .config.py.

config. This enables robust configuration setups. This can be used to open a connection on request and also from the inter- active Python shell or a script. we place the database right next to it.from_envvar('FLASKR_SETTINGS'. USERNAME='admin'. files can then easily be found. Database Path Operating systems know the concept of a current working directory for each process.""" rv = sqlite3.config. environment-specific configuration file. you cannot depend on this in web applications because you might have more than one application in the same process. For a real-world application. you will add a method that allows for easy connections to the specified database.root_path attribute can be used to get the path to the applica- tion. Together with the os. Lastly. it is a good idea to load a separate.from_envvar('FLASKR_SETTINGS'. silent=True) Simply define the environment variable FLASKR_SETTINGS that points to a config file to be loaded. PASSWORD='default' )) app. The silent switch just tells Flask to not complain if no such environment key is set. only variable names that are uppercase are considered. In addition to that. This will come in handy later.connect(app. For this reason the app. Flask allows you to import multiple configurations and it will use the setting defined in the last import.Row object to represent rows. from_envvar() can help achieve this. Note that in all cases. Flask will then initialize the variable from that module. SECRET_KEY='development key'. it’s recommended to use Instance Folders instead.path module. silent=True) The Config object works similarly to a dictionary. app. def connect_db(): """Connects to the specific database. so it can be updated with new val- ues. Unfortunately. Usually. The SECRET_KEY is needed to keep the client-side sessions secure.config['DATABASE']) 32 . Choose that key wisely and as hard to guess and complex as possible. This allows the rows to be treated as if they were dictionaries instead of tuples. In this example. You can create a simple database connection through SQLite and then tell it to use the sqlite3. you can use the from_object() method on the config object and provide it with an import name of a module.

Create the MANIFEST.in). In this case. Click provides Flask with en- hanced and extensible command line utilities.5 Step 3: Installing flaskr as a Package Flask is now shipped with built-in support for Click.py and MANIFEST.py schema.in The content of the setup. A useful pattern to manage a Flask application is to install your app following the Python Packaging Guide.Row return rv In the next section you will see how to run the application. You also need to add an __init__. 5. your code structure should be: /flaskr /flaskr __init__.py file for flaskr is: from setuptools import setup setup( name='flaskr'. the static and templates directories need to be included. ]. install_requires=[ 'flask'.py file to make the flaskr/flaskr directory a package. as well as the schema.in in the projects root directory. it is also necessary to specify any special files that should be included in your package (in the MANIFEST.sql setup. packages=['flaskr']. After these changes. rv.row_factory = sqlite3.in and add the following lines: 33 . include_package_data=True. ) When using setuptools. Continue with Step 3: Installing flaskr as a Package. Presently this involves creating two new files. Later in this tutorial you will see exactly how to extend the flask command line interface (CLI).py MANIFEST. setup.py /static /templates flaskr.

The FLASK_DEBUG flag enables or disables the interactive debugger. flaskr/.sql To simplify locating the application. you should be able to start up the application. Externally Visible Server Want your server to be publicly available? Check out the externally visible server section for more information. The above installation command assumes that it is run within the projects root direc- tory. you will get a 404 error because we don’t have any views yet. because it will allow users to execute code on the server! You will see a message telling you that server has started along with the address at which you can access it. add the following import statement into this file. The flaskr app is now installed in your virtualenv (see output of pip freeze). 34 . but first. the Flask development server needs the location of the app instance. When it is time to run the application.flaskr import app This import statement brings the application instance into the top-level of the appli- cation package. flaskr/__init__. The editable flag allows editing source code without having to reinstall the Flask app each time you make changes.graft flaskr/templates graft flaskr/static include flaskr/schema. Never leave debug mode activated in a production system.py: from . This import statement simplifies the location process. Continue with Step 4: Database Connections.flaskr. Do this with the following commands: export FLASK_APP=flaskr export FLASK_DEBUG=true flask run (In case you are on Windows you need to use set instead of export). you should get the database working. As usual. That will be addressed a little later. With that out of the way. Without it the export statement a few steps below would need to be export FLASK_APP=flaskr. it is recommended to install your Flask application within a virtualenv. With that said. When you head over to the server in your browser. go ahead and install the application with: pip install --editable . At this point you should be able to install the application.

teardown_appcontext def close_db(error): """Closes the database again at the end of the request. For in- stance. all you have to know is that you can store information safely on the g object. but by itself.5. 35 . Flask provides two contexts: the application context and the request context. but how can you properly disconnect? For that. For the time being.sqlite_db Now you know how to connect. you will need to make sure that only one request at a time uses the connection. Because database connections encapsulate a transaction. 'sqlite_db'): g. in which case the error is passed to the teardown function. Curious about what these contexts mean? Have a look at the The Application Context documentation to learn more. The first time the function is called. the request variable is the request object associated with the current request. What does this mean? Essentially.sqlite_db. So when do you put it on there? To do that you can make a helper function.sqlite_db = connect_db() return g. all you have to know is that there are special variables that use these. An elegant way to do this is by utilizing the application context. """ if not hasattr(g.6 Step 4: Database Connections You currently have a function for establishing a database connection with connect_db. It’s executed every time the application context tears down: @app.close() Functions marked with teardown_appcontext() are called every time the app context tears down. it will create a database connection for the current context. Creating and closing database connections all the time is very inefficient. Flask provides us with the teardown_appcontext() decorator. A tear- down can happen because of two reasons: either everything went well (the error pa- rameter will be None) or an exception happened. For the time being. whereas g is a general purpose variable associated with the current application con- text. it is not particularly useful. Continue to Step 5: Creating The Database. the app context is created before the request comes in and is destroyed (torn down) whenever the request finishes.""" if hasattr(g. and successive calls will return the already established connection: def get_db(): """Opens a new database connection if there is none yet for the current application context. 'sqlite_db'): g. so you will need to keep it around for longer. The tutorial will cover some more details of this later on.

you can put all of your application code into a single Python module. It’s a good idea to add a function that initializes the database for you.commit() @app.db < schema.sql The downside of this is that it requires the sqlite3 command to be installed.cli. For now just take a look at the code segment below. and put your new get_db and close_db functions below your ex- isting connect_db function (following the tutorial line-by-line). you can then access 36 .cursor().py: def init_db(): db = get_db() with app. Flaskr is a database powered application. is just below the connect_db function in flaskr. which can introduce errors. Within the function.7 Step 5: Creating The Database As outlined earlier. you can create a function and hook it into a flask command that initializes the database. and command. This also requires that you provide the path to the database. To do this. and more precisely. Flask will automatically create an application context which is bound to the right application. which is not necessarily the case on every system. 5. it’s important to create that schema. When the command executes.') The app. mode='r') as f: db.command('initdb') def initdb_command(): """Initializes the database.Hint: Where do I put this code? If you’ve been following along in this tutorial. to the application.""" init_db() print('Initialized the database.executescript(f. In Flask.command() decorator registers a new command with the flask script. Such a schema can be created by piping the schema.sql'. you might be wondering where to put the code from this step and the next. it is an application powered by a relational database system. and if your app grows larger.read()) db. If you need a moment to find your bearings.cli. Before starting the server for the first time. A logical place is to group these module-level functions together. You don’t have to.sql file into the sqlite3 command as follows: sqlite3 /tmp/flaskr. Such systems need a schema that tells them how to store that information. it’s a good idea not to. take a look at how the example source is organized.open_resource('schema. A good place to add this function.

SQLite3 and other transactional databases will not commit unless you explicitly tell it to.html template and return the rendered one: @app. You will want to keep an actual function around that initializes the database. It is used in this example to execute a script on the database connection. The connection object provided by SQLite can give you a cursor object. plural.8. the application context tears down and the database connection is released.Row row factory. This function opens a file from the resource location (the flaskr/flaskr folder) and allows you to read from it. there is a method to execute a complete script. On that cur- sor. you can start writing the view func- tions.route('/') def show_entries(): 37 .1 Show Entries This view shows all the entries stored in the database. The view function will pass the entries to the show_entries. The one with the highest id (the newest entry) will be on top. Continue with Step 6: The View Functions 5. Finally. Now. You will need four of them: 5. (For more information see Testing Flask Applications. you only have to commit the changes. The rows returned from the cursor look a bit like dictionaries because we are using the sqlite3. Troubleshooting If you get an exception later on stating that a table cannot be found.g and other things as you might expect. so that we can easily create databases in unit tests later on. It listens on the root of the application and will select title and text from the database.8 Step 6: The View Functions Now that the database connections are working. it is possible to create a database with the flask script: flask initdb Initialized the database. for example).flask.) The open_resource() method of the application object is a convenient helper function that will open a resource that the application provides. though. check that you did execute the initdb command and that your table names are correct (singular vs. When the script ends.

commit() flash('New entry was successfully posted') return redirect(url_for('show_entries')) Note that this view checks that the user is logged in (that is. text) values (?.3 Login and Logout These functions are used to sign the user in and out. if the logged_in key is present in the session and True). entries=entries) 5. the actual form is shown on the show_entries page. that key is set to True. and the user is asked again: @app. your app will be vulnerable to SQL injection when you use string formatting to build SQL statements. db = get_db() cur = db. and the user is redi- rected back to the show_entries page. In addition. If the user logged in successfully. [request. the template is notified about that. ?)'. 5.html'. This only responds to POST requests. Otherwise. as done in the example above. request.form['title']. text from entries order by id desc') entries = cur.config['USERNAME']: 38 .form['text']]) db. If an error occurred. Security Note Be sure to use question marks when building SQL statements. a message is flashed that informs the user that he or she was logged in successfully. If everything worked out well. Login checks the username and password against the ones from the configuration and sets the logged_in key for the session.8.route('/login'.method == 'POST': if request.8. it will flash() an information message to the next request and redirect back to the show_entries page: @app.execute('insert into entries (title. methods=['POST']) def add_entry(): if not session.form['username'] != app. 'POST']) def login(): error = None if request.execute('select title.fetchall() return render_template('show_entries.2 Add New Entry This view lets the user add new entries if they are logged in. See Using SQLite 3 with Flask for more.route('/add'. methods=['GET'.get('logged_in'): abort(401) db = get_db() db.

@app. so adding this functionality is fairly straight forward.config['PASSWORD']: error = 'Invalid password' else: session['logged_in'] = True flash('You were logged in') return redirect(url_for('show_entries')) return render_template('login. 39 . there are Flask extensions for the purpose of hashing passwords and veri- fying passwords against hashes. the method will delete the key from the dictionary if present or do nothing when that key is not in there. passwords should be both hashed and salted before being stored in a database or file. you will get an exception that Flask cannot find the templates. This means that unless you mark a value in the code with Markup or with the |safe filter in the template. removes that key from the session again.route('/logout') def logout(): session.html'. None) flash('You were logged out') return redirect(url_for('show_entries')) Security Note Passwords should never be stored in plain text in a production system.pop('logged_in'. If you plan to release a project based off this tutorial out into the world. 5. Jinja2 will ensure that special characters such as < or > are escaped with their XML equivalents. The templates are using Jinja2 syntax and have autoescaping enabled by default.9 Step 7: The Templates Now it is time to start working on the templates. error=error) The logout function. You can find a list of recommended Flask extensions here Continue with Step 7: The Templates. This tutorial uses plain text passwords for simplicity. error = 'Invalid username' elif request. There are also many general python libraries that can be used for hashing. There is a neat trick here: if you use the pop() method of the dict and pass a second parameter to it (the default). Fortunately. if you make requests with the app running. on the other hand. As you may have noticed.form['password'] != app. This is helpful because now it is not necessary to check if the user was logged in. We are also using template inheritance which makes it possible to reuse the layout of the website in all pages.

filename='style.1 layout.logged_in %} <a href="{{ url_for('login') }}">log in</a> {% else %} <a href="{{ url_for('logout') }}">log out</a> {% endif %} </div> {% for message in get_flashed_messages() %} <div class=flash>{{ message }}</div> {% endfor %} {% block body %}{% endblock %} </div> 5.9.html template from above to display the mes- sages.logged_in %} <form action="{{ url_for('add_entry') }}" method=post class=add-entry> <dl> <dt>Title: <dd><input type=text size=30 name=title> <dt>Text: 40 . Notice that the form is configured to to submit to the add_entry view function and use POST as HTTP method: {% extends "layout. The {% block body %} block can be replaced by a block of the same name (body) in a child template.css . the header and a link to log in (or log out if the user was already logged in). even if there is no 'logged_in' key in the session: <!doctype html> <title>Flaskr</title> <link rel=stylesheet type=text/css href="{{ url_for('static'.9.→') }}"> <div class=page> <h1>Flaskr</h1> <div class=metanav> {% if not session.Put the following templates into the templates folder: 5. It also displays the flashed messages if there are any. The session dict is available in the template as well and you can use that to check if the user is logged in or not. Note that the for loop iterates over the messages we passed in with the render_template() function.html" %} {% block body %} {% if session.2 show_entries. Note that in Jinja you can access missing attributes and items of objects / dicts which makes the following code work.html This template extends the layout.html This template contains the HTML skeleton.

5.9. margin: 0. <dd><textarea name=text rows=5 cols=40></textarea> <dd><input type=submit value=Share> </dl> </form> {% endif %} <ul class=entries> {% for entry in entries %} <li><h2>{{ entry.10 Step 8: Adding Style Now that everything else works.text|safe }} {% else %} <li><em>Unbelievable.2em. } 41 . } h2 { font-size: 1. No entries here so far</em> {% endfor %} </ul> {% endblock %} 5. which basically just displays a form to allow the user to login: {% extends "layout.css in the static folder: body { font-family: sans-serif. } h1 { border-bottom: 2px solid #eee. it’s time to add some style to the application. } h1.3 login.html This is the login template. Just create a stylesheet called style.title }}</h2>{{ entry. serif. h2 { color: #377ba8. background: #eee. h2 { font-family: 'Georgia'. h1. } a.html" %} {% block body %} <h2>Login</h2> {% if error %}<p class=error><strong>Error:</strong> {{ error }}{% endif %} <form action="{{ url_for('login') }}" method=post> <dl> <dt>Username: <dd><input type=text name=username> <dt>Password: <dd><input type=password name=password> <dd><input type=submit value=Login> </dl> </form> {% endblock %} Continue with Step 8: Adding Style.

padding: 0.11.1 Adding tests to flaskr Assuming you have seen the Testing Flask Applications section and have either written your own tests for flaskr or have followed along with the examples provided. } . background: white. } .entries { list-style: none. } Continue with Bonus: Testing the Application.. width: 35em.11. } .add-entry dl { font-weight: bold.page { margin: 2em auto.5em. 42 . One possible and recommended project structure is: flaskr/ flaskr/ __init__.8em. border-bottom: 1px solid #ccc.add-entry { font-size: 0.flash { background: #cee5F5.2 Running the tests At this point you can run the tests. padding: 0. padding: 0. } .9em. 5.11 Bonus: Testing the Application Now that you have finished the application and everything works as expected.entries li h2 { margin-left: -1em. border: 1px solid #aacbe2. border: 5px solid #ccc. The application above is used as a basic example of how to perform unit testing in the Testing Flask Applications section of the documentation.3em.in For now go ahead a create the tests/ directory as well as the test_flaskr. } .5em. it’s probably not a bad idea to add automated tests to simplify modifications in the future.8em. you might be wondering about ways to organize the project. Go there to see how easy it is to test Flask applications.py file. } .py static/ templates/ tests/ test_flaskr.metanav { text-align: right. margin-bottom: 1em. padding: 0. padding: 0.error { background: #f0d6d6. } .entries li { margin: 0. 5. 5.py MANIFEST.8em 1. font-size: 0. margin: 0. } .py setup.2em. background: #fafafa. Here pytest will be used.

run. (Recall the setup_requires ar- gument in setup. Go ahead and update the setup.py file to contain: from setuptools import setup setup( name='flaskr'.3 Testing + setuptools One way to handle testing is to integrate it with setuptools. pip install pytest Run and watch the tests pass. Otherwise pytest test will not be able to import the required components to test the application: pip install -e . 43 . packages=['flaskr']. as the setup.cfg.py file and creating a new file setup.py script has been called. include_package_data=True. One benefit of running the tests this way is that you do not have to install pytest.11. ) Now create setup. setup_requires=[ 'pytest-runner'.test 5.py) Following the standard rules of test-discovery your tests will be found.py): [aliases] test=pytest Now you can run: python setup. ].py test This calls on the alias created in setup.cfg which in turn runs pytest via pytest-runner. tests_require=[ 'pytest'. and hopefully pass. install_requires=[ 'flask'. ].Note: Make sure that pytest is installed in the same virtualenv as flaskr. ]. Here that requires adding a couple of lines to the setup.cfg in the project root (alongside setup. within the top-level flaskr/ directory as: py.

Integrating testing with setuptools is convenient because it is not necessary to actually download pytest or any other testing framework one might use. 44 .This is one possible way to run and manage testing. but there are other options such as nose. Here pytest is used.

45 .htm. head over to the official Jinja2 Template Documentation for more information. Changed in version 0.xhtml when using render_template(). • Flask inserts a couple of global functions and helpers into the Jinja2 context. • a template has the ability to opt in/out autoescaping with the {% autoescape %} tag.config) New in version 0. This requirement is necessary to enable rich extensions. 6.10: This is now always available. ad- ditionally to the values that are present by default. You are obviously free to use a different tem- plate engine. If you want information on the template engine’s syntax itself.html.1 Jinja Setup Unless customized. even in imported tem- plates. • autoescaping is enabled for all strings when using render_template_string().xml as well as . CHAPTER SIX TEMPLATES Flask leverages Jinja2 as template engine. . 6. .6. This section only gives a very quick introduction into how Jinja2 is integrated into Flask.2 Standard Context The following global variables are available within Jinja2 templates by default: config The current configuration object (flask. but you still have to install Jinja2 to run Flask itself. Jinja2 is configured by Flask as follows: • autoescaping is enabled for all templates ending in . An extension can depend on Jinja2 being present.

This is for example very helpful if you try to generate JavaScript on the fly. This variable is unavailable if the template was rendered without an active request context. that needs to access the request object you have two possibilities: 1. This is partially caused by performance considerations.html' import my_macro with context %} 6. Importing with context looks like this: {% from '_helpers.g). you import the macro “with context”.url_for() function. The difference is that by default these will not show up in the context of imported templates. Note that inside script tags no escaping must take place. session The current session object (flask. partially to keep things explicit.3 Standard Filters These filters are available in Jinja2 additionally to the filters provided by Jinja2 itself: tojson() This function converts the given object into JSON representation. get_flashed_messages() The flask. What does this mean for you? If you have a macro you want to import.10 if you intend to use it inside script tags: 46 . so make sure to disable escaping with |safe before Flask 0.session). This variable is unavail- able if the template was rendered without an active request context.request). or the attribute of the request object you are interested in. they are not global variables. This variable is unavailable if the template was rendered without an active request context. g The request-bound object for global variables (flask. you explicitly pass the request to the macro as parameter. url_for() The flask.get_flashed_messages() function. The Jinja Context Behavior These variables are added to the context of variables.request The current request object (flask. 2.

4 Controlling Autoescaping Autoescaping is the concept of automatically escaping special characters for you.template_filter('reverse') def reverse_filter(s): 47 .5 Registering Filters If you want to register your own filters in Jinja2 you have two ways to do that. Not doing so would not only cause user frustration by the inability to use these characters in text. and thus XHTML) are &. wrap the HTML string in a Markup object before passing it to the template. You can either put them by hand into the jinja_env of the application or use the template_filter() decorator. use the |safe filter to explicitly mark a string as safe HTML ({{ myvariable|safe }}) • Temporarily disable the autoescape system altogether. >. To disable the autoescape system in templates. (see Cross-Site Scripting (XSS)) Sometimes however you will need to disable autoescaping in templates. you can use the {% autoescape %} block: {% autoescape false %} <p>autoescaping is disabled here <p>{{ will_not_be_escaped }} {% endautoescape %} Whenever you do this. " as well as '. Spe- cial characters in the sense of HTML (or XML. but can also lead to security problems. • Inside the template. for example if they come from a system that generates secure HTML like a markdown to HTML converter. The two following examples work the same and both reverse an object: @app. 6. please be very cautious about the variables you are using in this block. <script type=text/javascript> doSomethingWith({{ user. There are three ways to accomplish that: • In the Python code. Because these characters carry specific meanings in documents on their own you have to replace them by so called “entities” if you want to use them for text. </script> 6. <.username|tojson|safe }}). This is in general the recommended way. This can be the case if you want to explicitly inject HTML into pages.

currency=u''): return u'{0:. for example if you have a Python list in context called mylist: {% for x in mylist | reverse %} {% endfor %} 6. Context processors run before the template is rendered and have the ability to inject new values into the template context. context processors exist in Flask.format(amount.context_processor def utility_processor(): def format_price(amount. but it gives an idea how this works.jinja_env. for all templates in the app: @app.2f}{1}'.filters['reverse'] = reverse_filter In case of the decorator the argument is optional if you want to use the function name as name of the filter.6 Context Processors To inject new variables automatically into the context of a template. return s[::-1] def reverse_filter(s): return s[::-1] app. Variables are not limited to values. A context processor is a function that returns a dictionary.user) The context processor above makes a variable called user available in the template with the value of g. 48 . a context processor can also make functions avail- able to templates (since Python allows passing around functions): @app.33) }} You could also build format_price as a template filter (see Registering Filters). you can use the filter in your templates in the same way as Jinja2’s builtin filters.context_processor def inject_user(): return dict(user=g. This example is not very interesting because g is available in templates anyways. but this demonstrates how to pass functions in a context processor. currency) return dict(format_price=format_price) The context processor above makes the format_price function available to all templates: {{ format_price(0. Once registered. The keys and values of this dictionary are then merged with the template context.user.

You can then use that with your favourite testing solution.app. 7.py) and create a unittest skeleton there: import os import flaskr import unittest import tempfile class FlaskrTestCase(unittest. The origin of this quote is unknown and while it is not entirely correct. Flask provides a way to test your application by exposing the Werkzeug test Client and handling the context locals for you. 7.config['DATABASE'] = tempfile. flaskr. it is also not far from the truth.mkstemp() flaskr. CHAPTER SEVEN TESTING FLASK APPLICATIONS Something that is untested is broken. you can safely make changes and instantly know if anything breaks. Untested applications make it hard to improve existing code and developers of untested applications tend to become pretty paranoid.app = flaskr.test_client() 49 . we will use the application from the Tutorial. In this documentation we will use the unittest package that comes pre-installed with Python.1 The Application First. we add a second module (flaskr_tests.config['TESTING'] = True self.db_fd. If you don’t have that application yet. get the sources from the examples.2 The Testing Skeleton In order to test the application.TestCase): def setUp(self): self.app.app. we need an application to test. If an application has automated tests.

app. we close the file and remove it from the filesystem in the tearDown() method.test_client() 50 .mkstemp() self. The mkstemp() function does two things for us: it returns a low-level file handle and a random file name. flaskr. we add a new test method to our class.config['DATABASE']) if __name__ == '__main__': unittest.config['DATABASE'] = tempfile. otherwise the import would have died with an exception.app_context(): flaskr. like this: class FlaskrTestCase(unittest. We just have to keep the db_fd around so that we can use the os.db_fd.close(self. What it does is disable the error catching during request handling so that you get bet- ter error reports when performing test requests against the application. If we now run the test suite. We can trigger test requests to the application. we should see the following output: $ python flaskr_tests.TestCase): def setUp(self): self. and the client will also keep track of cookies for us. This function is called before each individual test function is run. 7. To delete the database after the test.app. the latter we use as database name.main() The code in the setUp() method creates a new test client and initializes a new database.db_fd) os. Because SQLite3 is filesystem-based we can easily use the tempfile module to create a temporary database and initialize it.unlink(flaskr.app.app = flaskr. we already know that our flaskr applica- tion is syntactically valid.py ---------------------------------------------------------------------- Ran 0 tests in 0. This test client will give us a simple interface to the application.init_db() def tearDown(self): os.app.3 The First Test Now it’s time to start testing the functionality of the application. Additionally during setup the TESTING config flag is activated. To do this. close() function to close the file. with flaskr.000s OK Even though it did not run any actual tests. Let’s check that the application shows “No entries here so far” if we access the root of the application (/).

flaskr.init_db()

def tearDown(self):
os.close(self.db_fd)
os.unlink(flaskr.app.config['DATABASE'])

def test_empty_db(self):
rv = self.app.get('/')
assert b'No entries here so far' in rv.data

Notice that our test functions begin with the word test; this allows unittest to auto-
matically identify the method as a test to run.
By using self.app.get we can send an HTTP GET request to the application with the
given path. The return value will be a response_class object. We can now use the
data attribute to inspect the return value (as string) from the application. In this case,
we ensure that 'No entries here so far' is part of the output.
Run it again and you should see one passing test:

$ python flaskr_tests.py
.
----------------------------------------------------------------------
Ran 1 test in 0.034s

OK

7.4 Logging In and Out

The majority of the functionality of our application is only available for the adminis-
trative user, so we need a way to log our test client in and out of the application. To do
this, we fire some requests to the login and logout pages with the required form data
(username and password). And because the login and logout pages redirect, we tell
the client to follow_redirects.
Add the following two methods to your FlaskrTestCase class:

def login(self, username, password):
return self.app.post('/login', data=dict(
username=username,
password=password
), follow_redirects=True)

def logout(self):
return self.app.get('/logout', follow_redirects=True)

Now we can easily test that logging in and out works and that it fails with invalid
credentials. Add this new test to the class:

51

def test_login_logout(self):
rv = self.login('admin', 'default')
assert b'You were logged in' in rv.data
rv = self.logout()
assert b'You were logged out' in rv.data
rv = self.login('adminx', 'default')
assert b'Invalid username' in rv.data
rv = self.login('admin', 'defaultx')
assert b'Invalid password' in rv.data

7.5 Test Adding Messages

We should also test that adding messages works. Add a new test method like this:

def test_messages(self):
self.login('admin', 'default')
rv = self.app.post('/add', data=dict(
title='<Hello>',
text='<strong>HTML</strong> allowed here'
), follow_redirects=True)
assert b'No entries here so far' not in rv.data
assert b'&lt;Hello&gt;' in rv.data
assert b'<strong>HTML</strong> allowed here' in rv.data

Here we check that HTML is allowed in the text but not in the title, which is the in-
tended behavior.
Running that should now give us three passing tests:

$ python flaskr_tests.py
...
----------------------------------------------------------------------
Ran 3 tests in 0.332s

OK

For more complex tests with headers and status codes, check out the MiniTwit Exam-
ple from the sources which contains a larger test suite.

7.6 Other Testing Tricks

Besides using the test client as shown above, there is also the test_request_context()
method that can be used in combination with the with statement to activate a request
context temporarily. With this you can access the request, g and session objects like
in view functions. Here is a full example that demonstrates this approach:

52

import flask

app = flask.Flask(__name__)

with app.test_request_context('/?name=Peter'):
assert flask.request.path == '/'
assert flask.request.args['name'] == 'Peter'

All the other objects that are context bound can be used in the same way.
If you want to test your application with different configurations and there does not
seem to be a good way to do that, consider switching to application factories (see
Application Factories).
Note however that if you are using a test request context, the before_request()
and after_request() functions are not called automatically. However
teardown_request() functions are indeed executed when the test request con-
text leaves the with block. If you do want the before_request() functions to be called
as well, you need to call preprocess_request() yourself:
app = flask.Flask(__name__)

with app.test_request_context('/?name=Peter'):
app.preprocess_request()
...

This can be necessary to open database connections or something similar depending
on how your application was designed.
If you want to call the after_request() functions you need to call into
process_response() which however requires that you pass it a response object:
app = flask.Flask(__name__)

with app.test_request_context('/?name=Peter'):
resp = Response('...')
resp = app.process_response(resp)
...

This in general is less useful because at that point you can directly start using the test
client.

7.7 Faking Resources and Context

New in version 0.10.
A very common pattern is to store user authorization information and database con-
nections on the application context or the flask.g object. The general pattern for this is
to put the object on there on first usage and then to remove it on a teardown. Imagine
for instance this code to get the current user:

53

def get_user():
user = getattr(g, 'user', None)
if user is None:
user = fetch_current_user_from_database()
g.user = user
return user

For a test it would be nice to override this user from the outside without hav-
ing to change some code. This can be accomplished with hooking the flask.
appcontext_pushed signal:

from contextlib import contextmanager
from flask import appcontext_pushed, g

@contextmanager
def user_set(app, user):
def handler(sender, **kwargs):
g.user = user
with appcontext_pushed.connected_to(handler, app):
yield

And then to use it:

from flask import json, jsonify

@app.route('/users/me')
def users_me():
return jsonify(username=g.user.username)

with user_set(app, my_user):
with app.test_client() as c:
resp = c.get('/users/me')
data = json.loads(resp.data)
self.assert_equal(data['username'], my_user.username)

7.8 Keeping the Context Around

New in version 0.4.
Sometimes it is helpful to trigger a regular request but still keep the context around
for a little longer so that additional introspection can happen. With Flask 0.4 this is
possible by using the test_client() with a with block:

app = flask.Flask(__name__)

with app.test_client() as c:
rv = c.get('/?tequila=42')
assert request.args['tequila'] == '42'

54

If you were to use just the test_client() without the with block, the assert would
fail with an error because request is no longer available (because you are trying to use
it outside of the actual request).

7.9 Accessing and Modifying Sessions

New in version 0.8.
Sometimes it can be very helpful to access or modify the sessions from the test client.
Generally there are two ways for this. If you just want to ensure that a session has
certain keys set to certain values you can just keep the context around and access
flask.session:

with app.test_client() as c:
rv = c.get('/')
assert flask.session['foo'] == 42

This however does not make it possible to also modify the session or to access the ses-
sion before a request was fired. Starting with Flask 0.8 we provide a so called “session
transaction” which simulates the appropriate calls to open a session in the context of
the test client and to modify it. At the end of the transaction the session is stored. This
works independently of the session backend used:

with app.test_client() as c:
with c.session_transaction() as sess:
sess['a_key'] = 'a value'

# once this is reached the session was stored

Note that in this case you have to use the sess object instead of the flask.session
proxy. The object however itself will provide the same interface.

55

56

Even if your code is 100% correct. Here are some situations where perfectly fine code can lead to server errors: • the client terminated the request early and the application was still reading from the incoming data • the database server was overloaded and could not handle the query • a filesystem is full • a harddrive crashed • a backend server overloaded • a programming error in a library you are using • network connection of the server to another system failed And that’s just a small sample of issues you could be facing.1 Error Logging Tools Sending error mails. To use Sentry you need to install the raven client: 57 . Applications fail. Sooner or later you will see an exception in production. and we will cover some better setups to deal with errors. can become overwhelming if enough users are hitting the error and log files are typically never looked at. But there is more you can do. So how do we deal with that sort of problem? By default if your application runs in production mode. captures the full stack trace and local variables for debugging. CHAPTER EIGHT APPLICATION ERRORS New in version 0. and sends you mails based on new errors or frequency thresholds. even if just for critical ones. This is why we recommend using Sentry for dealing with application errors. servers fail. you will still see exceptions from time to time. Sentry aggregates duplicate errors. 8. Why? Because everything else involved will fail.3. It’s available as an Open Source project on GitHub and is also available as a hosted version which you can try for free. Flask will display a very simple page for you and log the exception to the logger.

2 Error handlers You might want to show custom error pages to the user when an error occurs..flask import Sentry sentry = Sentry(app.2.BadRequest) def handle_bad_request(e): return 'bad request!' app. 58 . 8.init_app(app) .$ pip install raven And then add this to your Flask app: from raven.register_error_handler(400.1 Registering Register error handlers using errorhandler() or register_error_handler(): @app. 8.exceptions. but the first one is more clear and leaves you with a function to call on your whim (and in tests).flask import Sentry sentry = Sentry(dsn='YOUR_DSN_HERE') def create_app(): app = Flask(__name__) sentry.exceptions. Note that werkzeug.contrib.errorhandler(werkzeug. Afterwards failures are automatically reported to Sentry and from there you can re- ceive error notifications.. lambda e: 'bad request!') Those two ways are equivalent. code == 400). Error handlers are normal Pluggable Views but instead of being registered for routes. they are registered for exceptions that are raised while trying to do something else. HTTPException subclasses like BadRequest from the example and their HTTP codes are interchangeable when handed to the registration methods or decorator (BadRequest.contrib. dsn='YOUR_DSN_HERE') Or if you are using factories you can also init it later: from raven. This can be done by registering error handlers. return app The YOUR_DSN_HERE value needs to be replaced with the DSN value you get from your Sentry installation.

11: Errorhandlers are now prioritized by specificity of the excep- tion classes they are registered for instead of the order they are registered in. 8. Flask uses the Python builtin logging system.0.0. 59 . and it can actually send you mails for errors which is probably what you want.handlers import SMTPHandler mail_handler = SMTPHandler('127. its class hierarchy is traversed. 8.debug: import logging from logging.com'] if not app. The reason for that is that Flask by default will just report to the WSGI error stream or stderr (depending on what’s available).g. and searched for in the exception classes for which handlers are registered.ERROR) app. The most specific handler is selected.addHandler(mail_handler) So what just happened? We created a new SMTPHandler that will send mails with the mail server listening on 127.You are however not limited to HTTPException or HTTP status codes but can register a handler for every exception class you like. these can also be provided. the more specific ConnectionRefusedError handler is called on the exception instance. Where this ends up is sometimes hard to find.3 Error Mails If the application runs in production mode (which it will do on your server) you might not see any log messages. 'server-error@example.com'. if an instance of ConnectionRefusedError is raised.2.logger. Here is how you can configure the Flask logger to send you mails for exceptions: ADMINS = ['yourname@example. 'YourApplication Failed') mail_handler. E. and a handler is registered for ConnectionError and ConnectionRefusedError. What you probably want instead is a mail the second the exception happened. Changed in version 0. ADMINS. If your mail server re- quires credentials. Then you get an alert and you can do something about it.2 Handling Once an exception instance is raised.1 to all the ADMINS from the address server- error@example.1'.0. Often it’s in your webserver’s log files.com with the subject “YourApplication Failed”. and its response is shown to the user. I can pretty much promise you however that if you only use a logfile for the application errors you will never look at it except for debugging an issue when a user reported it for you. For that check out the documentation for the SMTPHandler.setLevel(logging.0.

11. errors are logged to your webserver’s log automatically. so it’s your responsibility to warn in the code if something seems odd. 60 . and more importantly..5 Controlling the Log Format By default a handler will only write the message string into a file or send you that message as mail. • SysLogHandler .setLevel(logging. Please note that Flask itself will not issue any warnings in the core system.logger. just make sure to use a lower setting (I would recommend WARNING): if not app. You don’t have to do that in the log formatter format string.) file_handler.debug: import logging from themodule import TheHandlerYouWant file_handler = TheHandlerYouWant(. Before you run that in production. • RotatingFileHandler .will log to the system event log of a Windows system. Because we certainly don’t want to get a mail for warnings or other useless logs that might happen during request handling. By default as of Flask 0. • NTEventLogHandler .We also tell the handler to only send errors and more critical messages. 8.. you probably also want to log warnings.WARNING) app. A log record stores more information. Once you picked your log handler. where it did.logs messages to a file on the filesystem. this is what you want to use. The most interesting are probably the following: • FileHandler .addHandler(file_handler) 8.sends logs to a UNIX syslog. and it makes a lot of sense to configure your logger to also contain that information so that you have a better idea of why that error happened.logs messages to a file on the filesystem and will rotate after a certain number of messages. That will save you from a lot of frustration. Note that tracebacks are ap- pended to the log entry automatically. do like you did with the SMTP handler above. please also look at Controlling the Log Format to put more information into that error mail. Warnings however are not. If you are deploying on a Windows box. A formatter can be instantiated with a format string.4 Logging to a File Even if you get mails. There are a couple of handlers provided by the logging system out of the box but not all of them are useful for basic error logging. It’s a good idea to keep as much information around that might be required to debug a problem.

61 .5.setFormatter(Formatter(''' Message type: %(levelname)s Location: %(pathname)s:%(lineno)d Module: %(module)s Function: %(funcName)s Time: %(asctime)s Message: %(message)s ''')) 8.5.2 File logging from logging import Formatter file_handler.Here are some example setups: 8. consult the official documentation of the logging package for a full list. Note that this list is not complete.3 Complex Log Formatting Here is a list of useful formatting variables for the format string.5.1 Email from logging import Formatter mail_handler.setFormatter(Formatter( '%(asctime)s %(levelname)s: %(message)s ' '[in %(pathname)s:%(lineno)d]' )) 8.

By de- fault this is of the form "2003-07-08 16:49:45.addHandler(mail_handler) logger. While there is a method to configure all loggers at once in the logging package. formatTime(): called for asctime formatting. 'CRITICAL'). Instead. I would recommend figuring out which loggers you are interested in. %(funcName)s Name of function containing the logging call. computed as msg % args If you want to further customize the formatting. %(asctime)s Human-readable time when the LogRecord was created.logger. you don’t have to override it. getLogger('otherlibrary')] for logger in loggers: logger. %(message)s The logged message. 'INFO'.896" (the num- bers after the comma are millisecond portion of the time). 'WARNING'. 8. For more information. Other libraries might log themselves as well.6 Other Libraries So far we only configured the logger your application created itself. %(pathname)s Full pathname of the source file where the logging call was issued (if available). you can subclass the formatter. It is passed a LogRecord object and has to return the formatted string. getLogger('sqlalchemy'). It is passed an exc_info tuple and has to return a string. formatException() called for exception formatting. There might be a situation in which you want to have multiple separate applications running side by side in the same Python interpreter and then it becomes impossible to have different logging setups for those. The default is usually fine. getting the loggers with the getLogger() function and iterating over them to attach handlers: from logging import getLogger loggers = [app. head over to the official documentation. SQLAlchemy uses logging heavily in its core. If you want a different time format you can override this method. 'ERROR'.addHandler(file_handler) 62 . This can be changed by subclassing the formatter and overriding the formatTime() method. %(module)s Module (name portion of filename). Format Description %(levelname)s Text logging level for the message ('DEBUG'. For example. %(lineno)d Source line number where the logging call was issued (if avail- able). %(filename)s Filename portion of pathname. The formatter has three interesting methods: format(): handles the actual formatting. I would not recommend using it.

but be sure to do this temporarily in a controlled environment.e. Be sure to run under the same user account as the configured deployment to troubleshoot permission issues. If you would like to use another Python debugger.whether to enable debug mode and catch exceptions • use_debugger . Flask provides a debugger out of the box (see Debug Mode).. This section provides pointers when debugging deployment configuration and digging deeper with a full-featured Python debugger. Run Manually Having problems getting your application configured for production? If you have shell access to your host.2 Working with Debuggers To dig deeper. which is helpful in catching configuration issues.yaml (change the block as appropriate for your application.1 When in Doubt. Do not run in production with debug=True. of course): 63 . configure your application with logging and notifications as described in Application Errors. possibly to trace code execution.whether to reload and fork the process on exception debug must be True (i. note that debuggers interfere with each other. verify that you can run your application manually from the shell in the deployment environment. 9. You have to set some options in order to use your favorite debugger: • debug . If you’re using Aptana/Eclipse for debugging you’ll need to set both use_debugger and use_reloader to False. You can use Flask’s builtin development server with debug=True on your production host. 9. A possible useful pattern for configuration is to set the following in your config.whether to use the internal Flask debugger • use_reloader . exceptions must be caught) in order for the other two options to have any value. CHAPTER NINE DEBUGGING APPLICATION ERRORS For production applications.

use_reloader=use_debugger. set use_debugger=False app = create_app(config="config.py).0. debug=app. you could have something like: if __name__ == "__main__": # To allow aptana to receive errors.debug.run(use_debugger=use_debugger.0') 64 .config.FLASK: DEBUG: True DEBUG_WITH_APTANA: True Then in your application’s entry-point (main.yaml") if app. host='0.get('DEBUG_WITH_APTANA')) except: pass app.debug: use_debugger = True try: # Disable Flask's debugger if external debugger is requested use_debugger = not(app.0.

But this is also where you can have your own configuration. The way Flask is designed usually requires the configuration to be available when the application starts up. but there are better ways. There are different settings you might want to change depending on the application environment like toggling the debug mode. You can hardcode the configuration in the code.' ) 65 .config['DEBUG'] = True Certain configuration values are also forwarded to the Flask object so you can read and write them from there: app.config.3. Independent of how you load your config. setting the secret key. This is the place where Flask itself puts certain configuration values and also where exten- sions can put their configuration values.debug = True To update multiple keys at once you can use the dict. Applications need some kind of configuration. and other such environment-specific things.. which for many small applications is not actually that bad. there is a config object available which holds the loaded configuration values: The config attribute of the Flask object.1 Configuration Basics The config is actually a subclass of a dictionary and can be modified just like any dictionary: app = Flask(__name__) app.update( DEBUG=True. SECRET_KEY='. CHAPTER TEN CONFIGURATION HANDLING New in version 0.update() method: app. 10..

10.2 Builtin Configuration Values The following configuration values are used internally by Flask: 66 .

dev:5000') Note that localhost does not sup- port subdomains so setting this to “localhost” does not help. SESSION_COOKIE_HTTPONLY controls if the cookie should be set with the httponly flag. SERVER_NAME the name and port number of the server. Defaults to True.: 'myapp. Re- quired for subdomain support (e. If not set or explicitly set to None this is implicitly true if either TESTING or DEBUG is true. If this is not set. Non permanent sessions are not affected by this.8 this can also be an integer represent- ing seconds. USE_X_SENDFILE enable/disable x-sendfile LOGGER_NAME the name of the logger LOGGER_HANDLER_POLICY the policy of the default logging handler. APPLICATION_ROOT If the application does not occupy a whole do- . SECRET_KEY the secret key SESSION_COOKIE_NAME the name of the session cookie SESSION_COOKIE_DOMAIN the domain for the session cookie. SESSION_COOKIE_SECURE controls if the cookie should be set with the secure flag. Starting with Flask 0. If this is not set the cookie will be valid for all of APPLICATION_ROOT or if that is not set for '/'. the cookie will be valid for all subdomains of SERVER_NAME. PRESERVE_CONTEXT_ON_EXCEPTION By default if the application is in debug mode the request context is not popped on excep- tions to enable debuggers to introspect the data. You can also use this setting to force-enable it for non debug execution which might be useful to debug production applications (but also very risky). 'production' will only log in produc- tion and 'never' disables it entirely. The default is 'always' which means that the default logging handler is always active.timedelta object. This can be disabled by this key. If set to True (which is the de- fault) then the cookie is refreshed each request which automatically bumps the lifetime.DEBUG enable/disable debug mode TESTING enable/disable testing mode PROPAGATE_EXCEPTIONS explicitly enable or disable the propagation of exceptions. Setting a SERVER_NAME also by67 default enables URL generation without a re- quest context but with an application context. PERMANENT_SESSION_LIFETIME the lifetime of a permanent session as datetime. 'debug' will only activate logging in debug mode. Defaults to False. If set to False a set-cookie header is only sent if the session is modified.g. SESSION_REFRESH_EACH_REQUEST this flag controls how permanent sessions are refreshed. SESSION_COOKIE_PATH the path for the session cookie.

default_settings') app. This is also used for the session cookie. SESSION_COOKIE_SECURE New in version 0. JSON_SORT_KEYS.11: SESSION_REFRESH_EACH_REQUEST.config.9: PREFERRED_URL_SCHEME New in version 0.More on SERVER_NAME The SERVER_NAME key is used for the subdomain support.from_envvar('YOURAPPLICATION_SETTINGS') This first loads the configuration from the yourapplication. SESSION_COOKIE_HTTPONLY.default_settings module and then overrides the values with the contents of the file the YOURAPPLICATION_SETTINGS environment variable points to.8: TRAP_BAD_REQUEST_ERRORS. So if your server name is 'localhost' you will not be able to set a cookie for 'localhost' and every subdomain of it. like 'myapplication. LOGGER_HANDLER_POLICY. Most modern web browsers will not allow cross-subdomain cookies to be set on a server name without dots in it. Please keep in mind that not only Flask has the problem of not knowing what sub- domains are. This makes packaging and distributing your application possible via various package handling tools (Deploying with Setuptools) and finally modifying the configuration file afterwards.from_object('yourapplication.local' and add this name + the subdomains you want to use into your host config or setup a local bind. Please choose a different server name in that case. SESSION_COOKIE_DOMAIN.7: PROPAGATE_EXCEPTIONS. TEMPLATES_AUTO_RELOAD. this is required if you want to work with subdomains. EXPLAIN_TEMPLATE_LOADING 10. JSONIFY_PRETTYPRINT_REGULAR New in version 0. This environment variable can be set on Linux or OS X with the export command in the shell before starting the server: 68 . Because Flask cannot guess the subdomain part without the knowledge of the actual server name. SESSION_COOKIE_PATH.config.6: MAX_CONTENT_LENGTH New in version 0.10: JSON_AS_ASCII. APPLICATION_ROOT. ideally located outside the actual application package. New in version 0.3 Configuring from Files Configuration becomes more useful if you can store it in a separate file. TRAP_HTTP_EXCEPTIONS.5: SERVER_NAME New in version 0. your web browser does as well. PRESERVE_CONTEXT_ON_EXCEPTION New in version 0. So a common pattern is this: app = Flask(__name__) app.4: LOGGER_NAME New in version 0.

cfg $ python run-app. and a separate configuration that overrides the values as necessary as mentioned in the example above: 69 . 10.cfg The configuration files themselves are actual Python files. For a complete reference. The easiest way to handle this is to use a default configuration that is always loaded and part of the version control. You can use this to pass in configuration as needed. So make sure to use uppercase letters for your config keys.4 Configuration Best Practices The downside with the approach mentioned earlier is that it makes testing a little harder. Create your application in a function and register blueprints on it. 10. There should be at least separate configurations for the production server and the one used during development.$ export YOURAPPLICATION_SETTINGS=/path/to/settings. On Windows systems use the set builtin instead: >set YOURAPPLICATION_SETTINGS=\path\to\settings. There is no single 100% solution for this problem in general. Only values in uppercase are actually stored in the config object later on..1:5000/ * Restarting with reloader. read the Config object’s documentation.5 Development / Production Most applications need more than one configuration. but there are a couple of things you can keep in mind to improve that experience: 1. Here is an example of a configuration file: # Example configuration DEBUG = False SECRET_KEY = '?\xbf. That way you can create multiple instances of your application with different configura- tions attached which makes unittesting a lot easier.w\xee\x8d$0\x13\x8b83' Make sure to load the configuration very early on.py * Running on http://127.. 2. If you limit your- self to request-only accesses to the configuration you can reconfigure the object later on as needed.\xb4\x8d\xa3"<\x9c\xb0@\x0f5\xab. so that extensions have the ability to access the configuration when starting up.0. There are other methods on the config object as well to load from individual files.0. Do not write code that needs the configuration at import time.

Either populate the config with this default configuration or import it in your own configuration files before overriding values.config.py and you are done. What is very popular in the Django world is to make the import explicit in the config file by adding from yourapplication. If you are working often on dif- ferent projects you can even create your own script for sourcing that activates a virtualenv and exports the development configuration for you.default_settings') app.app = Flask(__name__) app.ProductionConfig') There are many different ways and it’s up to you how you want to manage your con- figuration files.from_object('configmodule. For some details about how to do that. • Use a tool like fabric in production to push code and configurations separately to the production server(s).py file and export YOURAPPLICATION_SETTINGS=/path/to/config. You could also inspect an environment vari- able like YOURAPPLICATION_MODE and set that to production. 70 . An interesting pattern is also to use classes and inheritance for configuration: class Config(object): DEBUG = False TESTING = False DATABASE_URI = 'sqlite://:memory:' class ProductionConfig(Config): DATABASE_URI = 'mysql://user@localhost/foo' class DevelopmentConfig(Config): DEBUG = True class TestingConfig(Config): TESTING = True To enable such a config you just have to call into from_object(): app.from_envvar('YOURAPPLICATION_SETTINGS') Then you just have to add a separate config.from_object('yourapplication.config. However here a list of good recommendations: • Keep a default configuration in version control. For example you could use imports or subclassing.config. However there are alternative ways as well. • Use an environment variable to switch between the configurations. head over to the Deploying with Fabric pattern. This can be done from outside the Python interpreter and makes development and deploy- ment much easier because you can quickly and easily switch between different configs without having to touch the code at all. development etc and import different hardcoded files based on that.default_settings import * to the top of the file and then overriding the changes by hand.

Flask 0. The behavior of relative paths in config files can be flipped between “relative to the application root” (the default) to “relative to instance folder” via the instance_relative_config switch to the application constructor: 71 . It refers to a new concept called the “instance folder”.8 introduces instance folders.10.instance_path. If the instance_path parameter is not provided the following default locations are used: • Uninstalled module: /myapp.8 a new attribute was introduced: Flask. For explicit con- figuration use the instance_path parameter: app = Flask(__name__. Since the config object provided loading of configuration files from relative filenames we made it possible to change the loading via filenames to be relative to the instance path if wanted. You can print the value of sys.8.py /instance • Installed module or package: $PREFIX/lib/python2. Flask for a long time made it possible to refer to paths relative to the application’s folder directly (via Flask.X/site-packages/myapp $PREFIX/var/myapp-instance $PREFIX is the prefix of your Python installation.py /instance • Uninstalled package: /myapp /__init__.root_path). instance_path='/path/to/instance/folder') Please keep in mind that this path must be absolute when provided. With Flask 0. Unfortu- nately however this only works well if applications are not packages in which case the root path refers to the contents of the package. This was also how many developers loaded configurations stored next to the application.prefix to see what the prefix is set to. The instance folder is designed to not be under version control and be deployment specific. You can either explicitly provide the path of the instance folder when creating the Flask application or you can let Flask autodetect the instance folder. It’s the perfect place to drop things that either change at runtime or configuration files. This can be /usr or the path to your virtualenv.6 Instance Folders New in version 0.

default_settings') app.read() # or via open_instance_resource: with app.instance_path.instance_path.from_pyfile('application.from_object('yourapplication. silent=True) The path to the instance folder can be found via the Flask.cfg') with open(filename) as f: config = f. instance_relative_config=True) Here is a full example of how to configure Flask to preload the config from a module and then override the config from a file in the config folder if it exists: app = Flask(__name__.config.cfg') as f: config = f. Example usage for both: filename = os.app = Flask(__name__. Flask also provides a shortcut to open a file from the instance folder with Flask. 'application.read() 72 .config.path. open_instance_resource(). instance_relative_config=True) app.open_instance_resource('application.cfg'.join(app.

be sure to also provide a sender unless you really want to listen for signals from all applications. 11. you can use the disconnect() method. Flask comes with a couple of signals and other extensions might provide more. This is especially true if you are developing an extension. CHAPTER ELEVEN SIGNALS New in version 0. When you subscribe to a signal. The big advantage of signals over handlers is that you can safely subscribe to them for just a split second. In contrast all signal handlers are executed in undefined order and do not modify any data. Say you want to know what templates were rendered as part of a request: signals allow you to do exactly that. signals allow certain senders to notify subscribers that something happened. What are signals? Signals help you decouple applications by sending notifications when actions occur elsewhere in the core framework or another Flask extensions. Also keep in mind that signals are intended to notify subscribers and should not encourage subscribers to modify data.1 Subscribing to Signals To subscribe to a signal. However. These temporary subscriptions are helpful for unittesting for example. You will notice that there are signals that appear to do the same thing like some of the builtin decorators do (eg: request_started is very similar to before_request()). In short. the sender is the application that issued the signal. The first ar- gument is the function that should be called when the signal is emitted.6. for example. the optional second argument specifies a sender.6. there are differences in how they work. there is integrated support for signalling in Flask. This support is provided by the excellent blinker library and will gracefully fall back if it is not available. The core before_request() handler. 73 . is executed in a specific order and is able to abort the request early by returning a response. For all core Flask signals. you can use the connect() method of a signal. Starting with Flask 0. To unsubscribe from a signal.

context = templates[0] 74 . the template object as well as context are appended to it. context = templates[0] assert template. context): recorded. recorded.connected_to(record.name == 'index. Additionally there is a convenient helper method (connected_to()) that allows you to temporarily subscribe a function to a signal with a context manager on its own. **extra): . Whenever a template is rendered. here is a helper context manager that can be used in a unittest to deter- mine which templates were rendered and what variables were passed to the template: from flask import template_rendered from contextlib import contextmanager @contextmanager def captured_templates(app): recorded = [] def record(sender. context)) template_rendered.html' assert len(context['items']) == 10 Make sure to subscribe with an extra **extra argument so that your calls don’t fail if Flask introduces new arguments to the signals.append((template. **extra): def record(sender. context)) return template_rendered. app) This can now easily be paired with a test client: with captured_templates(app) as templates: rv = app. you have to pass the list in as an argument: from flask import template_rendered def captured_templates(app. templates.append((template. context. template.get('/') assert rv. **extra): recorded.connect(record. All the template rendering in the code issued by the application app in the body of the with block will now be recorded in the templates variable. template.. app) try: yield recorded finally: template_rendered.disconnect(record.For example. template. app) The example above would then look like this: templates = [] with captured_templates(app..test_client(). Because the return value of the context manager cannot be specified that way.status_code == 200 assert len(templates) == 1 template.

signals. Passing Proxies as Senders 75 .Namespace class. 11. you can do so by calling the send() method.send(self) Try to always pick a good sender._get_current_object() as sender. For Extension Developers If you are writing a Flask extension and you want to gracefully degrade for missing blinker installations. If you are emitting a signal from a random function. It accepts a sender as first argument and optionally some keyword arguments that are forwarded to the signal subscribers: class Model(object): . you can pass current_app. you can use the blinker library directly.1. If you have a class that is emitting a signal. This is what is recommended most of the time: from blinker import Namespace my_signals = Namespace() Now you can create new signals like this: model_saved = my_signals. The most common use case are named signals in a custom Namespace. pass self as sender..3 Sending Signals If you want to emit a signal.Blinker API Changes The connected_to() method arrived in Blinker with version 1.. def save(self): model_saved.2 Creating Signals If you want to use signals in your own application. 11.. you can do so by using the flask. You can access the name of the signal with the name attribute.signal('model-saved') The name for the signal here makes it unique and also simplifies debugging.

so you can rely on flask. _get_current_object() instead.connect_via(app) def when_template_rendered(sender.1 you can also easily subscribe to signals by using the new connect_via() decorator: from flask import template_rendered @template_rendered. 76 . context. 11.Never pass current_app as sender to a signal. Context-local vari- ables are consistently available between request_started and request_finished.g and others as needed.6 Core Signals Take a look at Signals for a list of all builtin signals. template. Use current_app.name. 11. **extra): print 'Template %s is rendered with %s' % (template. context) 11.4 Signals and Flask’s Request Context Signals fully support The Request Context when receiving signals. Note the limitations described in Send- ing Signals and the request_tearing_down signal.5 Decorator Based Signal Subscriptions With Blinker 1. The reason for this is that current_app is a proxy and not the real application object.

users=users) This is simple and flexible.query. This is where pluggable class-based views come into place.all() return render_template('users. But this by itself is not helpful.all() return render_template('users. As the first step to convert this into a class based view you would do this: from flask.views import View class ShowUsers(View): def dispatch_request(self): users = User.html'.1 Basic Principle Consider you have a function that loads a list of objects from the database and renders into a template: @app. so let’s refactor the code a bit: 77 .7 introduces pluggable views inspired by the generic views from Django which are based on classes instead of functions.as_view('show_users')) As you can see what you have to do is to create a subclass of flask. but if you want to provide this view in a generic fashion that can be adapted to other models and templates as well you might want more flex- ibility. objects=users) app.View and im- plement dispatch_request().html'.add_url_rule('/users/'.query.route('/users/') def show_users(page): users = User. 12. view_func=ShowUsers.views. The string you pass to that function is the name of the endpoint that view will then have. The main intention is that you can replace parts of the implementations and this way have customizable pluggable views. CHAPTER TWELVE PLUGGABLE VIEWS New in version 0. Flask 0.7. Then we have to convert that class into an actual view function by using the as_view() class method.

html')) 12.get_objects()} return self.render_template(context) class UserView(ListView): def get_template_name(self): return 'users. When you have a class-based view the question comes up what self points to.views import View class ListView(View): def get_template_name(self): raise NotImplementedError() def render_template(self.get_template_name(). The way this works is that whenever the request is dispatched a new instance of the class is created and the dispatch_request() method is called with the parameters from the URL rule. For instance you can write a class like this: class RenderTemplateView(View): def __init__(self. **context) def dispatch_request(self): context = {'objects': self.add_url_rule('/about'.all() This of course is not that helpful for such a small example. context): return render_template(self.html' def get_objects(self): return User. The class itself is instantiated with the parameters passed to the as_view() function. template_name): self. view_func=RenderTemplateView.template_name = template_name def dispatch_request(self): return render_template(self. but it’s good enough to explain the basic principle.as_view( 'about_page'.2 Method Hints Pluggable views are attached to the application like a regular function by either using route() or better add_url_rule().query. template_name='about.template_name) And then you can register it like this: app. In order to move that information to the class you can provide a methods attribute that has this information: 78 .from flask. That however also means that you would have to provide the names of the HTTP methods the view supports when you attach this.

12. Instead you either have to decorate the return value of as_view() by hand: def user_required(f): """Checks whether user is logged in or raises error 401.""" def decorator(*args.. **kwargs): if not g. 'POST'] def dispatch_request(self): if request..add_url_rule('/users/'.views import MethodView class UserAPI(MethodView): def get(self): users = User.3 Method Based Dispatching For RESTful APIs it’s especially helpful to execute a different function for each HTTP method.method == 'POST': .user: abort(401) return f(*args. view_func=UserAPI.form) . .. app.from_form_data(request. app.MethodView you can easily do that.add_url_rule('/myview'..all() .as_view('users')) That way you also don’t have to provide the methods attribute.4 Decorating Views Since the view class itself is not the view function that is added to the routing system it does not make much sense to decorate the class itself.query.class MyView(View): methods = ['GET'...as_view('myview')) 12.. With the flask.. view_func=MyView.views. def post(self): user = User. It’s automatically set based on the methods defined in the class. **kwargs) return decorator 79 . Each HTTP method maps to a function with the same name (just in lowercase): from flask.

view = user_required(UserAPI. Let’s assume for the moment the view would look like this: class UserAPI(MethodView): def get(self. user_id): # delete a single user pass 80 . user_id): if user_id is None: # return a list of users pass else: # expose a single user pass def post(self): # create a new user pass def delete(self. you will notice that the API will require different URL rules that go to the same method view most of the time.8 there is also an alternative way where you can specify a list of decorators to apply in the class declaration: class UserAPI(MethodView): decorators = [user_required] Due to the implicit self from the caller’s perspective you cannot use regular view dec- orators on the individual methods of the view however. view_func=view) Starting with Flask 0.add_url_rule('/users/'.5 Method Views for APIs Web APIs are often working very closely with HTTP verbs so it makes a lot of sense to implement such an API based on the MethodView.as_view('users')) app. 12. That said. keep this in mind. For instance consider that you are exposing a user object on the web: URL Method Description /users/ GET Gives a list of all users /users/ POST Creates a new user /users/<id> GET Shows a single user /users/<id> PUT Updates a single user /users/<id> DELETE Deletes a single user So how would you go about doing that with the MethodView? The trick is to take advantage of the fact that you can provide multiple rules to the same view.

add_url_rule('/users/'. view_func=view_func. pk_type. 'PUT'. view_func=user_view. methods=['GET'. '/users/'. endpoint. user_id): # update a single user pass So how do we hook this up with the routing system? By adding two rules and explic- itly mentioning the methods for each: user_view = UserAPI. view_func=view_func. methods=['POST'. methods=['GET'. methods=['GET'. 'PUT'. view_func=view_func. defaults={pk: None}.]) app.add_url_rule('/users/<int:user_id>'.]) app. pk).add_url_rule(url. pk='id'.add_url_rule('/users/'. view_func=user_view. methods=['POST'.as_view(endpoint) app. view_func=user_view. methods=['GET'. pk_type='int'): view_func = view.]) app. 'DELETE']) If you have a lot of APIs that look similar you can refactor that registration code: def register_api(view. 'user_api'. url. defaults={'user_id': None}.add_url_rule(url. 'DELETE']) register_api(UserAPI. def put(self.as_view('user_api') app.]) app.add_url_rule('%s<%s:%s>' % (url. pk='user_id') 81 .

82 .

Consider. One of the design ideas behind Flask is that there are two different “states” in which code is executed. during request handling.9. While the application is in this state a few assumptions are true: • the programmer can modify the application object safely.request and others) point to the current request. There is a third state which is sitting in between a little bit. The application context is what powers the current_app context local. • no request handling happened so far • you have to have a reference to the application object in order to modify it. for instance. Sometimes you are dealing with an application in a way that is similar to how you interact with applications during request handling. 13. or a command line application. just that there is no request active. that you’re sitting in an interactive Python shell and interacting with the application. and it implicitly ends when the first request comes in. CHAPTER THIRTEEN THE APPLICATION CONTEXT New in version 0. • any code can get hold of these objects at any time. there is no magic proxy that can give you a reference to the application object you’re currently creating or modifying. The application setup state in which the application implicitly is on the module level. a couple of other rules exist: • while a request is active. 83 .1 Purpose of the Application Context The main reason for the application’s context existence is that in the past a bunch of functionality was attached to the request context for lack of a better solution. the context local objects (flask. It starts when the Flask object is instantiated. Since one of the pillars of Flask’s design is that you can have more than one application in the same Python process. In contrast.

For more information about that.So how does the code find the “right” application? In the past we recommended pass- ing applications around explicitly. As such it is the perfect place to store database connection information and other things. you can ignore the existence of the application context unless you need it.app_context(): # within this block. As a result. assuming they pick a sufficiently unique name and should put their information there. see Flask Extension Development. but that caused issues with libraries that were not designed with that in mind. Since creating such a request context is an unnecessarily expensive operation in case there is no request around. which was bound to the current request’s application reference. current_app points to app.2 Creating an Application Context There are two ways to make an application context. a RuntimeError will be raised. 13. print current_app. The second way is the explicit way using the app_context() method: from flask import Flask. an application context will be created alongside if this is necessary._app_ctx_stack.3 Locality of the Context The application context is created and destroyed as necessary. RuntimeError: Working outside of application context. If no request context has been pushed and an application context has not been explic- itly set. instead of on the flask. It never moves between threads and it will not be shared between requests.g object which is reserved for user code. 84 . A common workaround for that problem was to use the current_app proxy later on. The first one is implicit: whenever a request context is pushed. 13. The internal stack object is called flask. current_app app = Flask(__name__) with app. Extensions are free to store additional information on the topmost level. This allows you to generate URLs even in the absence of a request.name The application context is also used by the url_for() function in case a SERVER_NAME was configured. the application context was introduced.

close() The first time get_db() is called the connection will be established. The most common usage is to split resource management into two parts: 1. For instance. None) if db is None: db = g. 2. None) if db is not None: db. '_database'. When storing things on the application context unique names should be chosen as this is a place that is shared between Flask applications and extensions._database = connect_to_database() return db @app. '_database'. This is an example that connects to a database: import sqlite3 from flask import g def get_db(): db = getattr(g. a context teardown based resource deallocation.13. an implicit resource caching on the context. Generally there would be a get_X() function that creates resource X if it does not ex- ist yet and otherwise returns the same resource.4 Context Usage The context is typically used to cache resources that need to be created on a per-request or usage case.local import LocalProxy db = LocalProxy(get_db) That way a user can directly access db which internally calls get_db(). To make this im- plicit a LocalProxy can be used: from werkzeug.teardown_appcontext def teardown_db(exception): db = getattr(g. database connections are destined to go there. and a teardown_X() function that is registered as teardown handler. 85 .

86 .

line 1. CHAPTER FOURTEEN THE REQUEST CONTEXT This document describes the behavior in Flask 0.1 Diving into Context Locals Say you have a utility function that returns the URL the user should be redirected to. If you try to run this from a plain Python shell.args. in <module> AttributeError: 'NoneType' object has no attribute 'request' That makes a lot of sense because we currently do not have a request we could access. It is recommended that you read the The Application Context chapter first.com/') This context can be used in two ways. this is the exception you will see: >>> redirect_url() Traceback (most recent call last): File "<stdin>". it accesses the request object.get('next') or \ request.test_request_context('/?next=http://example. So we have to make a request and bind it to the current context.7 which is mostly in line with the old behavior but has some small. Either with the with statement or by calling the push() and pop() methods: 87 .referrer or \ url_for('index') As you can see. Imagine it would always redirect to the URL’s next parameter or the HTTP referrer or the index page: from flask import request. subtle differences. The test_request_context method can create us a RequestContext: >>> ctx = app. 14. url_for def redirect_url(): return request.

For more information of how to utilize the request context from the interactive Python shell. start_response) The method request_context() returns a new RequestContext object and uses it in combination with the with statement to bind the context.handle_exception(e)) return response(environ.request_context(environ): try: response = self. pop() removes it from the stack again. 14.make_response(self. head over to the Working with the Shell chapter. The request context internally works like a stack: The topmost level on the stack is the current active request. you will find a piece of code that looks very much like this: def wsgi_app(self. On popping the application’s teardown_request() functions are also executed. 88 .pop() Because the request context is internally maintained as a stack you can push and pop multiple times. environ): with self.>>> ctx. Everything that is called from the same thread from this point onwards until the end of the with statement will have access to the request globals (flask.com/' Until you call pop: >>> ctx. Another thing of note is that the request context will automatically also create an appli- cation context when it’s pushed and there is no application context for that application so far. push() adds the context to the stack on the very top. This is very handy to implement things like internal redirects.request and others).push() From that point onwards you can work with the request object: >>> redirect_url() u'http://example.full_dispatch_request() except Exception as e: response = self.2 How the Context Works If you look into how the Flask WSGI application internally works.

3 Callbacks and Errors What happens if an error occurs in Flask during request processing? This particular behavior changed in 0.7 because we wanted to make it easier to understand what is actually happening. When the request context is popped. the teardown_request() functions are called. At the end of the request the teardown_request() functions are executed. This always happens. An important change in 0. 4. before_request() functions are executed.get('/foo') # the teardown functions are still not called at that point 89 .test_client() as client: resp = client. In development mode however the exception is not further processed and bubbles up to the WSGI server.7 is that the internal server error is now no longer post processed by the after request callbacks and after request callbacks are no longer guar- anteed to be executed. This is important to know if the life of the request context is prolonged by using the test client in a with statement or when using the request context from the command line: with app. The return value of the view is then converted into an actual response object and handed over to the after_request() functions which have the chance to replace it or modify it in place. This way the internal dispatching code looks cleaner and is easier to customize and understand. Strictly speaking they are independent of the actual request handling as they are bound to the lifecycle of the RequestContext object.4 Teardown Callbacks The teardown callbacks are special callbacks in that they are executed at a different point. The new teardown functions are supposed to be used as a replacement for things that absolutely need to happen at the end of request. even in case of an unhandled exception down the road or if a before-request handler was not executed yet or at all (for example in test envi- ronments sometimes you might want to not execute before-request callbacks). 14. the other functions are no longer called. If the before_request() functions did not return a response. 3. the regular request handling kicks in and the view function that was matched has the chance to return a response. the 500 internal server handler is called. The new behavior is quite simple: 1. Before each request. 2. That way things like the interactive debugger can provide helpful debug information. In any case however the return value is treated as a replacement for the view’s return value. If one of these functions return a response. Now what happens on errors? In production mode if an exception is not caught.14.

but there are some exceptions where it is good to know that this object is an actual proxy: • The proxy objects do not fake their inherited types. 14.push() >>> ctx. • if the object reference is important (so for example for sending Signals) If you need to get access to the underlying object that is proxied.send(app) 90 .test_request_context() >>> ctx. print 'this runs after request' . you have to do that on the instance that is being proxied (see _get_current_object below).teardown_request . >>> ctx = app. Most of the time you don’t have to care about that. Certain parts of the test sys- tem might also temporarily create a request context without calling the before-request handlers. so if you want to perform actual instance checks.. you can use the _get_current_object() method: app = current_app. The reason behind this is that these proxies are shared between threads and they have to dispatch to the actual object bound to a thread behind the scenes as necessary. # even though the response ended and you have the response # object in your hand # only when the code reaches this point the teardown functions # are called. even if before-request call- backs were not executed yet but an exception happened.. Make sure to write your teardown-request handlers in a way that they will never fail..5 Notes On Proxies Some of the objects provided by Flask are proxies to other objects.pop() this runs after request >>> Keep in mind that teardown callbacks are always executed. def teardown_request(exception=None): .. Alternatively the same thing happens if another # request was triggered from the test client It’s easy to see the behavior from the command line: >>> app = Flask(__name__) >>> @app..._get_current_object() my_signal.

Starting with Flask 0.14. By default it’s linked to the setting of DEBUG. at the end of the request the request context is popped and all data associated with it is destroyed. In Flask 0. If the application is in debug mode the context is preserved. the request context was not popped so that the interactive debugger can still provide you with important information. During development however that can be problematic as you might want to have the information around for a longer time in case an exception occurred. in pro- duction mode it’s not. if an exception occurred. 91 .6 and earlier in debug mode.6 Context Preservation on Error If an error occurs or not. However it can be useful during development to get the same error preserving behavior as in development mode when attempting to debug an error that only occurs under production settings. Do not force activate PRESERVE_CONTEXT_ON_EXCEPTION in production mode as it will cause your application to leak memory on exceptions.7 you have finer control over that behavior by setting the PRESERVE_CONTEXT_ON_EXCEPTION configuration variable.

92 .

• Register a blueprint multiple times on an application with different URL rules. and other utilities through blueprints. but it is not actually an application. This is ideal for larger applications. • Register a blueprint on an application at a URL prefix and/or subdomain. static files. A blueprint does not have to implement applications or view func- tions. • Register a blueprint on an application for any of these cases when initializing a Flask extension. even multiple times. CHAPTER FIFTEEN MODULAR APPLICATIONS WITH BLUEPRINTS New in version 0. Why not have multiple application objects? You can do that (see Application Dispatching). 93 . and register a collection of blueprints. templates. 15. • Provide template filters.7. Pa- rameters in the URL prefix/subdomain become common view arguments (with defaults) across all view functions in the blueprint.1 Why Blueprints? Blueprints in Flask are intended for these cases: • Factor an application into a set of blueprints. a project could instantiate an application object. Flask uses a concept of blueprints for making application components and supporting common patterns within an application or across applications. but your applications will have separate configs and will be managed at the WSGI layer. Blueprints can greatly simplify how large applications work and provide a central means for Flask extensions to register operations on applications. Rather it is a blueprint of how to construct or extend an application. A blueprint in Flask is not a pluggable app because it is not actually an application – it’s a set of operations which can be registered on an application. A Blueprint object works similarly to a Flask application object. initialize several extensions.

4 Registering Blueprints So how do you register that blueprint? Like this: from flask import Flask from yourapplication. abort from jinja2 import TemplateNotFound simple_page = Blueprint('simple_page'. share application config. 15. 15. The downside is that you cannot unregister a blueprint once an application was created without having to destroy the whole application object. template_folder='templates') @simple_page. defaults={'page': 'index'}) @simple_page.Blueprints instead provide separation at the Flask level. 15. Additionally it will prefix the endpoint of the function with the name of the blueprint which was given to the Blueprint constructor (in this case also simple_page).2 The Concept of Blueprints The basic concept of blueprints is that they record operations to execute when reg- istered on an application.route('/'.route('/<page>') def show(page): try: return render_template('pages/%s. and can change an application object as necessary with being registered.route decorator the blueprint will record the intention of registering the function show on the application when it’s later registered. __name__. In this case we want to implement a blueprint that does simple rendering of static templates: from flask import Blueprint. render_template.3 My First Blueprint This is what a very basic blueprint looks like. Flask associates view functions with blueprints when dis- patching requests and generating URLs from one endpoint to another.html' % page) except TemplateNotFound: abort(404) When you bind a function with the help of the @simple_page.simple_page import simple_page app = Flask(__name__) app.register_blueprint(simple_page) 94 .

these are the generated rules: [<Rule '/static/<filename>' (HEAD. OPTIONS.show>. 15. If it points to an actual Python package that package (which is a folder on the filesystem) is the resource folder. OPTIONS. Blueprints however can also be mounted at different locations: app. <Rule '/pages/' (HEAD. The other two are for the show function of the simple_page blueprint. OPTIONS. As you can see. While multiple blueprints can originate from the same folder. GET) -> simple_page. 15. GET) -> static>.root_path property to see what the resource folder is: >>> simple_page. If it’s a module.1 Blueprint Resource Folder Like for regular applications.). GET) -> simple_page.If you check the rules registered on the application.show>] On top of that you can register blueprints multiple times though not every blueprint might respond properly to that. <Rule '/' (HEAD. GET) -> simple_page.5 Blueprint Resources Blueprints can provide resources as well. OPTIONS.root_path '/Users/username/TestProject/yourapplication' To quickly open sources from this folder you can use the open_resource() function: 95 . This argument specifies what logical Python module or package corre- sponds to the blueprint. GET) -> static>. OPTIONS. you will find these: [<Rule '/static/<filename>' (HEAD. The folder is inferred from the second argument to Blueprint which is usually __name__. it does not have to be the case and it’s usually not recommended. blueprints are considered to be contained in a folder.show>] The first one is obviously from the application itself for the static files. url_prefix='/pages') And sure enough. GET) -> simple_page.5. OPTIONS.register_blueprint(simple_page. they are also prefixed with the name of the blueprint and separated by a dot (. In fact it depends on how the blueprint is imple- mented if it can be mounted more than once.show>. Sometimes you might want to introduce a blueprint only for the resources it provides. <Rule '/pages/<page>' (HEAD. <Rule '/<page>' (HEAD. the package the mod- ule is contained in will be the resource folder. You can access the Blueprint.

Say the blueprint is registered for /admin the static folder will be at /admin/ static. Because the folder is called static here it will be available at the location of the blueprint + /static. The template folder is added to the search path of templates but with a lower prior- ity than the actual application’s template folder. The endpoint is named blueprint_name.5. template_folder='templates') For static files. The reason for the extra admin folder is to avoid getting our tem- plate overridden by a template named index. the best idea is to lay out your templates like this: 96 . __name__. the path can be absolute or relative to the blueprint resource folder. __name__.css') 15. static_folder='static') By default the rightmost part of the path is where it is exposed on the web. To further reiterate this: if you have a blueprint named admin and you want to render a template called index. filename='style.2 Static Files A blueprint can expose a folder with static files by providing a path to a folder on the filesystem via the static_folder keyword argument.3 Templates If you want the blueprint to expose templates you can do that by providing the tem- plate_folder parameter to the Blueprint constructor: admin = Blueprint('admin'.css') as f: code = f.open_resource('static/style.read() 15. It can either be an absolute path or one relative to the folder of the blueprint: admin = Blueprint('admin'.5.html which is specific to this blueprint. So if you have a blueprint in the folder yourapplication/admin and you want to ren- der the template 'admin/index. This also means that if you don’t want a blueprint template to be accidentally overridden.html in the actual application template folder.with simple_page. That way you can easily override templates that a blueprint provides in the actual application.static so you can generate URLs to it like you would do to the static folder of the application: url_for('admin.html' and you have provided templates as a tem- plate_folder you will have to create a file like this: yourapplication/admin/templates/ admin/index.html. make sure that no other blueprint or actual application template has the same relative path. When mul- tiple blueprints provide the same relative template path the first blueprint registered takes precedence over the others.static'.

use admin/index.6 Building URLs If you want to link from one page to another you can use the url_for() function just like you normally would do just that you prefix the URL endpoint with the name of the blueprint and a dot (. If you encounter problems loading the correct templates enable the EXPLAIN_TEMPLATE_LOADING config variable which will instruct Flask to print out the steps it goes through to locate templates on every render_template call. 15.html __init__. 15.): url_for('admin.html') More information on error handling see Custom Error Pages.7 Error Handlers Blueprints support the errorhandler decorator just like the Flask application object.index') This will link to admin. Here is an example for a “404 Page Not Found” exception: @simple_page.py And then when you want to render the template. 97 . you can use relative redi- rects by prefixing the endpoint with a dot only: url_for('.index') Additionally if you are in a view function of a blueprint or a rendered template and you want to link to another endpoint of the same blueprint.html as the name to look up the template by.yourpackage/ blueprints/ admin/ templates/ admin/ index. so it is easy to make Blueprint-specific custom error pages.index for instance in case the current request was dispatched to any other admin blueprint endpoint.errorhandler(404) def page_not_found(e): return render_template('pages/404.

98 .

Consider reading Flask Extension Development to develop your own Flask extension.4 Flask Before 0. 16. you can always create your own. instead you have to import from flaskext. For in- stance they add support for databases and other common tasks. CHAPTER SIXTEEN FLASK EXTENSIONS Flask extensions extend the functionality of Flask in various different ways.foo or flask_foo depending on how the extension is 99 . If you add a Flask extension as dependency to your requirements.txt or setup. If you have an extension called Flask-Foo or Foo-Flask it should be always importable from flask_foo: import flask_foo 16. If this is the case. There are no general rules in how extensions are supposed to behave but they are imported from common locations.ext package will not exist.8 If you are using Flask 0.3 Building Extensions While Flask Extension Registry contains many Flask extensions. you may not find an extension that fits your need.1 Finding Extensions Flask extensions are listed on the Flask Extension Registry and can be downloaded with easy_install or pip.7 or earlier the flask.py file they are usually installed with a simple command or when your application installs. 16. 16.2 Using Extensions Extensions typically have documentation that goes along that shows how to use it.

7 or earlier you should still import from the flask.py And here is how you can use it: import flaskext_compat flaskext_compat.distributed. If you want to develop an application that supports Flask 0.ext package. 100 . We provide you with a compatibility module that provides this package for older versions of Flask.activate() from flask.ext import foo Once the flaskext_compat module is activated the flask. You can download it from GitHub: flaskext_compat.ext will exist and you can start importing from there.

CHAPTER SEVENTEEN COMMAND LINE INTERFACE New in version 0. The way this script works is by providing access to all the commands on your Flask application’s Flask.1 Basic Usage After installation of Flask you will now find a flask script installed into your vir- tualenv. For instance mymodule:application would tell it to use the application object in the mymodule.11 is the built-in integration of the click com- mand line interface. an application needs to be discovered.py flask run 101 . Given a hello. For the flask script to work. In that imported file the name of the app needs to be called app or optionally be speci- fied after a colon.cli instance as well as some built-in commands that are always there. 17.py file. Flask extensions can also register more commands there if they desire so.py file with the application in it named app this is how it can be run. If you don’t want to install Flask or you have a special use-case you can also use python -m flask to accomplish exactly the same. Environment variables (On Windows use set instead of export): export FLASK_APP=hello flask run Or with a filename: export FLASK_APP=/path/to/hello. This enables a wide range of new features for the Flask ecosystem and your own applications.11. One of the nice new features in Flask 0. This is achieved by exporting the FLASK_APP environment variable. It can be either set to an import path or to a filename of a Python module that contains a Flask application.

17.""" click. Flask uses click for the command interface which makes creating custom commands very easy. 17. setup the correct application context and setup the local variables in the shell.echo('Init the db') The command will then show up on the command line: 102 .4 Running a Shell To run an interactive Python shell you can use the shell command: flask shell This will start up an interactive Python shell. make_shell_context() method of the application. That way every time you activate your virtualenv you automatically also activate the correct application name. By default you have access to your app and g.5 Custom Commands If you want to add more commands to the shell script you can do this easily.2 Virtualenv Integration If you are constantly working with a virtualenv you can also put the export FLASK_APP into your activate script by adding it to the bottom of the file. If set to 1 debug is enabled or 0 disables it: export FLASK_DEBUG=1 17. This is done by invoking the Flask.17.3 Debug Flag The flask script can also be instructed to enable the debug mode of the application automatically by exporting FLASK_DEBUG. For instance if you want a shell command to initialize the database you can do this: import click from flask import Flask app = Flask(__name__) @app.cli.command() def initdb(): """Initialize the database.

environ['YOURAPPLICATION_CONFIG']) Once this has happened you can make the flask command automatically pick it up: export YOURAPPLICATION_CONFIG=/path/to/config.6 Application Context Most commands operate on the application so it makes a lot of sense if they have the application context setup. For instance if you have a factory function that creates an application from a filename you could make a separate file that creates such an application from an environment variable.cli. This behavior is not available if a command is added later with add_command() or through other means.py with these contents: import os from yourapplication import create_app app = create_app(os.cfg export FLASK_APP=/path/to/autoapp. cli with the command() the callback will automatically be wrapped through cli. Flask won’t be able to figure out how to instantiate your application properly by itself.7 Factory Functions In case you are using factory functions to create your application (see Application Facto- ries) you will discover that the flask command cannot work with them directly. Be- cause of this reason the recommendation is to create a separate file that instantiates applications. It can also be disabled by passing with_appcontext=False to the decorator: @app. with_appcontext() which informs the cli system to ensure that an application con- text is set up. This could be a file named autoapp. This is not the only way to make this work. 103 .py From this point onwards flask will find your application.command(with_appcontext=False) def example(): pass 17. Another is the Custom Scripts support. Because of this. if you register a callback on app.$ flask initdb Init the db 17.

To explain all of this.py')) @click. 'wikiconfig. Next step is to create a FlaskGroup. you can also make your own “driver scripts”. This is very useful if you write reusable applications that you want to ship to users and they should be presented with a custom management script. To understand why you might want custom scripts you need to understand how click finds and executes the Flask application.FlaskGroup click group.cli will expect to be (indirectly at least) launched from a flask. cli package.py script that manages a hypothetical wiki application. so let’s go through all parts step by step. 2. This can either directly import an application object or create it (see Application Factories).""" if __name__ == '__main__': cli() That’s a lot of code for not much.cli import FlaskGroup def create_wiki_app(info): from yourwiki import create_app return create_app( config=os. In this case we just make an empty function with a help doc string that just does nothing and then pass the create_wiki_app 104 . Primarily it does not work with application factory functions (see Application Factories). There is one big caveat and that is. Since Flask uses click for the scripts there is no reason you cannot hook these scripts into any click application. With a custom script you don’t have this problem as you can fully customize how the application will be created. We will go through the details afterwards: import os import click from flask. This is simple but it has some limitations.environ.get('WIKI_CONFIG'.17. create_app=create_wiki_app) def cli(): """This is a management script for the wiki application. If you use the flask script you specify the application to work with on the command line or environment variable as an import name. that commands registered to Flask. In this case we load the config from an environment vari- able.cli. Primarily we are here interested in the FlaskGroup click group.8 Custom Scripts While the most common way is to use the flask command. 1.group(cls=FlaskGroup. 3. here is an example manage. First we import the click library as well as the click extensions from the flask. The next thing we do is defining a function that is invoked with the script info object (ScriptInfo) from Flask and its purpose is to fully import and create the application. This is necessary so that the commands know which Flask application they have to work with.

All is rounded up by invoking the script. function as a factory function.commands:cli '''. ) Inside mypackage/commands.commands] my-command=mypackage. 105 . This is useful to provide extra functionality that Flask itself cannot ship.py file that declares an entrypoint that points to a click command: Example setup.. However there is a second way to add CLI plugins to Flask which is through setuptools. 4.py you can then export a Click object: import click @click.py: from setuptools import setup setup( name='flask-my-extension'.""" Once that package is installed in the same virtualenv as Flask itself you can run flask my-command to invoke your command. .9 CLI Plugins Flask extensions can always patch the Flask. Whenever click now needs to operate on a Flask application it will call that func- tion with the script info and ask for it to be created.command() def cli(): """This is an example command.cli instance with more commands if they want.. 17. If you make a Python package that should export a Flask command line plugin you can ship a setup. entry_points=''' [flask.

106 .

This will immediately launch a local server exactly the same way the flask script does. For instance the reloader can be disabled: $ flask run --no-reload 18.run() This works well for the common case but it does not work well for development which is why from Flask 0. Example: if __name__ == '__main__': app.1 Command Line The flask command line script (Command Line Interface) is strongly recommended for development because it provides a superior reload experience due to how it loads the application.run() method. The basic usage is like this: $ export FLASK_APP=my_application $ export FLASK_DEBUG=1 $ flask run This will enable the debugger. The individual features of the server can be controlled by passing more arguments to the run option. The best one is the flask command line utility but you can also continue using the Flask.2 In Code The alternative way to start the application is through the Flask.11 there are multiple built-in ways to run a development server. the reloader and then start the server on http://localhost:5000/. The reason for this 107 .11 onwards the flask method is recommended. 18.run() method. CHAPTER EIGHTEEN DEVELOPMENT SERVER Starting with Flask 0.

sometimes crashing without message or dying when a syntax or import error happens).is that due to how the reload mechanism works there are some bizarre side-effects (like executing certain code twice. 108 . It is however still a perfectly valid method for invoking a non automatic reloading application.

But the code you want to test might depend on them. 19. just import your application and start playing around. For more information see Command Line Interface. because it does not require any specific setup upfront. For instance the shell is automatically initialized with a loaded application context. CHAPTER NINETEEN WORKING WITH THE SHELL New in version 0. but also for unittesting and other situations that require a faked request context. There are however some handy helpers to make playing around in the shell a more pleasant experience. so what can you do? This is where some helper functions come in handy. The main issue with interactive console sessions is that you’re not triggering a request like a browser does which means that g. request and others are not available. Flask itself does not come with an interactive shell.1 Command Line Interface Starting with Flask 0.2 Creating a Request Context The easiest way to create a proper request context from the shell is by using the test_request_context method which creates us a RequestContext: >>> ctx = app. 19.3. One of the reasons everybody loves Python is the interactive shell. Keep in mind however that these functions are not only there for interactive shell usage.11 the recommended way to work with the shell is the flask shell command which does a lot of this automatically for you. It basically allows you to execute Python commands in real time and immediately get results back.test_request_context() 109 . Generally it’s recommended that you read the The Request Context chapter of the doc- umentation first.

To shutdown a request. There you could also define some more helper methods for common things such as initializing the database. create yourself a module with stuff you want to star import into your interactive session. Just call preprocess_request(): >>> ctx = app. So this is the perfect place to automatically tear down resources that were needed by the request context (such as database connections). you need to trick a bit before the after request functions (trig- gered by process_response()) operate on a response object: >>> app.pop() 19.3 Firing Before/After Request By just creating a request context. but in the shell it’s easier to use the push() and pop() methods by hand: >>> ctx.pop() The functions registered as teardown_request() are automatically called when the context is popped. 19.4 Further Improving the Shell Experience If you like the idea of experimenting in a shell. dropping tables etc. you still don’t have run the code that is normally run before a request.Normally you would use the with statement to make this request object active. This might result in your database being unavailable if you are connecting to the database in a before-request callback or the current user not being stored on the g object etc.response_class()) <Response 0 bytes [200 OK]> >>> ctx. Just put them into a module (like shelltools) and import from there: >>> from shelltools import * 110 .process_response(app.push() >>> app. This however can easily be done yourself. in that case just ignore it.push() From that point onwards you can work with the request object until you call pop: >>> ctx.preprocess_request() Keep in mind that the preprocess_request() function might return a response object.test_request_context() >>> ctx.

CHAPTER TWENTY PATTERNS FOR FLASK Certain things are common enough that the chances are high you will find them in most web applications.html . (Make sure to delete all . There are more user contributed snippets and patterns in the Flask Snippet Archives.html login.py. otherwise things would most likely break) You should then end up with something like that: /yourapplication /yourapplication 111 . just create a new folder yourapplication inside the existing one and move everything below it. the database connection is closed again.1. 20.css /templates layout.py to __init__.1 Simple Packages To convert that into a larger one. In that case..py /static style. For example quite a lot of applications are using relational databases and user authentication. 20.. chances are they will open a database connection at the beginning of the request and get the information of the currently logged in user.1 Larger Applications For larger applications it’s a good idea to use a package instead of a module. At the end of the request.html index. That is quite simple. Imagine a small application looks like this: /yourapplication yourapplication.pyc files first. Then rename yourapplication.

Similiarly you can turn on “debug mode” with this environment variable: export FLASK_DEBUG=true In order to install and run the application you need to issue the following commands: pip install -e . Import the view module after the application object is created.py /static style. packages=['yourapplication']. all the view functions (the ones with a route() decorator on top) have to be imported in the __init__. the Flask application object creation has to be in the __init__. 2..py file. Not the object itself. __init__. 112 . install_requires=[ 'flask'.css /templates layout.py next to the inner yourapplication folder with the following contents: from setuptools import setup setup( name='yourapplication'. but the module it is in.py will not work. The only thing you have to remember is the following quick checklist: 1.html login. But how do you run your application now? The naive python yourapplication/ __init__. ) In order to run the application you need to export an environment variable that tells Flask where to find the application instance: export FLASK_APP=yourapplication If you are outside of the project directory make sure to provide the exact path to your application directory. just add a new file called setup. flask run What did we gain from this? Now we can restructure the application a bit into multiple modules. That way each module can import it safely and the __name__ variable will resolve to the correct package.. ].html .py file. But that is not a big problem.html index. Let’s just say that Python does not want modules in pack- ages to be the startup file. include_package_data=True.

py). Circular Imports Every Python programmer hates them.py /yourapplication __init__.css /templates layout. Be advised that this is a bad idea in general but here it is actually fine. 113 . In this case views.1.html .route('/') def index(): return 'Hello World!' You should then end up with something like that: /yourapplication setup.py views.html index.Here’s an example __init__.html login. and yet we just added some: circular imports (That’s when two modules depend on each other.py would look like: from yourapplication import app @app. 20.py: from flask import Flask app = Flask(__name__) import yourapplication.views And this is what views.py and just ensuring the module is imported and we are doing that at the bottom of the file.. For a gentle intro- duction into this topic refer to the Modular Applications with Blueprints chapter of the documentation.py /static style.2 Working with Blueprints If you have larger applications it’s recommended to divide them into smaller groups where each group is implemented with the help of a blueprint. There are still some problems with that approach but if you want to use decorators there is no way around that. Check out the Becoming Big section for some inspiration how to deal with that..py depends on __init__. The reason for this is that we are not actually using the views in __init__.

from_pyfile(config_filename) from yourapplication. you can have multiple instances of the same application running in the same application process which can be handy. So how would you then actually implement that? 20.views. Blueprint.2 Application Factories If you are already using packages and blueprints for your application (Modular Appli- cations with Blueprints) there are a couple of really nice ways to further improve the experience.2. You can however use it from within a request. You can have instances of the application with different settings to test every case. Imagine you want to run different versions of the same ap- plication.frontend import frontend app.init_app(app) from yourapplication. A common pattern is creating the application object when the blueprint is imported. Multiple instances. Of course you could have multiple instances with different configs set up in your webserver.route('/') def index(): return render_template(current_app. How do you get access to the application with the config? Use current_app: from flask import current_app.config. Like this: def create_app(config_filename): app = Flask(__name__) app.register_blueprint(admin) app. render_template admin = Blueprint('admin'. __name__.admin import admin from yourapplication.views. 2.register_blueprint(frontend) return app The downside is that you cannot use the application object in the blueprints at import time. but if you use factories. url_prefix='/admin') @admin. But if you move the creation of this object into a function.config['INDEX_TEMPLATE']) 114 . you can then create multiple instances of this app later. Testing.model import db db.1 Basic Factories The idea is to set up the application in a function. So why would you want to do this? 1.20.

rather.Here we look up the name of a template in the config.model import db db.init_app(app) Using this design pattern.3 Using Applications So to use such an application you then have to create the application first in a sep- arate file otherwise the flask command won’t be able to find it. you should not do something along those lines: def create_app(config_filename): app = Flask(__name__) app.py (or equivalent): db = SQLAlchemy() and in your application. For more information about the design of extensions refer to Flask Extension Development. 20.py file that creates such an application: from yourapplication import create_app app = create_app('/path/to/config. Here an example exampleapp.py (or equivalent): def create_app(config_filename): app = Flask(__name__) app.cfg') It can then be used with the flask command: export FLASK_APP=exampleapp flask run 115 . so one extension object can be used for multiple apps.from_pyfile(config_filename) from yourapplication. no application-specific state is stored on the extension ob- ject. as an example.config.2 Factories & Extensions It’s preferable to create your extensions and app factories so that the extension object does not initially get bound to the application.from_pyfile(config_filename) db = SQLAlchemy(app) But.2. 20. Using Flask-SQLAlchemy.config. in model.2.

This would allow you to run a Django and a Flask application in the same interpreter side by side if you want.serving import run_simple app = Flask(__name__) app. Werkzeug provides a builtin server for development available at werkzeug.3. In order to use the interactive debugger.serving import run_simple run_simple('localhost'. Use a full-blown WSGI server.2.20. debugging must be enabled both on the ap- plication and the simple server. use_reloader=True) Note that run_simple is not intended for use in production. call a function from a blueprint when the application is setting up so that you have a place to modify attributes of the application (like hooking in before / after request handlers etc. They run different configurations and are dispatched on the WSGI level. see Deployment Options. you can improve it. 20. application. The following changes are straightforward and possible: 1. run_simple(): from werkzeug. Add in WSGI middlewares when the application is creating if necessary. The fundamental difference from the module approach is that in this case you are run- ning the same or different Flask applications that are entirely isolated from each other.3 Application Dispatching Application dispatching is the process of combining multiple Flask applications on the WSGI level. The usefulness of this depends on how the applications work internally. Here is the “hello world” example with debugging and run_simple: from flask import Flask from werkzeug. You can combine not only Flask applications but any WSGI application.serving.) 3.debug = True 116 . make it possible to pass in configuration values for unittests so that you don’t have to create config files on the filesystem 2. For development.4 Factory Improvements The factory function from above is not very clever so far.1 Working with this Document Each of the techniques and examples below results in an application object that can be run with any WSGI server. For production. 20. 5000.

In order to develop your application to support creating new instances in functions have a look at the Application Factories pattern.DispatcherMiddleware. wsgi. that is really easy to implement. it is dynamically created and remem- bered: 117 . For instance you configure your webserver to dispatch all requests for all subdomains to your application and you then use the subdomain information to create user-specific in- stances. For example you could have your main application run on / and your backend inter- face on /backend: from werkzeug. use_debugger=True. 5000. Assuming the application is created inside a function and you can call that function to instantiate it.3 Dispatch by Subdomain Sometimes you might want to use multiple instances of the same application with different configurations. If that application does not exist yet.wsgi import DispatcherMiddleware from frontend_app import application as frontend from backend_app import application as backend application = DispatcherMiddleware(frontend.3. Once you have your server set up to listen on all subdomains you can use a very simple WSGI application to do the dynamic application creation. { '/backend': backend }) 20.route('/') def hello_world(): return 'Hello World!' if __name__ == '__main__': run_simple('localhost'. use_reloader=True.@app. app. The idea here is that each Flask application is a valid WSGI application and they are combined by the dispatcher middleware into a larger one that is dispatched based on prefix. You write your own WSGI application that looks at the request that comes and delegates it to your Flask application. A very common example would be creating applications per subdomain.3.2 Combining Applications If you have entirely separated applications and you want them to work next to each other in the same Python interpreter process you can take advantage of the werkzeug. The perfect level for abstraction in that regard is the WSGI layer. use_evalex=True) 20.

split(':')[0] assert host.get_application(environ['HTTP_HOST']) return app(environ.lock = Lock() self.instances[subdomain] = app return app def __call__(self.from threading import Lock class SubdomainDispatcher(object): def __init__(self.4 Dispatch by Path Dispatching by a path on the URL is very similar.exceptions import NotFound def make_app(subdomain): user = get_user_for_subdomain(subdomain) if user is None: # if there is no user for that subdomain we still have # to return a WSGI application that handles that request.get(subdomain) if app is None: app = self. # We can then just return the NotFound() exception as # application which will render a default 404 page. environ. get_user_for_subdomain from werkzeug.lock: app = self. host): host = host. 'Configuration error' subdomain = host[:-len(self. create_app): self. make_app) 20.instances = {} def get_application(self.com'. domain. start_response) This dispatcher can then be used like this: from myapplication import create_app.instances.') with self. Instead of looking at the Host header to figure out the subdomain one simply looks at the request path up to the first slash: 118 .domain).create_app(subdomain) self. start_response): app = self.create_app = create_app self.endswith(self.domain = domain self. # You might also redirect the user to the main page then return NotFound() # otherwise create the application for the specific user return create_app(user) application = SubdomainDispatcher('example.3.rstrip('.domain)].

create_app = create_app self. default_app.instances = {} def get_application(self.get(prefix) if app is None: app = self. start_response) The big difference between this and the subdomain one is that this one falls back to another application if the creator function returns None: from myapplication import create_app. default_app.get_application(peek_path_info(environ)) if app is not None: pop_path_info(environ) else: app = self.4 Implementing API Exceptions It’s very common to implement RESTful APIs on top of Flask.lock: app = self.instances[prefix] = app return app def __call__(self.default_app return app(environ. The better solution than using abort to signal errors for invalid API usage is to im- 119 .lock = Lock() self.create_app(prefix) if app is not None: self. start_response): app = self.from threading import Lock from werkzeug. get_user_for_prefix def make_app(prefix): user = get_user_for_prefix(prefix) if user is not None: return create_app(user) application = PathDispatcher(default_app. make_app) 20. create_app): self.instances. One of the first things that developers run into is the realization that the builtin exceptions are not expressive enough for APIs and that the content type of text/html they are emitting is not very useful for API consumers.wsgi import pop_path_info. peek_path_info class PathDispatcher(object): def __init__(self.default_app = default_app self. environ. prefix): with self.

status_code=None.message return rv A view can now raise that exception with an error message.plement your own exception type and install an error handler for it that produces the errors in the format the user is expecting.4.4. The reason for this is that there is no handler registered for this error class.1 Simple Exception Class The basic idea is to introduce a new exception that can take a proper human readable message. message.payload or ()) rv['message'] = self.2 Registering an Error Handler At that point views can raise that error.__init__(self) self.payload = payload def to_dict(self): rv = dict(self.errorhandler(InvalidUsage) def handle_invalid_usage(error): response = jsonify(error.to_dict()) response. payload=None): Exception.3 Usage in Views Here is how a view can use that functionality: 120 . a status code for the error and some optional payload to give more context for the error. 20.message = message if status_code is not None: self.4. Additionally some extra payload can be provided as a dictionary through the payload parameter.status_code = status_code self. This is a simple example: from flask import jsonify class InvalidUsage(Exception): status_code = 400 def __init__(self. but it would immediately result in an internal server error.status_code return response 20. That however is easy to add: @app. 20.status_code = error.

The code below checks if the language code is not yet in the dictionary of URL values and if the endpoint wants a value named 'lang_code': @app.url_defaults def add_language_code(endpoint.. values): if 'lang_code' in values or not g. Flask 0.lang_code = lang_code .route('/<lang_code>/about') def about(lang_code): g. status_code=410) 20. The idea is that you might have a bunch of resources with common parts in the URL that you don’t always explicitly want to provide..7. URL processors are especially helpful when combined with blueprints.1 Internationalized Application URLs Consider an application like this: from flask import Flask. They can automatically inject values into a call for url_for() automatically. For the latter.route('/<lang_code>/') def index(lang_code): g. We will handle both application specific URL processors here as well as blueprint specifics.lang_code: 121 .@app.5.lang_code = lang_code . @app.5 Using URL Processors New in version 0. but if you want to generate URLs from one function to another you would have to still provide the language code explicitly which can be annoying. a decorator could be used to simplify this. 20.. For instance you might have a bunch of URLs that have the language code in it but you don’t want to have to handle it in every single function yourself. g app = Flask(__name__) @app. this is where url_defaults() functions come in. This is an awful lot of repetition as you have to handle the language code setting on the g object yourself in every single function.7 introduces the concept of URL processors.route('/foo') def get_foo(): raise InvalidUsage('This view is gone'.. Sure.

url_map. The reverse of that function are url_value_preprocessor()s.lang_code The method is_endpoint_expecting() of the URL map can be used to figure out if it would make sense to provide a language code for the given endpoint.. g app = Flask(__name__) @app.url_value_preprocessor def pull_lang_code(endpoint. The idea is that they pull information out of the values dictionary and put it somewhere else: @app.2 Internationalized Blueprint URLs Because blueprints can automatically prefix all URLs with a common string it’s easy to automatically do that for every function. None) @app. None) That way you no longer have to do the lang_code assignment to g in every function.lang_code: return if app. 'lang_code'): values['lang_code'] = g.route('/<lang_code>/') def index(): .lang_code = values. but the more beautiful solution is using a blueprint.url_defaults def add_language_code(endpoint.pop('lang_code'.lang_code = values... @app.is_endpoint_expecting(endpoint.5.route('/<lang_code>/about') def about(): . Once the 'lang_code' is popped from the values dictionary and it will no longer be forwarded to the view function reducing the code to this: from flask import Flask.url_value_preprocessor def pull_lang_code(endpoint. They are executed right after the request was matched and can execute code based on the URL values. You can further improve that by writing your own decorator that prefixes URLs with the language code. 'lang_code'): values['lang_code'] = g.lang_code @app. values): if 'lang_code' in values or not g. return if app.pop('lang_code'. values): g.. values): g.url_map. Furthermore blueprints can have per- 122 .is_endpoint_expecting(endpoint. 20.

you will already have pip and setuptools on your system. Otherwise. @bp. The best known feature of this system is the entry point sup- port which allows one package to declare an “entry point” that another package can hook into to extend the other package. __name__. and all the libraries you can find on PyPI are distributed with either setup- tools or distutils. • package registry: setuptools registers your package with your Python installa- tion. a basic module installation system shipped with Python to also support various more complex constructs that make larger applications easier to distribute: • support for dependencies: a library or application can declare a list of other libraries it depends on which will be installed automatically for you.. • installation manager: pip can install other libraries for you.9) or Python 3 (>=3. g bp = Blueprint('frontend'.pop('lang_code') @bp. values): g. g.setdefault('lang_code'.url_defaults def add_language_code(endpoint.route('/about') def about(): . 20. It extends distutils.org.6 Deploying with Setuptools Setuptools. This makes it possible to query information provided by one package from another package.. 123 .4) installed from python. is an extension library that is commonly used to distribute Python libraries and extensions.lang_code) @bp.lang_code = values.route('/') def index(): . If you have Python 2 (>=2.url_value_preprocessor def pull_lang_code(endpoint. url_prefix='/<lang_code>') @bp.. you will need to install them yourself..7.blueprint URL processors which removes a whole lot of logic from the url_defaults() function because it no longer has to check if the URL is really interested in a 'lang_code' parameter: from flask import Blueprint. values): values. Flask itself.

If you want setuptools to lookup the packages for you automatically.6. packages=['yourapplication']. 20. include_package_data and zip_safe might not be.in file and install all the entries that match as package data. version='1. packages=find_packages() ) Most parameters to the setup function should be self explanatory.0'. 124 . zip_safe=False. include_package_data tells setuptools to look for a MANIFEST.py next to your application. If you want to fully automate the process.py and you are not using a module.. head over to the Larger Applications pattern to see how this can be done. find_packages setup( .In this case we assume your application is called yourapplication. you have setuptools available on your system. but a package. install_requires=['Flask'] ) Please keep in mind that you have to list subpackages explicitly. In general you probably don’t want your packages to be installed as zip files because some tools do not support them and they make debugging a lot harder. but because everybody will look for a file with that name.. If you have not yet converted your application into a package. long_description=__doc__. include_package_data=True. The name of the file is only convention. The zip_safe flag can be used to force or prevent zip Archive creation.1 Basic Setup Script Because you have Flask installed. Flask already depends upon setuptools. We will use this to distribute the static files and templates along with the Python module (see Distributing Resources). you can use the find_packages function: from setuptools import setup. also read the Deploying with Fabric chapter. A basic setup. you better not change it.py file for a Flask application looks like this: from setuptools import setup setup( name='Your Application'. A working deployment with setuptools is the first step into more complex and more automated deployment scenarios. Your setup code always goes into a file named setup. Standard disclaimer applies: you better use a virtualenv.

<=1. This file lists all the files that should be added to your tarball: recursive-include yourapplication/templates * recursive-include yourapplication/static * Don’t forget that even if you enlist them in your MANIFEST. Each item in that list is the name of a package that should be pulled from PyPI on installation.6.in file next to your setup.6.2 Tagging Builds It is useful to distinguish between release and development builds.20. What if you want to depend on a package that cannot be found on PyPI and won’t be because it is an internal package you don’t want to share with anyone? Just do it as if there was a PyPI entry and provide a list of alternative locations where setuptools should look for tarballs: dependency_links=['http://example.0. 20.0' ] As mentioned earlier.tar. is to create a MANIFEST. The reason for this is that setuptools does not know which files to add for you.tar. [egg_info] tag_build = . Add a setup. 'SQLAlchemy>=0.dev20160314.in file.dev” and the current date appended: flaskr-1.cfg file to configure these options. By default it will always use the most recent version.gz.6. dependencies are pulled from PyPI. Here some examples: install_requires=[ 'Flask>=0.3 Distributing Resources If you try to install the package you just created. What you should do. they won’t be in- stalled for you unless you set the include_package_data parameter of the setup function to True! 20. you will notice that folders like static or templates are not installed for you.4 Declaring Dependencies Dependencies are declared in the install_requires parameter as a list.py file.7. gz.0.com/yourfiles'] 125 . py release sdist will create a release package with only the version: flaskr-1.6'. Running python setup. but you can also provide minimum and maximum version requirements.2'.dev tag_date = 1 [aliases] release = egg_info -RDb ‘’ Running python setup. 'BrokenPackage>=0.py sdist will create a development package with ”.

In combination with a properly set up Python package (Larger Applications) and a good concept for configurations (Configuration Handling) it is very easy to deploy Flask applications to external servers.7 Deploying with Fabric Fabric is a tool for Python similar to Makefiles but with the ability to execute com- mands on a remote server. Before we get started. It is named fabfile. It will install your application into the virtualenv’s site- packages folder and also download and install all dependencies: $ python setup. you can use the develop command instead: $ python setup. but for this example we chose Apache + mod_wsgi because it’s very easy to setup and has a simple way to reload applications without root access. 20. This tutorial assumes the latest version of Fabric. You can then continue to work on the code without having to run install again after each change. You can of course use your own favourite server there. 20.6. • In the following example we are using mod_wsgi for the remote servers. 20.7.Make sure that page has a directory listing and the links on the page are pointing to the actual tarballs with their correct filenames as this is how setuptools will find the files.py and executed by the fab command. provide the URL to that server.1 Creating the first Fabfile A fabfile is what controls what Fabric executes.py install If you are developing on the package and also want the requirements to be installed.py develop This has the advantage of just installing a link to the site-packages folder instead of copying the data over.py file (Deploying with Setuptools).py script with the install parameter. • The application already has to be a package and requires a working setup. here a quick checklist of things we have to ensure upfront: • Fabric 1. All the functions defined in that file will show up as fab subcom- 126 .5 Installing / Developing To install your application (ideally into a virtualenv) just run the setup. If you have an internal company server that contains the packages.0 has to be installed locally.

wsgi file to trigger a reload in mod_wsgi run('touch /var/www/yourapplication. creating a command in the fabfile for this is overkill. So how do we bootstrap a new server into our infrastructure? This now depends on the number of servers we want to set up. capture=False) def deploy(): # figure out the package name and version dist = local('python setup. To deploy the current version of the code on the remote server you would use this command: $ fab pack deploy However this requires that our server already has the /var/www/yourapplication folder created and /var/www/yourapplication/env to be a virtual environment.com'. This is a basic first example that has the ability to upload the current source code to the server and install it into a pre-existing virtual environment: from fabric.hosts = ['server1. In this case we will add them to the fabfile.gz' % dist # upload the package to the temporary folder on the server put('dist/%s' % filename. These hosts can be defined either in the fabfile or on the command line.mands. They are executed on one or more hosts.py sdist --formats=gztar'.wsgi') 20. But obviously you can do that.wsgi file on the server.tar. capture=True). 'server2. Fur- thermore are we not creating the configuration or .user = 'appuser' # the servers where the commands are executed env.7.com'] def pack(): # build the package local('python setup.2 Running Fabfiles Now how do you execute that fabfile? You use the fab command.example.api import * # the user to use for the remote commands env.strip() filename = '%s. '/tmp/%s' % filename) # install the package in the application's virtualenv with pip run('/var/www/yourapplication/env/bin/pip install /tmp/%s' % filename) # remove the uploaded package run('rm -r /tmp/%s' % filename) # touch the .py --fullname'.example. In that case you 127 . If we just have one application server (which the majority of applications will have).

(See mod_wsgi (Apache) for more information) So now the question is.from_object('yourapplication.environ['YOURAPPLICATION_CONFIG'] = '/var/www/yourapplication/application.config. This is a short example that does exactly that: import os os. So we have to put the configuration in a place where the application will able to find it.wsgi file so that we can automatically reload the application by touching it.default_config') app. where do the application. Create a new Apache config for yourapplication and activate it.wsgi and application.7. the application will find the correct configuration file by looking up the YOURAPPLICATION_CONFIG environment variable.wsgi file to the server and the configuration file for the application (eg: application.3 The WSGI File The WSGI file has to import the application and also to set an environment variable so that the application knows where to look for the config. 20. Make sure to ac- tivate watching for changes of the . Create the directory structure in /var/www: $ mkdir /var/www/yourapplication $ cd /var/www/yourapplication $ virtualenv --distribute env 2.example.com bootstrap To setup a new server you would roughly do these steps: 1.from_envvar('YOURAPPLICATION_CONFIG') This approach is explained in detail in the Configuration Handling section of the docu- mentation.config.cfg' from yourapplication import app The application itself then has to initialize itself like this to look for the config at that environment variable: app = Flask(__name__) app. Configuration files 128 .cfg) 3. Upload a new application.4 The Configuration File Now as mentioned above.7.cfg files come from? 20.would probably call it setup or bootstrap and then pass the servername explicitly on the command line: $ fab -H newserver.

• hook in testing functionality so that you can deploy to an external server and run the test suite. Then symlink the file that is active for the server into the location where it’s expected (eg: /var/www/ yourapplication). It could initialize a new virtual environment. We have set up the servers so that they have their virtual environments and activated apache configs. 20. 20.5 First Deployment Now we can do our first deployment. That way you can also easily go back to older versions.py file we will automatically pull in the required libraries into our virtual environment.7. in our case here we only expect one or two servers and we can upload them ahead of time by hand. Thanks to the setup.7. so you do not version them usually. Now we can pack up the application and deploy it: $ fab pack deploy Fabric will now connect to all servers and run the commands as written down in the fabfile. • Put configuration files into a separate version control repository and symlink the active configs into place. • You could also put your application code into a repository and check out the latest version on the server and then install. setup apache appropriately etc.6 Next Steps From that point onwards there is so much that can be done to make deployment actu- ally fun: • Create a bootstrap command that initializes new servers. A popular approach is to store configuration files for different servers in a sepa- rate version control repository and check them out on all servers.have the unfriendly quality of being different on all computers. First it will execute pack so that we have our tarball ready and then it will execute deploy and upload the source code to all servers and install it there. 129 . Either way. Working with Fabric is fun and you will notice that it’s quite magical to type fab deploy and see your application being deployed automatically to one or more remote servers.

.. If you want to use this code outside a request context you can use it in a Python shell by opening the application context by hand: 130 .close() Now._app_ctx_stack.1 Connect on Demand The upside of this approach (connecting on first use) is that this will only open the connection if truly necessary. the application must either have an active application con- text (which is always true if there is a request in flight) or create an application context itself.cursor() . None) if db is None: db = g. Note: Please keep in mind that the teardown request and appcontext functions are al- ways executed. Example: @app.9 or older you need to use flask.8 Using SQLite 3 with Flask In Flask you can easily implement the opening of database connections on demand and closing them when the context dies (usually at the end of the request). Here is a simple example of how you can use SQLite 3 with Flask: import sqlite3 from flask import g DATABASE = '/path/to/database. Because of this we have to make sure here that the database is there before we close it. 20._database = sqlite3.connect(DATABASE) return db @app. '_database'.teardown_appcontext def close_connection(exception): db = getattr(g. Whenever the context is destroyed the database connection will be terminated. None) if db is not None: db. '_database'.route('/') def index(): cur = get_db(). Note: if you use Flask 0.20. to use the database.top instead of g as the flask.g object was bound to the request and not application context.db' def get_db(): db = getattr(g.8. even if a before-request handler failed or was never executed. At that point the get_db function can be used to get the current database connec- tion.

These are namedtuple s. FirstName. executing and fetching the results: def query_db(query. so we can access them either by index or by key. args) rv = cur. one=False): cur = get_db().. which are much nicer to deal with.db to get the current open database connection. and MiddleInitial: >>> # You can get values based on the row's name >>> r['FirstName'] John >>> # Or. in order to get dictionaries instead of tuples. It is executed for every result returned from the database to convert the result.row_factory = make_dicts This will make the sqlite3 module return dicts for this database connection. print(value) 1 John Doe M Additionally. Even more simply.row_factory = sqlite3. For example.Row This would use Row objects rather than dicts to return the results of queries.with app. a row factory function is use- ful. you can get them based on index >>> r[1] John # Row objects are also iterable: >>> for value in r: . this could be inserted into the get_db function we created above: def make_dicts(cursor.Row called r for the rows id. value) for idx.app_context(): # now you can use get_db() 20.close() return (rv[0] if rv else None) if one else rv 131 . it is a good idea to provide a query function that combines getting the cursor. row): return dict((cursor. value in enumerate(row)) db.execute(query.description[idx][0]. To simplify working with SQLite. we could place this in get_db instead: db. For instance. LastName.2 Easy Querying Now in each request handling function you can access g. as- suming we have a sqlite3.fetchall() cur. args=()..8.

use a question mark in the statement and pass in the arguments as a list.executescript(f. [the_username].cursor(). 'has the id'.commit() You can then create such a database from the Python shell: >>> from yourapplication import init_db >>> init_db() 20. 'has the id'. makes working with the database much more pleasant than it is by just using the raw cursor and connection objects. It’s a good idea to provide a function that creates the database based on that schema. one=True) if user is None: print 'No such user' else: print the_username. 20. in combination with a row factory.8.9 SQLAlchemy in Flask Many people prefer SQLAlchemy for database access. Never directly add them to the SQL statement with string formatting because this makes it possible to attack the application using SQL Injections. In this case it’s encouraged to use a package instead of a module for your flask application and drop the models into 132 .This handy little function. user['user_id'] To pass variable parts to the SQL statement.3 Initial Schemas Relational databases need schemas. Here is how you can use it: for user in query_db('select * from users'): print user['username']. user['user_id'] Or if you just want a single result: user = query_db('select * from users where username = ?'.sql'. This function can do that for you: def init_db(): with app.read()) db.open_resource('schema. so applications often ship a schema.app_context(): db = get_db() with app. mode='r') as f: db.sql file that creates the database.

If you are wondering why we don’t have to care about threads here (like we did in the SQLite3 example above with the g object): that’s because SQLAlchemy does that for us already with the scoped_session.orm import scoped_session.query = db_session.1 Flask-SQLAlchemy Extension Because SQLAlchemy is a common database abstraction layer and object relational mapper that requires a little bit of configuration effort.db'.ext.metadata.py module for your application: from sqlalchemy import create_engine from sqlalchemy. it makes a lot of sense.a separate module (Larger Applications). While that is not necessary. autoflush=False. To use SQLAlchemy in a declarative way with your application. Here’s the example database. convert_unicode=True) db_session = scoped_session(sessionmaker(autocommit=False. bind=engine)) Base = declarative_base() Base.9. just subclass the Base class that was created by the code above. Flask will automatically remove database sessions at the end of the request or when the application shuts down: 133 . You can download Flask-SQLAlchemy from PyPI.2 Declarative The declarative extension in SQLAlchemy is the most recent method of using SQLAlchemy. there is a Flask extension that handles that for you. sessionmaker from sqlalchemy. In addition to the following text I recommend the official documenta- tion on the declarative extension.declarative import declarative_base engine = create_engine('sqlite:////tmp/test. Otherwise # you will have to import them first before calling init_db() import yourapplication.create_all(bind=engine) To define your models. There are four very common ways to use SQLAlchemy.query_property() def init_db(): # import all modules here that might define models so that # they will be registered properly on the metadata. It allows you to define tables and models in one go. I will outline each of them here: 20. This is recommended if you want to get started quickly. you just have to put the following code into your application module.models Base. similar to how Django works. 20.9.

g. unique=True) def __init__(self.filter(User. The main difference is that you define tables and classes separately and map them together. primary_key=True) name = Column(String(50). Integer.3 Manual Object Relational Mapping Manual object relational mapping has a few upsides and a few downsides versus the declarative approach from above.all() [<User u'admin'>] >>> User.first() <User u'admin'> 20.add(u) >>> db_session.from yourapplication. e. email=None): self.database import Base class User(Base): __tablename__ = 'users' id = Column(Integer.name = name self. name=None.database import init_db >>> init_db() You can insert entries into the database like this: >>> from yourapplication.commit() Querying is simple as well: >>> User.database import db_session @app.py.database import db_session >>> from yourapplication.email = email def __repr__(self): return '<User %r>' % (self.teardown_appcontext def shutdown_session(exception=None): db_session. 134 .9.name == 'admin'). String from yourapplication.remove() Here is an example model (put this into models.): from sqlalchemy import Column.name) To create the database you can use the init_db function: >>> from yourapplication. It’s more flexible but a little more to type.query.query. unique=True) email = Column(String(120). 'admin@localhost') >>> db_session.models import User >>> u = User('admin'.

database import db_session @app. 135 . Column('name'.name = name self. Column. Integer. unique=True). bind=engine)) def init_db(): metadata.create_all(bind=engine) As in the declarative approach. Column('id'. String from sqlalchemy. MetaData from sqlalchemy.query_property() def __init__(self. Integer. email=None): self.db'. String(120). Column('email'. primary_key=True). String(50).py module for your application: from sqlalchemy import create_engine.email = email def __repr__(self): return '<User %r>' % (self.orm import mapper from yourapplication.name) users = Table('users'.teardown_appcontext def shutdown_session(exception=None): db_session. convert_unicode=True) metadata = MetaData() db_session = scoped_session(sessionmaker(autocommit=False. Put this into your application module: from yourapplication.remove() Here is an example table and model (put this into models.py): from sqlalchemy import Table. users) Querying and inserting works exactly the same as in the example above. unique=True) ) mapper(User. name=None.orm import scoped_session. autoflush=False. so make sure to also split up your application into multiple modules in a package. you need to close the session after each request or application context shutdown. db_session class User(object): query = db_session. Here is an example database.database import metadata. sessionmaker engine = create_engine('sqlite:////tmp/test. metadata.In general it works like the declarative approach.

u'admin'. name='admin'. 136 . u'admin'.db'. the good old problem of file uploads. autoload=True) To insert data you can use the insert method.select(users. We have to get a connection first so that we can use a transaction: >>> con = engine. head over to the website.c. 20.execute(). metadata.id == 1).c. convert_unicode=True) metadata = MetaData(bind=engine) Then you can either declare the tables in your code like in the examples above. or automatically load them: from sqlalchemy import Table users = Table('users'. Table engine = create_engine('sqlite:////tmp/test.first() >>> r['name'] u'admin' You can also pass strings of SQL statements to the execute() method: >>> engine.first() (1.execute(). u'admin@localhost') For more information about SQLAlchemy.9.20.id == 1).insert(). MetaData. To query your database. The basic idea of file uploads is actually quite simple. [1]). you use the engine directly or use a connection: >>> users. email='admin@localhost') SQLAlchemy will automatically commit for us.connect() >>> con.select(users.4 SQL Abstraction Layer If you just want to use the database system (and SQL) abstraction layer you basically only need the engine: from sqlalchemy import create_engine. It basically works like this: 1.execute('select * from users where id = :1'.first() (1. A <form> tag is marked with enctype=multipart/form-data and an <input type=file> is placed in that form.10 Uploading Files Ah yes. u'admin@localhost') These results are also dict-like tuples: >>> r = users.execute(users.

3.10. Let’s look at the bootstrapping code for our application: import os from flask import Flask. 'POST']) def upload_file(): if request. redirect. 'gif']) app = Flask(__name__) app. use the save() method of the file to save the file permanently somewhere on the filesystem.url) file = request. 'jpg'. methods=['GET'. 20.lower() in ALLOWED_EXTENSIONS @app.config['UPLOAD_FOLDER'] = UPLOAD_FOLDER So first we need a couple of imports.method == 'POST': # check if the post request has the file part if 'file' not in request. Also make sure to disallow .'. That way you can make sure that users are not able to upload HTML files that would cause XSS problems (see Cross-Site Scripting (XSS)).php files if the server executes them.1 A Gentle Introduction Let’s start with a very basic application that uploads a file to a specific upload folder and displays a file to the user. 'png'. Most should be straightforward. 'pdf'. Why do we limit the extensions that are allowed? You probably don’t want your users to be able to upload everything there if the server is directly sending out the data to the client.' in filename and \ filename. browser also # submit a empty part without filename if file.filename == '': 137 . request. 2. right? :) Next the functions that check if an extension is valid and that uploads the file and redirects the user to the URL for the uploaded file: def allowed_file(filename): return '.rsplit('.route('/'. but who has PHP installed on their server. 'jpeg'.files: flash('No file part') return redirect(request.files['file'] # if user does not select file. The application accesses the file from the files dictionary on the request object. secure_filename() is explained a little bit later. the werkzeug. url_for from werkzeug. 1)[1]. The UPLOAD_FOLDER is where we will store the uploaded files and the ALLOWED_EXTENSIONS is the set of allowed file exten- sions.utils import secure_filename UPLOAD_FOLDER = '/path/to/the/uploads' ALLOWED_EXTENSIONS = set(['txt'.

that is. In the upload_file() we redirect the user to url_for('uploaded_file'.path.filename) file.route('/uploads/<filename>') 138 . This does require some knowledge about how the application looks like./. flash('No selected file') return redirect(request. For the moment just remember: always use that function to secure a filename before storing it directly on the filesystem.bashrc" Assuming the number of .join(app.bashrc' Now one last thing is missing: the serving of the uploaded files././ is correct and you would join this with the UPLOAD_FOLDER the user might have the ability to modify a file on the server’s filesystem he or she should not modify./home/username/. All submitted form data can be forged./. hackers are patient :) Now let’s look how that function works: >>> secure_filename('.. Information for the Pros So you’re interested in what that secure_filename() function does and what the prob- lem is if you’re not using it? So just imagine someone would send the following infor- mation as filename to your application: filename = ". filename=filename)...5 we can use a function that does that for us: from flask import send_from_directory @app.save(os.. filename=filename)) return ''' <!doctype html> <title>Upload new File</title> <h1>Upload new File</h1> <form method=post enctype=multipart/form-data> <p><input type=file name=file> <input type=submit value=Upload> </form> ''' So what does that secure_filename() function actually do? Now the problem is that there is that principle called “never trust user input”. This is also true for the filename of an uploaded file.url) if file and allowed_file(file.bashrc') 'home_username_./. So we write the uploaded_file() function to return the file of that name. filename)) return redirect(url_for('uploaded_file'../. / uploads/filename. and filenames can be dan- gerous...config['UPLOAD_FOLDER'].filename): filename = secure_filename(file./home/username/.. but trust me../. As of Flask 0.

def uploaded_file(filename): return send_from_directory(app. So how exactly does Flask handle uploads? Well it will store them in the webserver’s memory if the files are reasonable small otherwise in a temporary location (as returned by tempfile. This feature was added in Flask 0.3 Upload Progress Bars A while ago many developers had the idea to read the incoming file in small chunks and store the upload progress in the database to be able to poll the progress with JavaScript from the client.add_url_rule('/uploads/<filename>'. filename) Alternatively you can register uploaded_file as build_only rule and use the SharedDataMiddleware. { '/uploads': app. build_only=True) app. For more information on that consult the Werkzeug documentation on file handling. Request app = Flask(__name__) app. 20.10. Long story short: the client asks the server every 5 seconds how much it has transmitted already. Do you realize the irony? The client is asking for something it should already know. but you can limit that by setting the MAX_CONTENT_LENGTH config key: from flask import Flask.6. 'uploaded_file'. This also works with older versions of Flask: from werkzeug import SharedDataMiddleware app.config['MAX_CONTENT_LENGTH'] = 16 * 1024 * 1024 The code above will limited the maximum allowed payload to 16 megabytes. If a larger file is transmitted. 139 .6 but can be achieved in older versions as well by subclassing the request object. Flask will raise an RequestEntityTooLarge exception.config['UPLOAD_FOLDER']. 20.wsgi_app = SharedDataMiddleware(app.wsgi_app.gettempdir()). But how do you specify the maximum file size after which an upload is aborted? By default Flask will happily accept file uploads to an un- limited amount of memory.2 Improving Uploads New in version 0.config['UPLOAD_FOLDER'] }) If you now run the application everything should work as expected.10.

Well.cache import SimpleCache cache = SimpleCache() If you want to use memcached.11.1:11211']) If you are using App Engine.contrib. So then the idea is that you actually put the result of that calculation into a cache for some time. Because the common pattern for file uploads exists almost unchanged in all applica- tions dealing with uploads.20.0. If you are using the development server you can create a SimpleCache object. similar to how Flask objects are created. There are JavaScript libraries like jQuery that have form plugins to ease the construction of progress bar. at least it’s the easiest way to speed up things.10. has some very basic cache support. This is how you connect to such an memcached server then: from werkzeug. make sure to have one of the memcache modules supported (you get them from PyPI) and a memcached server running somewhere. you can connect to the App Engine memcache server easily: from werkzeug. there is also a Flask extension called Flask-Uploads that implements a full fledged upload mechanism with white and blacklisting of exten- sions and more.contrib. It supports multiple cache backends.contrib. throw some caches in.11 Caching When your application runs slow.cache import GAEMemcachedCache cache = GAEMemcachedCache() 140 .4 An Easier Solution Now there are better solutions that work faster and are more reliable.0.1 Setting up a Cache You create a cache object once and keep it around. normally you want to use a memcached server. 20. What does a cache do? Say you have a function that takes some time to complete but the results would still be good enough if they were 5 min- utes old. that one is a simple cache that keeps the item stored in the memory of the Python interpreter: from werkzeug. 20.cache import MemcachedCache cache = MemcachedCache(['127. but Werkzeug. one of the libraries it is based on. Flask itself does not provide caching for you.

If something is in the cache. rv.get('my-item') To add items to the cache. timeout=5 * 60) return rv 20. Since the original function is replaced. This allows some really neat things for web applications.11.get('my-item') if rv is None: rv = calculate_value() cache.12 View Decorators Python has a really interesting feature called function decorators. This is how to use them: To get an item from the cache call get() with a string as key name. imagine you have a view that should only be used by people that are logged in. dec- orators can be used to inject additional functionality to one or more functions. it is returned.wraps() to handle this for you. you need to remember to copy the original function’s information to the new function. The first argument is the key and the second the value that should be set. A decorator is a function that wraps and replaces another function.2 Using a Cache Now how can one use such a cache? There are two very important operations: get() and set(). But there are use cases for implementing your own decorator. url_for def login_required(f): 141 . This example assumes that the login page is called 'login' and that the current user is stored in g. Here a full example how this looks like normally: def get_my_item(): rv = cache. 20. request.set('my-item'.1 Login Required Decorator So let’s implement such a decorator. they should be redirected to the login page. use the set() method instead. If a user goes to the site and is not logged in. Also a timeout can be provided after which the cache will automatically remove item.user and is None if there is no-one logged in. from functools import wraps from flask import g.12.20. Because each view in Flask is a function. Otherwise that function will return None: rv = cache. Use functools. This is a good example of a use case where a decorator is an excellent solution. redirect. For instance. The route() decorator is the one you probably used already.

@wraps(f) def decorated_function(*args. 3. Sounds aw- ful? Unfortunately it is a little bit more complex. always remember that the route() decorator is the outermost.user is None: return redirect(url_for('login'. get the value for that key from the cache.route('/secret_page') @login_required def secret_page(): pass Note: The next value will exist in request. then retrieve it from request. The decorated function will then work as follows 1. Here the code: 142 . Notice that we are using a function that first creates the decorator that then decorates the function. Here is an example cache function. You can do this with a hidden input tag.12. We’re assuming you have set up a cache like mentioned in Caching. You’ll have to pass it along when sending the POST request from the login form. get the unique cache key for the current request base on the current path. **kwargs): if g. When apply- ing further decorators. <input type="hidden" value="{{ request. It generates the cache key from a specific prefix (actually a format string) and the current path of the request. otherwise the original function is called and the return value is stored in the cache for the timeout provided (by default 5 minutes). '') }}"/> 20.get('next'.args. A decorator would be nice for that. apply it as innermost decorator to a view function. next=request.url)) return f(*args.form when logging the user in. @app.args after a GET request for the login page. 2. **kwargs) return decorated_function To use the decorator. If the cache returned something we will return that value.2 Caching Decorator Imagine you have a view function that does an expensive calculation and because of that you would like to cache the generated results for a certain amount of time. but the code should still be straight- forward to read.

When the decorated function returns. **kwargs): cache_key = key % request.path rv = cache.html'. key='view/%s'): def decorator(f): @wraps(f) def decorated_function(*args.set(cache_key. That way you can still use the redirect function or return simple strings. The idea of that decorator is that you return a dictionary with the values passed to the template from the view function and the template is automatically ren- dered.get(cache_key) if rv is not None: return rv rv = f(*args.route('/') @templated('index. the following three examples do exactly the same: @app. timeout=timeout) return rv return decorated_function return decorator Notice that this assumes an instantiated cache object is available.12. the dictionary returned is passed to the template rendering function. **kwargs) cache. If None is returned. Here is the code for that decorator: 143 . With that.from functools import wraps from flask import request def cached(timeout=5 * 60. rv. if something else than a dictionary is returned we return it from the function unchanged. see Caching for more information.route('/') @templated() def index(): return dict(value=42) As you can see. an empty dictionary is assumed. if no template name is provided it will use the endpoint of the URL map with dots converted to slashes + '.html') def index(): return dict(value=42) @app. 20. value=42) @app. Otherwise the provided template name is used.html'.route('/') def index(): return render_template('index.3 Templating Decorator A common pattern invented by the TurboGears guys a while back is a templating decorator.

144 .routing import Rule app = Flask(__name__) app. When you are working with WTForms you have to define your forms as classes first. This is possible with this decorator. **kwargs) if ctx is None: ctx = {} elif not isinstance(ctx.url_map. One of them is WTForms which we will handle here.endpoint \ . endpoint='index')) @app. code quickly becomes very hard to read. '/') + '.13 Form Validation with WTForms When you have to work with form data submitted by a browser view.12. **kwargs): template_name = template if template_name is None: template_name = request. If you find yourself in the situation of having many forms.html' ctx = f(*args.from functools import wraps from flask import request.replace('.4 Endpoint Decorator When you want to use the werkzeug routing system for more flexibility you need to map the endpoint as defined in the Rule to a view function.add(Rule('/'. There are libraries out there designed to make this process easier to manage. **ctx) return decorated_function return decorator 20. dict): return ctx return render_template(template_name. I recommend breaking up the application into multiple modules (Larger Applications) for that and adding a separate module for the forms. For example: from flask import Flask from werkzeug. render_template def templated(template=None): def decorator(f): @wraps(f) def decorated_function(*args.endpoint('index') def my_index(): return "Hello world" 20.'. you might want to give it a try.

route('/register'. but that’s not a requirement. the usage of this form looks like this: @app.1 The Forms This is an example form for a typical registration page: from wtforms import Form. [validators.username. methods=['GET'. StringField.Length(min=4.validate(): user = User(form. form=form) Notice we’re implying that the view is using SQLAlchemy here (SQLAlchemy in Flask). form. 20. Things to remember: 1.method == 'POST' and form.form) if request. call the validate() method. 145 .data) db_session. [validators. False otherwise. of course. message='Passwords must match') ]) confirm = PasswordField('Repeat Password') accept_tos = BooleanField('I accept the TOS'.13. 'POST']) def register(): form = RegistrationForm(request.Getting the most out of WTForms with an Extension The Flask-WTF extension expands on this pattern and adds a few little helpers that make working with forms and Flask more fun. form. You can get it from PyPI.<NAME>. BooleanField.EqualTo('confirm'. to validate the data. which will return True if the data validates. Adapt the code as necessary. to access individual values from the form.add(user) flash('Thanks for registering') return redirect(url_for('login')) return render_template('register. validators. PasswordField.password. max=25)]) email = StringField('Email Address'. max=35)]) password = PasswordField('New Password'. [validators.Length(min=6. access form.2 In the View In the view function. create the form from the request form value if the data is submitted via the HTTP POST method and args if the data is submitted as GET.data.data. 2. [ validators. 3.DataRequired()]) 20.email. validators class RegistrationForm(Form): username = StringField('Username'.html'.13.data.DataRequired().

To make it even nicer. class='username') to add a class to the input element. head over to the WTForms website. Here is the register.html" import render_field %} <form method=post> <dl> {{ render_field(form.html template for the function we used above. When you pass the form to the templates.errors %} <ul class=errors> {% for error in field. you can call render_field(form.username.3 Forms in Templates Now to the template side.confirm) }} {{ render_field(form.email) }} {{ render_field(form. which renders the field for us. Here’s an example _formhelpers. So.20.html template with such a macro: {% macro render_field(field) %} <dt>{{ field.label }} <dd>{{ field(**kwargs)|safe }} {% if field. you can easily render them there. Look at the following example template to see how easy this is. we can write a macro that renders a field with label and a list of errors if there are any. The keyword arguments will be inserted as HTML attributes. which takes ad- vantage of the _formhelpers.13. WTForms does half the form generation for us already.accept_tos) }} </dl> <p><input type=submit value=Register> </form> For more information about WTForms. for example.errors %} <li>{{ error }}</li> {% endfor %} </ul> {% endif %} </dd> {% endmacro %} This macro accepts a couple of keyword arguments that are forwarded to WTForm’s field function.password) }} {{ render_field(form.html template: {% from "_formhelpers. 146 .username) }} {{ render_field(form. Note that WTForms returns standard Python unicode strings. so we have to tell Jinja2 that this data is already HTML-escaped with the |safe filter.

1 Base Template This template.html" %} {% block title %}Index{% endblock %} {% block head %} {{ super() }} <style type="text/css"> . All the block tag does is tell the template engine that a child template may override those portions of the template.14 Template Inheritance The most powerful part of Jinja is template inheritance. which we’ll call layout.css') }}"> <title>{% block title %}{% endblock %} . filename='style. } </style> 147 . Copyright 2010 by <a href="http://domain.invalid/">you</a>.My Webpage</title> {% endblock %} </head> <body> <div id="content">{% block content %}{% endblock %}</div> <div id="footer"> {% block footer %} &copy.html. It’s easiest to understand it by starting with an example. defines a simple HTML skeleton docu- ment that you might use for a simple two-column page. It’s the job of “child” tem- plates to fill the empty blocks with content: <!doctype html> <html> <head> {% block head %} <link rel="stylesheet" href="{{ url_for('static'.20.14.2 Child Template A child template might look like this: {% extends "layout. Sounds complicated but is very basic. 20. 20.important { color: #336699.14. Template inheritance allows you to build a base “skeleton” template that contains all the common elements of your site and defines blocks that child templates can override. the {% block %} tags define four blocks that child templates can fill in. {% endblock %} </div> </body> </html> In this example.

20.1 Simple Flashing So here is a full example: from flask import Flask.method == 'POST': if request. 'POST']) def login(): error = None if request. first it locates the parent. It tells the template engine that this template “extends” another template.form['username'] != 'admin' or \ request. Note that browsers and sometimes web servers enforce a limit on cookie sizes. The extends tag must be the first tag in the template. If the user does not get enough feedback they will probably end up hating the application.{% endblock %} {% block content %} <h1>Index</h1> <p class="important"> Welcome on my awesome homepage.15 Message Flashing Good applications and user interfaces are all about feedback.form['password'] != 'secret': error = 'Invalid credentials' else: flash('You were successfully logged in') 148 .15. flash.route('/login'. render_template. {% endblock %} The {% extends %} tag is the key here. The flashing system basically makes it possible to record a message at the end of a request and access it next request and only next request. url_for app = Flask(__name__) app. When the template system evaluates this template. This is usually combined with a layout template that does this.secret_key = 'some_secret' @app. redirect.route('/') def index(): return render_template('index. To render the contents of a block defined in the parent template. \ request. Flask provides a really simple way to give feedback to a user with the flashing system.html') @app. This means that flashing messages that are too large for session cookies causes message flashing to fail silently. methods=['GET'. 20. use {{ super() }}.

html: {% extends "layout.html: {% extends "layout.html" %} {% block body %} <h1>Login</h1> {% if error %} <p class=error><strong>Error:</strong> {{ error }} {% endif %} <form method=post> <dl> <dt>Username: <dd><input type=text name=username value="{{ request. 149 . return redirect(url_for('index')) return render_template('login.html template which inherits from layout.username }}"> <dt>Password: <dd><input type=password name=password> </dl> <p><input type=submit value=Login> </form> {% endblock %} 20.html'.15. error=error) And here is the layout.html template which does the magic: <!doctype html> <title>My Application</title> {% with messages = get_flashed_messages() %} {% if messages %} <ul class=flashes> {% for message in messages %} <li>{{ message }}</li> {% endfor %} </ul> {% endif %} {% endwith %} {% block body %}{% endblock %} Here is the index.2 Flashing With Categories New in version 0.3.form.html template which also inherits from layout.html" %} {% block body %} <h1>Overview</h1> <p>Do you want to <a href="{{ url_for('login') }}">log in?</a> {% endblock %} And here is the login.

For example error messages could be displayed with a red background. To flash a message with a different category. 'error') Inside the template you then have to tell the get_flashed_messages() function to also return the categories. Optionally you can pass a list of categories which filters the results of get_flashed_messages(). The default cate- gory if nothing is provided is 'message'. Alternative categories can be used to give the user better feedback. {% with errors = get_flashed_messages(category_filter=["error"]) %} {% if errors %} <div class="alert-message block-message error"> <a class="close" href="#">×</a> <ul> {%.3 Filtering Flash Messages New in version 0. 20. This is useful if you wish to render each category in a sepa- rate block. One might also use the category to add a prefix such as <strong>Error:</strong> to the message. message in messages %} <li class="{{ category }}">{{ message }}</li> {% endfor %} </ul> {% endif %} {% endwith %} This is just one example of how to render these flashed messages.9. The loop looks slightly different in that situation then: {% with messages = get_flashed_messages(with_categories=true) %} {% if messages %} <ul class=flashes> {% for category. just use the second argument to the flash() function: flash(u'Invalid password provided'.15.for msg in errors %} <li>{{ msg }}</li> {% endfor -%} </ul> </div> {% endif %} {% endwith %} 150 .It is also possible to provide categories when flashing a message.

1/jquery. so how can we do that? A simple method would be to add a script tag to our page that sets a global variable to the prefix to the root of the application. but if we are using jQuery we should not hardcode the path to the application but make that dynamic. filename='jquery.min. But what if you later decide to move your application to a different location? For example to http://example. you have to download it first and place it in the static folder of your application and then ensure it’s loaded. Something like this: <script type=text/javascript> $SCRIPT_ROOT = {{ request.script_root|tojson|safe }}.1 Loading jQuery In order to use jQuery. This has the advantage that your website will probably load faster for users if they went to at least one other website before using the same jQuery version from Google because it will already be in the browser cache.16 AJAX with jQuery jQuery is a small JavaScript library commonly used to simplify working with the DOM and JavaScript in general. very similar to how Python primi- tives (numbers.js') }}">\x3C/script>')</script> In this case you have to put jQuery into your static folder as a fallback. strings. It is the perfect tool to make web applications more dynamic by exchanging JSON between server and client.js"></script> <script>window. but it will first try to load it directly from Google.googleapis. 20. filename='jquery.2 Where is My Site? Do you know where your application is? If you are developing the answer is quite simple: it’s on localhost port something and directly on the root of that server.com/myapp? On the server side this never was a problem because we were using the handy url_for() function that could answer that question for us. JSON itself is a very lightweight transport format.16.write('<script src="{{ url_for('static'. </script> 151 .9. Ideally you have a layout template that is used for all pages where you just have to add a script statement to the bottom of your <body> to load jQuery: <script type=text/javascript src="{{ url_for('static'.16.js') }}"></script> Another method is using Google’s AJAX Libraries API to load jQuery: <script src="//ajax.jQuery || document. dicts and lists) look like which is widely supported and very easy to parse.20.com/ajax/libs/jquery/1. It became popular a few years ago and quickly replaced XML as trans- port format in web applications. 20.

html') As you can see I also added an index method here that renders a template.route('/') def index(): return render_template('index. This tem- plate will load jQuery as above and have a little form we can add two numbers and a link to trigger the function on the server side. This makes it possible for Flask to automatically mark the result as HTML safe. Note that we are using the get() method here which will never fail. 152 . type=int) return jsonify(result=a + b) @app.args. This is a really ridiculous example and is something you usually would do on the client side alone. but we are inside a script block here where different rules apply. render_template. This is especially handy for code that is triggered by a script (APIs.get('a'. jsonify. 0. 20. 0. If the key is missing a default value (here 0) is returned. In Flask 0.args.The |safe is necessary in Flask before 0.route('/_add_numbers') def add_numbers(): a = request.get('b'. This also means that there must never be any </ between the script tags.16.) because you don’t need special error reporting in that case. |tojson is kind enough to do the right thing here and escape slashes for you ({{ "</script>"|tojson|safe }} is rendered as "<\/ script>"). but a simple example that shows how you would use jQuery and Flask nonetheless: from flask import Flask. type=int) b = request.10 it goes a step further and escapes all HTML tags with unicode escapes.3 JSON View Functions Now let’s create a server side function that accepts two URL arguments of numbers which should be added together and then sent back to the application in a JSON object. Everything until </script> is handled as script. Usually this would be necessary. JavaScript etc. Information for Pros In HTML the script tag is declared CDATA which means that entities will not be parsed. request app = Flask(__name__) @app.10 so that Jinja does not escape the JSON encoded string with HTML rules. Furthermore it can convert values to a specific type (like in our case int).

. Here’s the HTML code needed for our little application (index.4 The HTML Your index. func) sends a GET request to url and will send the con- tents of the data object as query parameters. or do that on the top. $. 153 . }). }). Note that we can use the $SCRIPT_ROOT variable here that we set earlier. 3.text(data.result). $(function() { . it will call the given function with the return value as argument. the default behavior will not kick in (in this case.bind('click'.16. It is usually a better idea to have that in a separate script file: <script type=text/javascript> $(function() { $('a#calculate').val(). $('selector') selects an element and lets you operate on it.getJSON($SCRIPT_ROOT + '/_add_numbers'. }) specifies code that should run once the browser is done loading the basic parts of the page. b: $('input[name="b"]').getJSON(url.html template either has to extend a layout. function(data) { $("#result"). element.. return false. { a: $('input[name="a"]'). If you don’t get the whole picture. If that function returns false.html). }).bind('event'.20. 4.val() }. func) specifies a function that should run when the user clicked on the element. just a very quick explanation of the little bit of code above: 1.html template with jQuery loaded and the $SCRIPT_ROOT variable set. Once the data arrived. function() { $. navigate to the # URL). </script> <h1>jQuery Example</h1> <p><input type=text size=5 name=a> + <input type=text size=5 name=b> = <span id=result>?</span> <p><a href=# id=calculate>calculate server side</a> I won’t go into detail here about how jQuery works. 2. Notice that we also drop the script directly into the HTML here. data. download the sourcecode for this example from GitHub.

Keep in mind that Flask will not set the error code for you. A terribly good idea is to have a nice page there.20. but it is called when an error happens and is passed that error. It’s a very good idea to make sure there is actually something useful on a 404 page.17. 410 Gone Did you know that there the “404 Not Found” has a brother named “410 Gone”? Few people actually implement that. So make sure the user is not lost when they try to access a forbidden resource. 20.1 Common Error Codes The following error codes are some that are often displayed to the user.17 Custom Error Pages Flask comes with a handy abort() function that aborts a request with an HTTP error code early. If you are not deleting documents permanently from the database but just mark them as deleted. you will have to send a 403 code for disallowed resources. at least a link back to the index. you made a mistake typing that URL” message. Here an example implementation for a “404 Page Not Found” exception: 154 . even if the application behaves correctly: 404 Not Found The good old “chap. do the user a favour and use the 410 code instead and display a message that what they were looking for was deleted for all eternity.2 Error Handlers An error handler is a function. The error is most likely a HTTPException. So common that even novices to the internet know that 404 means: damn. but in one case it can be a different error: a handler for internal server errors will be passed other exception instances as well if they are uncaught. so make sure to also provide the HTTP status code when returning a response. 403 Forbidden If you have some kind of access control on your website. the thing I was looking for is not there. An error handler is registered with the errorhandler() decorator and the error code of the exception. but nothing fancy. Depending on the error code it is less or more likely for the user to actually see such an error. because your application will fail sooner or later (see also: Application Errors). Please note that if you add an error handler for “500 Internal Server Error”. It will also provide a plain black and white error page for you with a basic description.17. 20. but the idea is that resources that previously existed and got deleted answer with 410 instead of 404. just like a view function. 500 Internal Server Error Usually happens on programming errors or if the server is overloaded. Flask will not trigger it if it’s running in Debug mode.

route('/user/<username>') def user(username): pass 155 .html" %} {% block title %}Page Not Found{% endblock %} {% block body %} <h1>Page Not Found</h1> <p>What you were looking for is just not there. Instead of using decorators. <p><a href="{{ url_for('index') }}">go somewhere nice</a> {% endblock %} 20.1 Converting to Centralized URL Map Imagine the current application looks somewhat like this: from flask import Flask app = Flask(__name__) @app.18 Lazily Loading Views Flask is usually used with the decorators.route('/') def index(): pass @app. This can be a problem if your application has to import quick.18. The system that enables having a central URL map is the add_url_rule() function.html').errorhandler(404) def page_not_found(e): return render_template('404.from flask import render_template @app. you have a file that sets up the application with all URLs. So if you suddenly notice that your application outgrows this approach you can fall back to a centralized URL mapping. Decorators are simple and you have the URL right next to the function that is called for that specific URL. However there is a downside to this approach: it means all your code that uses decorators has to be imported upfront or Flask will never actually find your function. 404 An example template might be this: {% extends "layout. 20. It might have to do that on systems like Google’s App Engine or other systems.

view(*args.user) 20. py) but without any decorator: def index(): pass def user(username): pass And then a file that sets up an application which maps the functions to URLs: from flask import Flask from yourapplication import views app = Flask(__name__) app. The trick is to actually load the view function as needed. Then you can define your central place to combine the views like this: from flask import Flask from yourapplication. *args.rsplit('.views. This can be accom- plished with a helper class that behaves just like a function but internally imports the real function on first use: from werkzeug import import_string. This is used by Flask internally to figure out how to name the URL rules in case you don’t provide a name for the rule yourself. cached_property class LazyView(object): def __init__(self.add_url_rule('/user/<username>'.index) app.2 Loading Late So far we only split up the views and the routing.add_url_rule('/'.add_url_rule('/'.helpers import LazyView app = Flask(__name__) app. import_name): self.18.'. but the module is still loaded up- front.import_name) def __call__(self.Then. self. view_func=views. view_func=views.__name__ = import_name.index')) 156 .__module__. view_func=LazyView('yourapplication. with the centralized approach you would have one file with the views (views.import_name = import_name @cached_property def view(self): return import_string(self. **kwargs): return self. **kwargs) What’s important here is is that __module__ and __name__ are properly set. 1) self.

['/']) # add two routes to a single function endpoint url_rules = ['/user/'. This pattern requires a running MongoDB server and the MongoKit library installed. and by wrapping view_func in a LazyView as needed.views. This pattern shows how to use MongoKit.add_url_rule(url_rule. I will outline each of them here: 20. def url(import_name. **options) # add a single route to the index view url('views. Here an example app. There are two very common ways to use MongoKit.index'.' + import_name) for url_rule in url_rules: app. url_rules=[]. url_rules) One thing to keep in mind is that before and after request handlers have to be in a file that is imported upfront to work properly on the first request.app.user'.user')) You can further optimize this in terms of amount of keystrokes needed to write this by having a function that calls into add_url_rule() by prefixing a string with the project name and a dot. **options): view = LazyView('yourapplication.1 Declarative The default behavior of MongoKit is the declarative one that is based on common ideas from Django or the SQLAlchemy declarative extension. 20.py module for your application: from flask import Flask from mongokit import Connection.'/user/<username>'] url('views.add_url_rule('/user/<username>'. a document mapper library. Document # configuration MONGODB_HOST = 'localhost' MONGODB_PORT = 27017 # create the little application object 157 . view_func=view. view_func=LazyView('yourapplication.19.19 MongoKit in Flask Using a document database rather than a full DBMS gets more common these days. to integrate with MongoDB. The same goes for any kind of remaining decorator.

This sometimes makes it more to type but also makes it blazingly fast. 'email': unicode. just subclass the Document class that is imported from Mon- goKit.name) # register the User document with our current connection connection. . MongoKit does not have something like a session. MongoKit is just schemaless too. 'email': max_length(120) } use_dot_notation = True def __repr__(self): return '<User %r>' % (self. If you’ve seen the SQLAlchemy pattern you may wonder why we do not have a session and even do not define a init_db function here.from_object(__name__) # connect to the database connection = Connection(app. MongoDB is schemaless.config['MONGODB_PORT']) To define your models.g.app = Flask(__name__) app. app.→format(length)) return validate class User(Document): structure = { 'name': unicode. Here is an example document (put this also into app. On the other hand. a valida- tor for the maximum character length and uses a special MongoKit feature called use_dot_notation.): from mongokit import ValidationError def max_length(length): def validate(value): if len(value) <= length: return True # must have %s in error format string to have mongokit place key in there raise ValidationError('%s must be at most {} characters long'. 158 .py. This means you can modify the data structure from one insert query to the next without any prob- lem. e. but implements some validation to ensure data integrity. } validators = { 'name': max_length(50). On the one hand. Per default MongoKit behaves like a python dictionary but with use_dot_notation set to True you can use your documents like you use models in nearly any other ORM by using dots to separate between attributes.config['MONGODB_HOST'].register([User]) This example shows you how to define your schema (named structure).config.

insert(user) MongoKit will automatically commit for us.User() >>> user['name'] = u'admin' >>> user['email'] = u'admin@localhost' >>> user.models import User >>> collection = connection['test'].users >>> user = collection.find_one({'name': u'admin'}) {u'_id': ObjectId('4c271729e13823182f000000').User. Note that this example does not show how to couple it with Flask. >>> collection = connection['test'].→'admin@localhost'} These results are also dict-like objects: 159 . u'email': u .You can insert entries into the database like this: >>> from yourapplication.User.19.find()) [<User u'admin'>] >>> collection. you must not use a com- mon str type for either name or email but unicode.users >>> user = {'name': u'admin'. this is somewhat the same as a table in the SQL world. u'email': u . 'email': u'admin@localhost'} >>> collection.2 PyMongo Compatibility Layer If you just want to use PyMongo. Querying is simple as well: >>> list(collection.database import connection >>> from yourapplication. u'name': u'admin'. We have to get a collection first. you use the collection directly: >>> list(collection. see the above MongoKit code for examples: from MongoKit import Connection connection = Connection() To insert data you can use the insert method. u'name': u'admin'.→'admin@localhost'}] >>> collection. You may use this process if you need the best performance to get. you can do that with MongoKit as well.save() Note that MongoKit is kinda strict with used column types.find_one({'name': u'admin'}) <User u'admin'> 20. To query your database.find()) [{u'_id': ObjectId('4c271729e13823182f000000').

with this name. This is not a requirement but a de-facto standard supported by all relevant browsers.ico')) If you want to save the extra redirect request you can also write a view using send_from_directory(): import os from flask import send_from_directory @app.ico') }}"> That’s all you need for most browsers. but we may as well specify it to avoid the extra guessing. 'favicon.join(app. The above will serve the icon via your application and if possible it’s better to config- ure your dedicated web server to serve it.ico') def favicon(): return send_from_directory(os. of course. filename='favicon. filename='favicon. to get browsers to find your icon. This helps to distin- guish your website and to give it a unique brand. Now. It should be 16 × 16 pixels and in the ICO file format. head over to the website. refer to the web server’s documentation. Put the icon in your static directory as favicon. as it will always be the same.find_one({'name': u'admin'}) >>> r['email'] u'admin@localhost' For more information about MongoKit. First. however some really old ones do not support this standard. If your application is not mounted at the root path of the domain you either need to configure the web server to serve the icon at the root or if you can’t do that you’re out of luck.path.add_url_rule('/favicon.microsoft. you need an icon.>>> r = collection.20 Adding a favicon A “favicon” is an icon used by browsers for tabs and bookmarks.ico. 20.root_path. If however your application is the root you can simply route a redirect: app. 'static').icon') We can leave out the explicit mimetype and it will be guessed.route('/favicon.ico'. mimetype='image/vnd. A common question is how to add a favicon to a Flask application.ico'. The old de-facto standard is to serve this file. for example: <link rel="shortcut icon" href="{{ url_for('static'. 160 . So. at the website root. redirect_to=url_for('static'. the correct way is to add a link tag in your HTML.

so be careful there in debug environments with profilers and other things you might have enabled.update_template_context(context) t = app. The trick is to have an inner function that uses a generator to generate data and to then invoke that function and pass it to a response object: from flask import Response @app.jinja_env. When you are generating the data on the fly though. much more than you want to keep in memory.get_template(template_name) rv = t.21. 20.1 See also • The Favicon article on Wikipedia 20. 20.21.20. how do you send that back to the client without the roundtrip to the filesystem? The answer is by using generators and direct responses.route('/large. This functionality is not directly exposed by Flask because it is quite uncommon.20.1 Basic Usage This is a basic view function that generates a lot of CSV data on the fly.'.enable_buffering(5) return rv 161 . but you can easily do it yourself: from flask import Response def stream_template(template_name. **context): app.csv') def generate_large_csv(): def generate(): for row in iter_all_rows(): yield '.2 Streaming from Templates The Jinja2 template engine also supports rendering templates piece by piece.21 Streaming Contents Sometimes you want to send an enormous amount of data to the client. Note though that some WSGI middlewares might break streaming. mimetype='text/csv') Each yield expression is directly sent to the browser.stream(context) rv.join(row) + '\n' return Response(generate().

enable_buffering(size). 5 is a sane default. The template is then evaluated as the stream is iterated over. Since each time you do a yield the server will flush the content to the client you might want to buffer up a few items in the template which you can do with rv.route('/stream') def streamed_response(): def generate(): yield 'Hello ' yield request. Response @app.args['name'] yield '!' return Response(stream_with_context(generate())) Without the stream_with_context() function you would get a RuntimeError at that point.html') def render_large_template(): rows = iter_all_rows() return Response(stream_template('the_template. Note that when you stream data. It is created as necessary either by a view function or by some other component in the system. 20. One way is to avoid the situation.html'.@app.9. there is no response object yet.22 Deferred Request Callbacks One of the design principles of Flask is that response objects are created and passed down a chain of potential callbacks that can modify them or replace them. Sometimes however moving 162 . request. When the request handling starts.9 provides you with a helper that can keep the request context around during the execution of the generator: from flask import stream_with_context. 20. the request context is already gone the moment the function executes. rows=rows)) The trick here is to get the template object from the Jinja2 environment on the appli- cation and to call stream() instead of render() which returns a stream object instead of a string. Since we’re bypassing the Flask template render functions and using the template object itself we have to make sure to update the render context ourselves by calling update_template_context(). Flask 0. But what happens if you want to modify the response at a point where the response does not exist yet? A common example for that would be a before-request function that wants to set a cookie on the response object.3 Streaming with Context New in version 0. Very often that is possible. For instance you can try to move that logic into an after-request callback instead.21.route('/my-large-page.

But we still need to call them. 'after_request_callbacks'.1 The Decorator The following decorator is the key. It registers a function on a list on the g object: from flask import g def after_this_request(f): if not hasattr(g.22. This way you can defer code execution from anywhere in the application.after_request def call_after_request_callbacks(response): for callback in getattr(g.before_request def detect_user_language(): language = request.cookies. we can register a function to be called at the end of the request.3 A Practical Example At any time during a request.append(f) return f 20.22. 20. For this the following function needs to be registered as after_request() callback: @app. ()): callback(response) return response 20.that code there is just not a very pleasant experience or makes code look very awk- ward. As an alternative possibility you can attach a bunch of callback functions to the g object and call them at the end of the request. For example you can remember the current language of the user in a cookie in the before-request function: from flask import request @app.get('user_lang') if language is None: language = guess_language_from_request() @after_this_request def remember_language(response): 163 .2 Calling the Deferred Now you can use the after_this_request decorator to mark a function to be called at the end of the request.after_request_callbacks.22.after_request_callbacks = [] g. 'after_request_callbacks'): g.

response. 'OPTIONS' ]) bodyless_methods = frozenset(['GET'.bodyless_methods: environ['CONTENT_LENGTH'] = '0' return self. 'replace') environ['REQUEST_METHOD'] = method if method in self. 'POST'. 'PATCH'. 'DELETE']) def __init__(self.app = app def __call__(self.language = language 20.23 Adding HTTP Method Overrides Some HTTP proxies do not support arbitrary HTTP methods or newer HTTP methods (such as PATCH). 'HEAD'. environ. This can easily be accomplished with an HTTP middleware: class HTTPMethodOverrideMiddleware(object): allowed_methods = frozenset([ 'GET'. start_response): method = environ.allowed_methods: method = method. In that case it’s possible to “proxy” HTTP methods through another HTTP method in total violation of the protocol. 'DELETE'. language) g.get('HTTP_X_HTTP_METHOD_OVERRIDE'.app(environ. '').wsgi_app) 164 .wsgi_app = HTTPMethodOverrideMiddleware(app. The way this works is by letting the client do an HTTP POST request and set the X-HTTP-Method-Override header and set the value to the intended HTTP method (such as PATCH). 'PUT'. 'HEAD'.set_cookie('user_lang'. 'OPTIONS'.encode('ascii'.upper() if method in self. start_response) To use this with Flask this is all that is necessary: from flask import Flask app = Flask(__name__) app. app): self.

bytes): rv = self.readline(size_hint) self.form or anything of that nature._hash To use this._stream.hexdigest() return 'Hash was: %s' % checksum 165 .update(rv) return rv def readline(self._hash._hash = hashlib. stream): self. checksum = hash._stream. The following example calculates the SHA1 checksum of the incoming data as it gets read and stores it in the WSGI environment: import hashlib class ChecksumCalcStream(object): def __init__(self.read(bytes) self.environ stream = ChecksumCalcStream(env['wsgi.sha1() def read(self._stream = stream self.input']) env['wsgi. Example usage: @app. methods=['POST']) def special_api(): hash = generate_checksum(request) # Accessing this parses the input stream files = request. This seems inconvenient when you want to calculate the checksum of the incoming request data. before_request_handlers for instance should be careful not to access it).input'] = stream return stream.20. Fortunately this is however very simple to change by wrapping the input stream.24 Request Content Checksums Various pieces of code can consume the request data and preprocess it. (Eg: be careful accessing request. all you need to do is to hook the calculating stream in before the request starts consuming data. This is necessary sometimes for some APIs. size_hint): rv = self.route('/special-api'.update(rv) return rv def generate_checksum(request): env = request.files # At this point the hash is fully constructed. For instance JSON data ends up on the request object already read and processed. form data ends up there as well but goes through a different code path._hash.

config['CELERY_RESULT_BACKEND'].25 Celery Based Background Tasks Celery is a task queue for Python with batteries included. configures it with the broker from the ap- plication config.update(app. **kwargs): with app.Task = ContextTask return celery The function creates a new Celery object. *args.25. updates the rest of the Celery config from the Flask config and then creates a subclass of the task that wraps the task execution in an application context. it must be possible for other modules to import it. It used to have a Flask inte- gration but it became unnecessary after some restructuring of the internals of Celery with Version 3. this is called the celery application. **kwargs) celery.config['CELERY_BROKER_URL']) celery.config) TaskBase = celery.__call__(self. This guide fills in the blanks in how to properly use Celery with Flask but assumes that you generally already read the First Steps with Celery guide in the official Celery documentation.app_context(): return TaskBase. While you can use Celery without any reconfiguration with Flask. 166 . broker=app. so it can be installed with standard Python tools like pip or easy_install: $ pip install celery 20. For instance you can place this in a tasks module. *args. just for Celery. backend=app.Task class ContextTask(TaskBase): abstract = True def __call__(self.conf. like creating tasks and managing workers. It serves the same purpose as the Flask object in Flask.1 Installing Celery Celery is on the Python Package Index (PyPI). it becomes a bit nicer by subclassing tasks and adding support for Flask’s application contexts and hooking it up with the Flask configura- tion. 20.import_name.25.2 Configuring Celery The first thing you need is a Celery instance.20. Since this instance is used as the entry-point for everything you want to do in Celery. This is all that is necessary to properly integrate Celery with Flask: from celery import Celery def make_celery(app): celery = Celery(app.

20.datastructures import ImmutableOrderedMultiDict class MyRequest(Request): """Request subclass to override request parameter storage""" parameter_storage_class = ImmutableOrderedMultiDict class MyFlask(Flask): 167 . For example.task() def add_together(a. b): return a + b This task can now be called in the background: >>> result = add_together. CELERY_RESULT_BACKEND='redis://localhost:6379' ) celery = make_celery(flask_app) @celery. Request from werkzeug.config.wait() will never actually return.25. 20.25.wait() 65 20.26 Subclassing Flask The Flask class is designed for subclassing.4 Running the Celery Worker Now if you jumped in and already executed the above code you will be disappointed to learn that your .update( CELERY_BROKER_URL='redis://localhost:6379'.celery worker The your_application string has to point to your application’s package or module that creates the celery object. You can do that by running celery as a worker: $ celery -A your_application.3 Minimal Example With what we have above this is the minimal example of using Celery with Flask: from flask import Flask flask_app = Flask(__name__) flask_app.delay(23. 42) >>> result. you may want to override how request parameters are handled to pre- serve their order: from flask import Flask. That’s because you also need to run celery.

"""Flask subclass using the custom request class""" request_class = MyRequest This is the recommended approach for overriding or augmenting Flask’s internal func- tionality. 168 .

run() calls you might have in your applica- tion file are inside an if __name__ == '__main__': block or moved to a separate file. Watch Out Please make sure in advance that any app. look up the server documentation about how to use a WSGI app with it.1 mod_wsgi (Apache) If you are using the Apache webserver. Some of the options available for properly running Flask in production are documented here.2 Self-hosted options 21. 21. consider using mod_wsgi. 169 . CHAPTER TWENTYONE DEPLOYMENT OPTIONS While lightweight and easy to use. If you want to deploy your Flask application to a WSGI server not listed here.1 Hosted options • Deploying Flask on Heroku • Deploying Flask on OpenShift • Deploying Flask on Webfaction • Deploying Flask on Google App Engine • Deploying Flask on AWS Elastic Beanstalk • Sharing your Localhost Server with Localtunnel • Deploying on Azure (IIS) • Deploying on PythonAnywhere 21. Flask’s built-in server is not suitable for produc- tion as it doesn’t scale well and by default serves only one request at a time. Just remember that your Flask application object is the actual WSGI application.2.

wsgi file before the import: 170 .) you can install it as follows: # yum install mod_wsgi On FreeBSD install mod_wsgi by compiling the www/mod_wsgi port or by using pkg_add: # pkg install ap22-mod_wsgi2 If you are using pkgsrc you can install mod_wsgi by compiling the www/ap2-wsgi pack- age.. Alternatively there is the option to just patch the path in the .wsgi file. OpenSUSE.: /var/www/yourapplication) and make sure that yourapplication and all the libraries that are in use are on the python load path. Creating a . Installing mod_wsgi If you don’t have mod_wsgi installed yet you have to either install it using a package manager or compile it yourself. Keep in mind that you will have to actually install your application into the virtualenv as well. If you don’t want to install it system wide consider using a virtual python instance. Just restart the server.wsgi file To run your application you need a yourapplication.Just make sure it’s not called because this will always start a local WSGI server which we do not want if we deploy that application to mod_wsgi. For most applications the following file should be sufficient: from yourapplication import app as application If you don’t have a factory function for application creation but a singleton instance you can directly import that one as application.g. This file contains the code mod_wsgi is executing on startup to get the application object. The object called application in that file is then used as application. Store that file somewhere that you will find it again (e. etc. The mod_wsgi installation instructions cover source installations on UNIX systems. If you encounter segfaulting child processes after the first apache reload you can safely ignore them. If you are using Ubuntu/Debian you can apt-get it and activate it as follows: # apt-get install libapache2-mod-wsgi If you are using a yum based distribution (Fedora.

On a Windows system. the syntax for directory permissions has changed from httpd 2.deny Allow from all to httpd 2. '/path/to/the/application') Configuring Apache The last thing you have to do is to create an Apache configuration file for your ap- plication. eliminate those lines: <VirtualHost *> ServerName example.wsgi <Directory /var/www/yourapplication> WSGIProcessGroup yourapplication WSGIApplicationGroup %{GLOBAL} Order deny.import sys sys.path.com WSGIScriptAlias / C:\yourdir\yourapp.allow Allow from all </Directory> </VirtualHost> Note: There have been some changes in access control configuration for Apache 2. 171 .4.allow Allow from all </Directory> </VirtualHost> Note: WSGIDaemonProcess isn’t implemented in Windows and Apache will refuse to run with the above configuration.insert(0. Most notably.com WSGIDaemonProcess yourapplication user=user1 group=group1 threads=5 WSGIScriptAlias / /var/www/yourapplication/yourapplication.2 Order allow. In this example we are telling mod_wsgi to execute the application under a different user for security reasons: <VirtualHost *> ServerName example.wsgi <Directory C:\yourdir> Order deny.4 syntax Require all granted For more information consult the mod_wsgi documentation.

If you want to use a virtual environment with mod_wsgi you have to modify your . Support for Automatic Reloading To help deployment tools you can activate support for automatic reloading.py file you symlinked into the site-packages folder.run() call in your application file that is not guarded by an if __name__ == '__main__': condition. Make sure the folders the application needs access to have the proper privileges set and the application runs as the correct user (user and group parameter to the WSGIDaemonProcess directive) Problem: application dies with an error on print Keep in mind that mod_wsgi dis- allows doing anything with sys. mod_wsgi will reload all the daemon processes for us. You can disable this protection from the config by setting the WSGIRestrictStdout to off: WSGIRestrictStdout Off Alternatively you can also replace the standard out in the .stderr Problem: accessing resources gives IO errors Your application probably is a single .wsgi file. instead you either have to put the folder into the pythonpath the file is stored in. the module filename is used to locate the resources and for symlinks the wrong filename is picked up. errorlog shows SystemExit ignored You have an app. Problem: application gives permission errors Probably caused by your application running as the wrong user.wsgi file slightly. 172 .stdout and sys. just add the following directive to your Directory section: WSGIScriptReloading On Working with Virtual Environments Virtual environments have the advantage that they never install the required depen- dencies system wide so you have a better control over what is used where. Either remove that run() call from the file and move it into a separate run. follow this guide to troubleshoot: Problem: application does not run. or convert your application into a package. The reason for this is that for non-installed packages. Whenever something changes the . Please be aware that this does not work.Troubleshooting If your application does not run.stderr.wsgi file with a differ- ent stream: import sys sys. For that.stdout = sys.py file or put it into such an if block.

It’s a pre-fork worker model ported from Ruby’s Unicorn project.2.py' with open(activate_this) as file_: exec(file_.wsgi import WSGIServer from yourapplication import app http_server = WSGIServer((''.2 Standalone WSGI Containers There are popular servers written in Python that contain WSGI applications and serve HTTP. For example. Keep in mind that the path has to be absolute. Gunicorn Gunicorn ‘Green Unicorn’ is a WSGI HTTP Server for UNIX. app) http_server.0.0.0. These servers stand alone when they run.Add the following lines to the top of your . 21.0. dict(__file__=activate_this)) For Python 3 add the following lines to the top of your .serve_forever() 173 .read(). Note the section on Proxy Setups if you run into issues.wsgi file: activate_this = '/path/to/env/bin/activate_this. 5000). to run a Flask application with 4 worker processes (-w 4) binding to localhost port 4000 (-b 127.1:4000 myproject:app Gevent Gevent is a coroutine-based Python networking library that uses greenlet to provide a high-level synchronous API on top of libev event loop: from gevent. Running a Flask application on this server is quite simple: gunicorn myproject:app Gunicorn provides many command-line options – see gunicorn -h.1:4000): gunicorn -w 4 -b 127.py' execfile(activate_this. dict(__file__=activate_this)) This sets up the load paths according to the settings of the virtual environment. you can proxy to them from your web server.wsgi file: activate_this = '/path/to/env/bin/activate_this. It supports both eventlet and greenlet.

error_log /var/log/nginx/error. or you can fix them in middleware. location / { proxy_pass http://127.log. Here’s a simple nginx configuration which proxies to an application served on local- host at port 8000. } } If your httpd is not providing these headers. on port 8080. Twisted Web comes with a standard WSGI container which can be controlled from the command line using the twistd utility: twistd web --wsgi myproject. For example. proxy_redirect off. proxy_set_header X-Forwarded-Proto $scheme. server_name _.0.app Proxy Setups If you deploy your application using one of these servers behind an HTTP proxy you will need to rewrite a few headers in order for the application to work.log. proxy_set_header Host $host. Twisted Web supports many flags and options. setting appropriate headers: server { listen 80. non-blocking event- driven networking library. a mature.Twisted Web Twisted Web is the web server shipped with Twisted. and the twistd utility does as well. see twistd -h and twistd web -h for more information. The two prob- lematic values in the WSGI environment usually are REMOTE_ADDR and HTTP_HOST. Werkzeug ships a fixer that will solve some common setups. but you might want to write your own WSGI middleware for specific setups. to run a Twisted Web server in the foreground.1:8000/.app This example will run a Flask application called app from a module named myproject. You can configure your httpd to pass these headers.0. the most common setup invokes the host being set from X-Forwarded-Host and the remote address from X-Forwarded-For: 174 . access_log /var/log/nginx/access. with an application from myproject: twistd -n web --port 8080 --wsgi myproject. proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for. proxy_set_header X-Real-IP $remote_addr.

app = app def __call__(self. see FastCGI and Standalone WSGI Containers for other options.run() calls you might have in your applica- tion file are inside an if __name__ == '__main__': block or moved to a separate file. To use your WSGI appli- cation with uWSGI protocol you will need a uWSGI server first. 175 .app(environ.wsgi_app = ProxyFix(app.fixers import ProxyFix app. environ. '') if host: environ['HTTP_HOST'] = host return self. Make sure to have it installed to follow along. start_response) app.wsgi_app) Trusting Headers Please keep in mind that it is a security issue to use such a middleware in a non-proxy setup because it will blindly trust the incoming headers which might be forged by malicious clients. the application server can serve uWSGI. and cherokee. Just make sure it’s not called because this will always start a local WSGI server which we do not want if we deploy that application to uWSGI. If you want to rewrite the headers from another header.contrib.3 uWSGI uWSGI is a deployment option on servers like nginx.wsgi_app = CustomProxyFix(app. lighttpd.wsgi_app) 21. The most popular uWSGI server is uwsgi. uWSGI is both a protocol and an application server. start_response): host = environ. and HTTP protocols. Watch Out Please make sure in advance that any app.from werkzeug. app): self.get('HTTP_X_FHOST'. you might want to use a fixer like this: class CustomProxyFix(object): def __init__(self.2. FastCGI. which we will use for this guide.

Configuring nginx A basic flask nginx configuration looks like this: location = /yourapplication { rewrite ^ /yourapplication/. use the following command: $ uwsgi -s /tmp/yourapplication. app is the callable inside of your application (usually the line reads app = Flask(__name__). } location @yourapplication { include uwsgi_params.sock --manage-script-name --mount / .sock. and cherokee.Starting your app with uwsgi uwsgi is designed to operate on WSGI callables found in python modules. uwsgi_pass unix:/tmp/yourapplication. since its smarter about that. To use your WSGI applica- tion with any of them you will need a FastCGI server first. you can use a single / instead of /yourapplication. } location /yourapplication { try_files $uri @yourapplication. } location @yourapplication { include uwsgi_params. Given a flask application in myapp. } 21. uwsgi_pass unix:/tmp/yourapplication. } This configuration binds the application to /yourapplication. You might also need to add --plugin python or --plugin python3 depending on which python version you use for your project. The most popular one is flup which we will use for this guide. It is used together with the --mount directive which will make requests to /yourapplication be directed to myapp:app. Make sure to have it installed to follow along.py. 176 . myapp refers to the name of the file of your flask application (without extension) or the module which provides app.→yourapplication=myapp:app The --manage-script-name will move the handling of SCRIPT_NAME to uwsgi. see uWSGI and Standalone WSGI Containers for other options. If your application is accessi- ble at root level.4 FastCGI FastCGI is a deployment option on servers like nginx. you need to also add --virtualenv /path/to/virtual/environment. If you want to have it in the URL root its a bit simpler: location / { try_files $uri @yourapplication.sock.2. lighttpd. If you want to deploy your flask application inside of a virtual environment.

Let’s call it yourapplication. however nginx and older versions of lighttpd need a socket to be explicitly passed to communicate with the FastCGI server. Make sure to set the executable bit on that file so that the servers can execute it: # chmod +x /var/www/yourapplication/yourapplication.fcgi Configuring Apache The example above is good enough for a basic Apache deployment but your . The following example uses FastCgiServer to start 5 instances of the application which will handle all incoming requests: LoadModule fastcgi_module /usr/lib64/httpd/modules/mod_fastcgi.fcgi import WSGIServer from yourapplication import app if __name__ == '__main__': WSGIServer(app).Watch Out Please make sure in advance that any app. There are few ways to configure your application so that yourapplication.com/yourapplication. Save the yourapplication.g.fcgi does not appear in the URL.fcgi -idle-timeout 300 -processes 5 177 . bindAddress='/path/to/fcgi. example.run() The path has to be the exact same path you define in the server config.sock'). Just make sure it’s not called because this will always start a local WSGI server which we do not want if we deploy that application to FastCGI.server.run() calls you might have in your applica- tion file are inside an if __name__ == '__main__': block or moved to a separate file. It makes sense to have that in /var/www/yourapplication or something similar. Creating a . A preferable way is to use the ScriptAlias and SetHandler config- uration directives to route requests to the FastCGI server. For that to work you need to pass the path to the socket to the WSGIServer: WSGIServer(application.fcgi: #!/usr/bin/python from flup.fcgi file somewhere you will find it again.fcgi/news/.so FastCgiServer /var/www/html/yourapplication/app.run() This is enough for Apache to work.fcgi file will appear in your application URL e.fcgi file First you need to create the FastCGI server file.

*)$ yourapplication.fcgi <Files ~ (\.htaccess: <IfModule mod_fcgid.server.fcgi/$1 [QSA. you can use the FastCgiExternalServer directive instead. you can use WSGI middleware to remove yourapplication. app): 178 .insert(0. '<your_local_path>/lib/python2.<VirtualHost *> ServerName webapp1. for example on a shared web host.fcgi/ <Location /> SetHandler fastcgi-script </Location> </VirtualHost> These processes will be managed by Apache.com DocumentRoot /var/www/html/yourapplication AddHandler fastcgi-script fcgi ScriptAlias / /var/www/html/yourapplication/app.0.0. If you’re using a standalone FastCGI server.6/site-packages') from flup. Note that in the fol- lowing the path is not real.fcgi)> SetHandler fcgid-script Options +FollowSymLinks +ExecCGI </Files> </IfModule> <IfModule mod_rewrite.c> Options +FollowSymlinks RewriteEngine On RewriteBase / RewriteCond %{REQUEST_FILENAME} !-f RewriteRule ^(.L] </IfModule> Set yourapplication. it’s simply used as an identifier to other directives such as AliasMatch: FastCgiServer /var/www/html/yourapplication -host 127.fcgi: #!/usr/bin/python #: optional path to your local python site-packages folder import sys sys. Set .fcgi import WSGIServer from yourapplication import app class ScriptNameStripper(object): def __init__(self.1:3000 If you cannot set ScriptAlias.c> AddHandler fcgid-script .mydomain.path.fcgi from the URLs.

If you want the application to work in the URL root you have to work around a lighttpd bug with the LighttpdCGIRootFix middle- ware.rewrite-once = ( "^(/static($|/.fcgi". Also. start_response): environ['SCRIPT_NAME'] = '' return self.*)$" => "/yourapplication. alias and rewrite modules. A basic Flask FastCGI configuration for nginx looks like this: 179 .app = app def __call__(self. environ.run() Configuring lighttpd A basic FastCGI configuration for lighttpd looks like that: fastcgi.sock". self. This configuration binds the application to /yourapplication.fcgi" => (( "socket" => "/tmp/yourapplication-fcgi.*))$" => "$1". "check-local" => "disable". Configuring nginx Installing FastCGI applications on nginx is a bit different because by default no FastCGI parameters are forwarded. Make sure to apply it only if you are mounting the application the URL root.app(environ.url = ( "/static/" => "/path/to/your/static" ) url.fcgi$1" ) Remember to enable the FastCGI.server = ("/yourapplication. "bin-path" => "/var/www/yourapplication/yourapplication. "max-procs" => 1 )) ) alias. see the Lighty docs for more information on FastCGI and Python (note that explicitly passing a socket to run() is no longer necessary). start_response) app = ScriptNameStripper(app) if __name__ == '__main__': WSGIServer(app). "^(/.

fastcgi_pass unix:/tmp/yourapplication-fcgi. Very often the only thing the server log tells you is something along the lines of “premature end of headers”. you have to do it by yourself. and note that this is a manual solution which does not persist across system restart: $ screen $ /var/www/yourapplication/yourapplication. } location @yourapplication { include fastcgi_params.fcgi script inside GNU screen. using a SysV init.d script. } This configuration binds the application to /yourapplication. you can always run the . See man screen for details. fastcgi_param SCRIPT_NAME "".g. You can look around for other FastCGI process managers or write a script to run your . fastcgi_pass unix:/tmp/yourapplication-fcgi. } Running FastCGI Processes Since nginx and others do not load FastCGI apps. fastcgi_param SCRIPT_NAME $fastcgi_script_name. e. fastcgi_param PATH_INFO $fastcgi_script_name.fcgi and that your web server user is www-data: $ su www-data $ cd /var/www/yourapplication 180 . Super- visor can manage FastCGI processes.sock. fastcgi_split_path_info ^(/yourapplication)(. This example assumes your application is called application. } location /yourapplication { try_files $uri @yourapplication. In order to debug the application the only thing that can really give you ideas why it breaks is switching to the correct user and executing the application by hand.fcgi Debugging FastCGI deployments tend to be hard to debug on most web servers. } location @yourapplication { include fastcgi_params.*)$.location = /yourapplication { rewrite ^ /yourapplication/ last.fcgi file at boot. If you want to have it in the URL root it’s a bit simpler because you don’t have to figure out how to calculate PATH_INFO and SCRIPT_NAME: location / { try_files $uri @yourapplication. For a temporary solution.sock. fastcgi_param PATH_INFO $fastcgi_path_info.

CGI is sup- ported by all major servers but usually has a sub-optimal performance. 21.handlers import CGIHandler from yourapplication import app CGIHandler(). Creating a .cgi: #!/usr/bin/python from wsgiref.stdout is overridden by something that doesn’t write into the HTTP response.$ python application. you will also have to make sure that your code does not contain any print statements. in <module> ImportError: No module named yourapplication In this case the error seems to be “yourapplication” not being on the python path.run() calls you might have in your applica- tion file are inside an if __name__ == '__main__': block or moved to a separate file. where execution happens in a CGI-like environment.cgi into a cgi-bin (and use mod_rewrite or something similar to rewrite the URL) or let the 181 . Let’s call it yourapplication. • The code depending on environment variables that are not set by the web server. • Different python interpreters being used.cgi file First you need to create the CGI application file. Common problems are: • Relative paths being used.5 CGI If all other deployment methods do not work. Either just copy the . Watch Out Please make sure in advance that any app.fcgi Traceback (most recent call last): File "yourapplication. Don’t rely on the current working directory.run(app) Server Setup Usually there are two ways to configure the server.fcgi". This is also the way you can use a Flask application on Google’s App Engine. CGI will work for sure.2. line 4. Just make sure it’s not called because this will always start a local WSGI server which we do not want if we deploy that application to CGI / app engine. With CGI. or that sys.

server point to the file directly. In Apache for example you can put something like this into the config: ScriptAlias /app /path/to/the/application. sitting in the public directory you want your app to be available.cgi On shared webhosting.*)$ /path/to/the/application.cgi/$1 [L] For more information consult the documentation of your webserver. 182 . though. you might not have access to your Apache config. works too but the ScriptAlias directive won’t work in that case: RewriteEngine On RewriteCond %{REQUEST_FILENAME} !-f # Don't interfere with static files RewriteRule ^(.htaccess. In this case. a file called .

and Signals. it became useful to a wide audience. hook points. 22. Dig deeper on the APIs you use.3 Subclass. The Flask class has many methods designed for subclassing. You can quickly add or customize behavior by subclassing Flask (see the linked method docs) and using that subclass wherever you instantiate an application class. CHAPTER TWENTYTWO BECOMING BIG Here are your options when growing your codebase or scaling your application. and documents its internal utilities so that you can find the hook points needed for your project. Read the source. This works well with Applica- tion Factories. 22.1 Read the Source. Explore the many extensions in the community. Flask’s code is written to be read. Flask started in part to demonstrate how to build your own framework on top of existing well-used tools Werkzeug (WSGI) and Jinja (templating). Look for ways in which your project can be refactored into a collection of utilities and Flask extensions. The API docs are full of available overrides. and look for the customizations which are available out of the box in a Flask release.2 Hook. Flask sticks to documented APIs in upstream libraries. and as it developed. As you grow your codebase. You can provide custom classes for things like the request and response objects. don’t just use Flask – understand it. and look for patterns to build your own extensions if you do not find the tools you need. its documentation is published so you can use its internal APIs. 22. See Subclassing Flask for an example. 183 . Extend.

forking should be the very last resort. fork Flask. depending on the number of changes. the data store you want to use and the Python implementation and webserver you are running on.22.6 Scale like a pro. You can introduce WSGI middleware to wrap your Flask instances and introduce fixes and changes at the layer between your Flask application and your HTTP server. This idea is also reflected in the license of Flask. 22. However the majority of servers are using either threads. For many web applications the complexity of the code is less an issue than the scaling for the number of users or data entries expected. You don’t have to contribute any changes back if you decide to modify the framework. The downside of forking is of course that Flask extensions will most likely break be- cause the new framework has a different import name. For every project there is the point where the underlying framework gets in the way (due to assumptions the original developers had). Scaling well means for example that if you double the amount of servers you get about twice the performance. There is only one limiting factor regarding scaling in Flask which are the context local proxies. Scaling bad means that if you add a new server the application won’t perform any better or would not even support a second server. If your server uses some kind of concurrency that is not based on threads or greenlets. Werkzeug includes several middlewares. This is not unique to Flask. They depend on context which in Flask is defined as being either a thread. Many people use patched and modified versions of their framework to counter shortcomings. Flask is just the paste that glues those together. process or greenlet. greenlets or separate pro- cesses to achieve concurrency which are all methods well supported by the underlying Werkzeug library. Be- cause of that. If none of the above options work. The Application Dispatching chapter shows in detail how to apply middleware. 22. These libraries do the majority of the work. 184 . Furthermore integrating up- stream changes can be a complex process. The majority of code of Flask is within Werkzeug and Jinja2.4 Wrap with middleware. the framework would be a very complex system to begin with which causes a steep learning curve and a lot of user frustration. Flask will no longer be able to support these global proxies. This is natural because if this would not be the case. Flask by itself is only limited in terms of scaling by your application code.5 Fork.

22. don’t hesitate to contact the developers on the mailinglist or IRC channel. caused by Flask. If you find an obstacle in your way. The Flask developers keep the framework accessible to users with codebases big and small. The best way for the Flask and Flask extension developers to improve the tools for larger applications is getting feedback from users. 185 .7 Discuss with the community.

186 .

Part II
API REFERENCE

If you are looking for information on a specific function, class or method, this part of
the documentation is for you.

187

188

CHAPTER

TWENTYTHREE

API

This part of the documentation covers all the interfaces of Flask. For parts where Flask
depends on external libraries, we document the most important right here and provide
links to the canonical documentation.

23.1 Application Object

class flask.Flask(import_name, static_path=None, static_url_path=None,
static_folder=’static’, template_folder=’templates’, in-
stance_path=None, instance_relative_config=False,
root_path=None)
The flask object implements a WSGI application and acts as the central object.
It is passed the name of the module or package of the application. Once it is
created it will act as a central registry for the view functions, the URL rules,
template configuration and much more.
The name of the package is used to resolve resources from inside the package
or the folder the module is contained in depending on if the package parameter
resolves to an actual python package (a folder with an __init__.py file inside)
or a standard module (just a .py file).
For more information about resource loading, see open_resource().
Usually you create a Flask instance in your main module or in the __init__.py
file of your package like this:

from flask import Flask
app = Flask(__name__)

About the First Parameter
The idea of the first parameter is to give Flask an idea of what belongs to your
application. This name is used to find resources on the filesystem, can be used
by extensions to improve debugging information and a lot more.
So it’s important what you provide there. If you are using a single module,
__name__ is always the correct value. If you however are using a package, it’s

189

usually recommended to hardcode the name of your package there.
For example if your application is defined in yourapplication/app.py you
should create it with one of the two versions below:

app = Flask('yourapplication')
app = Flask(__name__.split('.')[0])

Why is that? The application will work even with __name__, thanks to how re-
sources are looked up. However it will make debugging more painful. Certain
extensions can make assumptions based on the import name of your application.
For example the Flask-SQLAlchemy extension will look for the code in your ap-
plication that triggered an SQL query in debug mode. If the import name is not
properly set up, that debugging information is lost. (For example it would only
pick up SQL queries in yourapplication.app and not yourapplication.views.frontend)

New in version 0.7: The static_url_path, static_folder, and template_folder parame-
ters were added.
New in version 0.8: The instance_path and instance_relative_config parameters
were added.
New in version 0.11: The root_path parameter was added.
Parameters
• import_name – the name of the application package
• static_url_path – can be used to specify a different path
for the static files on the web. Defaults to the name of the
static_folder folder.
• static_folder – the folder with static files that should be
served at static_url_path. Defaults to the 'static' folder in the
root path of the application.
• template_folder – the folder that contains the templates that
should be used by the application. Defaults to 'templates'
folder in the root path of the application.
• instance_path – An alternative instance path for the applica-
tion. By default the folder 'instance' next to the package or
module is assumed to be the instance path.
• instance_relative_config – if set to True relative filenames
for loading the config are assumed to be relative to the instance
path instead of the application root.
• root_path – Flask by default will automatically calculate the
path to the root of the application. In certain situations this
cannot be achieved (for instance if the package is a Python 3
namespace package) and needs to be manually defined.

190

add_template_filter(f, name=None)
Register a custom template filter. Works exactly like the template_filter()
decorator.
Parameters name – the optional name of the filter, otherwise the
function name will be used.
add_template_global(f, name=None)
Register a custom template global function. Works exactly like the
template_global() decorator.
New in version 0.10.
Parameters name – the optional name of the global function, other-
wise the function name will be used.
add_template_test(f, name=None)
Register a custom template test. Works exactly like the template_test()
decorator.
New in version 0.10.
Parameters name – the optional name of the test, otherwise the
function name will be used.
add_url_rule(rule, endpoint=None, view_func=None, **options)
Connects a URL rule. Works exactly like the route() decorator. If a
view_func is provided it will be registered with the endpoint.
Basically this example:

@app.route('/')
def index():
pass

Is equivalent to the following:

def index():
pass
app.add_url_rule('/', 'index', index)

If the view_func is not provided you will need to connect the endpoint to a
view function like so:

app.view_functions['index'] = index

Internally route() invokes add_url_rule() so if you want to customize the
behavior via subclassing you only need to change this method.
For more information refer to URL Route Registrations.
Changed in version 0.2: view_func parameter added.
Changed in version 0.6: OPTIONS is added automatically as method.
Parameters

191

• rule – the URL rule as string
• endpoint – the endpoint for the registered URL rule. Flask
itself assumes the name of the view function as endpoint
• view_func – the function to call when serving a request to the
provided endpoint
• options – the options to be forwarded to the underlying Rule
object. A change to Werkzeug is handling of method options.
methods is a list of methods this rule should be limited to
(GET, POST etc.). By default a rule just listens for GET (and
implicitly HEAD). Starting with Flask 0.6, OPTIONS is implicitly
added and handled by the standard request handling.
after_request(f )
Register a function to be run after each request.
Your function must take one parameter, an instance of response_class and
return a new response object or the same (see process_response()).
As of Flask 0.7 this function might not be executed at the end of the request
in case an unhandled exception occurred.
after_request_funcs = None
A dictionary with lists of functions that should be called after each request.
The key of the dictionary is the name of the blueprint this function is active
for, None for all requests. This can for example be used to close database
connections. To register a function here, use the after_request() decorator.
app_context()
Binds the application only. For as long as the application is bound to the
current context the flask.current_app points to that application. An appli-
cation context is automatically created when a request context is pushed if
necessary.
Example usage:

with app.app_context():
...

New in version 0.9.
app_ctx_globals_class
The class that is used for the g instance.
Example use cases for a custom class:
1.Store arbitrary attributes on flask.g.
2.Add a property for lazy per-request database connectors.
3.Return None instead of AttributeError on unexpected attributes.
4.Raise exception if an unexpected attr is set, a “controlled” flask.g.

192

It will basically calculate the path to a folder named instance next to your main file or the package. Blueprints can be at- tached multiple times so this dictionary does not tell you how often they got attached.g object is now applica- tion context scoped.10 to app_ctx_globals_class because the flask. In Flask 0. This can for example be used to open database connections or getting hold of the currently logged in user. New in version 0.10. use the before_request() decorator.7. New in version 0. blueprints = None all the attached blueprints in a dictionary by name. before_request(f ) Registers a function to run before each request. Commands registered 193 . New in version 0.9 this property was called request_globals_class but it was changed in 0. The function will be called without any arguments and its return value is ignored. New in version 0. To register a function here. The function will be called without any arguments. before_request_funcs = None A dictionary with lists of functions that should be called at the beginning of the request. cli = None The click command line context for this application. New in version 0.8. use the before_first_request() decorator.8. before_first_request_funcs = None A lists of functions that should be called at the beginning of the first request to this instance. None for all requests. To register a function here. The key of the dictionary is the name of the blueprint this function is active for.8. it’s handled as if it was the return value from the view and further request handling is stopped. alias of _AppCtxGlobals auto_find_instance_path() Tries to locate the instance path if it was not provided to the constructor of the application class. If the function returns a non-None value. before_first_request(f ) Registers a function to be run before the first request to this instance of the application.

Group object. 194 . Changed in version 0.6.Default values for certain config options. config = None The configuration dictionary as Config. This behaves exactly like a regular dictionary but supports additional methods to load a config from files.11: Environment. This is an instance of a click. New in version 0.7.11. Defaults to Config. create_jinja_environment() Creates the Jinja2 environment based on jinja_options and select_jinja_autoescape().5. alias of Config context_processor(f ) Registers a template context processor function. create_global_jinja_loader() Creates the loader for the Jinja2 environment. Can be used to override just the loader and keeping the rest unchanged. create_url_adapter(request) Creates a URL adapter for the given request.7 this also adds the Jinja2 globals and filters after initialization. The global loader dispatches between the loaders of the application and the individual blueprints. The URL adapter is created at a point where the request context is not yet set up so the request is passed explicitly.9: This can now also be called without a request object when the URL adapter is created for the application context. config_class The class that is used for the config attribute of this app. New in version 0. The default commands are provided by Flask itself and can be overridden. New in version 0. 2. Changed in version 0. Example use cases for a custom class: 1. It’s discouraged to override this function.Access to config values through attributes in addition to keys. Since 0. New in version 0.auto_reload set in accordance with TEMPLATES_AUTO_RELOAD configuration option. Instead one should override the jinja_loader() function instead. here show up in the flask command once the application has been dis- covered. Override this function to customize the behavior.

9. Example: @app. do_teardown_appcontext(exc=<object object>) Called when an application context is popped.endpoint('example. In debug mode the debugger will kick in when an unhandled exception occurs and the integrated server will automatically reload the application if changes in the code are detected. do_teardown_request(exc=<object object>) Called after the actual request dispatching and will call every as teardown_request() decorated function. Each key points to another dictionary where the key is the status code of the http exception. This is not actually called by the Flask object itself but is always triggered when the request context is popped. Matches the URL and returns the return value of the view or error handler. endpoint(endpoint) A decorator to register a function as an endpoint.endpoint') def example(): return "example" Parameters endpoint – the name of the endpoint error_handler_spec = None A dictionary of all registered error handlers. ‘SE Default configuration parameters. Set this to True to enable debugging of the application. This does not have to be a response object. New in version 0. In order to convert the return value to a proper response object. this code was moved to the new full_dispatch_request(). Changed in version 0. This attribute can also be configured from the config with the DEBUG config- uration key.7: This no longer does the exception handling. dispatch_request() Does the request dispatching. That way we have a tighter control over certain resources under testing environments. Defaults to False.9: Added the exc argument. call make_response(). The key is None for error handlers active on the application. Previously this was al- ways using the current exception information. ‘USE_X_SENDFILE’: False. This works pretty much the same as do_teardown_request() but for the application context. Changed in version 0. default_config = ImmutableDict({‘JSON_AS_ASCII’: True. otherwise the key is the name of the blueprint.debug The debug flag. The special key None points to a list of tuples 195 .

Parameters code_or_exception – the code as integer for the han- dler.errorhandler(404) def page_not_found(error): return 'This page does not exist'.7: One can now additionally also register custom excep- tion types that do not necessarily have to be a subclass of the HTTPException class.extensions['extensionname'] = SomeObject() 196 . To register a error handler. For exam- ple this is where an extension could store database engines and similar things. where the first item is the class for the instance check and the second the error handler function. 404 app. The first None refers to the active blueprint. For backwards compatibility extensions should register themselves like this: if not hasattr(app. New in version 0. errorhandler(code_or_exception) A decorator that is used to register a function given an error code. 404 You can also register handlers for arbitrary exceptions: @app.extensions = {} app.7: Use register_error_handler() instead of modifying error_handler_spec directly. 'extensions'): app.errorhandler(DatabaseError) def special_exception_handler(error): return 'Database connection failed'. use the errorhandler() decorator. 500 You can also register a function as error handler without using the errorhandler() decorator. The following example is equivalent to the one above: def page_not_found(error): return 'This page does not exist'.error_handler_spec[None][404] = page_not_found Setting error handlers via assignments to error_handler_spec however is discouraged as it requires fiddling with nested dictionaries and the special case for arbitrary exception types. for application wide error handlers. New in version 0. If the error handler should be application wide None shall be used. or an arbitrary exception extensions = None a place where extensions can store application specific state. Example: @app.

By default.js'): return 60 return flask. name): if name.3. New in version 0.7. If no such handler exists. name) New in version 0. endpoint. For example. By default this will invoke the registered error handlers and fall back to returning the exception as response. this method is called. New in version 0.endswith('.9. New in version 0. the key would be 'foo'.get_send_file_max_age(self. this function returns SEND_FILE_MAX_AGE_DEFAULT from the con- figuration of current_app. handle_url_build_error(error. The key must match the name of the extension module.js files to 60 seconds: class MyFlask(flask. a default 500 internal server error message is dis- played. values) Handle BuildError on url_for(). This allows subclasses to change the behavior when sending files based on the filename. oth- erwise it is logged and the handler for a 500 internal server error is used. 197 . New in version 0. handle_http_exception(e) Handles an HTTP exception. get_send_file_max_age(filename) Provides default cache_timeout for the send_file() functions. otherwise.Flask. In debug mode the exception will be re-raised immediately.lower().8. full_dispatch_request() Dispatches the request and on top of that performs request pre and post- processing as well as HTTP exception catching and error handling. Static file functions such as send_from_directory() use this function. For example in case of a “Flask-Foo” extension in flask_foo.7. that time- out is used. New in version 0.Flask): def get_send_file_max_age(self. If a cache_timeout is given in send_file(). to set the cache timeout for . and send_file() calls this function on current_app when the given cache_timeout is None. handle_exception(e) Default exception handling that kicks in when an exception occurs that is not caught. got_first_request This attribute is set to True if the application started handling the first re- quest.3.

A special case are HTTPExceptions which are forwarded by this function to the handle_http_exception() method.5. New in version 0. inject_url_defaults(endpoint.ext. instance_path = None Holds the path to the instance folder.with_’]}) Options that are passed directly to the Jinja2 environment.11. New in version 0.7: This method is deprecated with 0. jinja_options = ImmutableDict({‘extensions’: [’jinja2. init_jinja_globals() Deprecated. This function will either return a response value or reraise the exception with the same traceback. New in version 0.5. New in version 0.7. New in version 0. New in version 0. Override create_jinja_environment() instead. values) Injects the URL defaults for the given endpoint directly into the values dic- tionary passed.7.8. Changed in version 0.7.5. alias of Environment jinja_loader The Jinja loader for this package bound object. iter_blueprints() Iterates over all blueprints by the order they were registered. This is used internally and automatically called on URL building.11.ext. Used to initialize the Jinja2 globals. handle_user_exception(e) This method is called whenever an exception occurs that should be handled. has_static_folder This is True if the package bound object’s container has a folder for static files. 198 . jinja_env The Jinja2 environment used to load templates.autoescape’. jinja_environment The class that is used for the Jinja environment. New in version 0. ‘jinja2. New in version 0.

10. New in version 0. logger A logging.8.4. New in version 0. New in version 0.logger. alias of JSONDecoder json_encoder The JSON encoder class to use.warning('A warning occurred (%d apples)'. This is called by handle_exception() if debugging is disabled and right before the handler is called.logger.Logger object for this application. Instead of overriding this method we recommend replacing the session_interface. make_null_session() Creates a new instance of a missing session. The default configuration is to log to stderr if the application is in debug mode. The default implementation logs the exception as error on the logger.10. The in- stance_relative parameter is passed in from the constructor of Flask (there named instance_relative_config) and indicates if the config should be relative to the instance path or the root path of the application. This can be changed through subclassing to change the default behavior of OPTIONS re- sponses.8. logger_name The name of the logger to use. New in version 0.error('An error occurred') New in version 0. 199 . Defaults to JSONDecoder.debug('A value for debugging') app.3. alias of JSONEncoder log_exception(exc_info) Logs an exception.logger. By default the logger name is the package name passed to the constructor. Here some examples: app. make_default_options_response() This method is called to create the default OPTIONS response. 42) app. This logger can be used to (surprise) log messages. New in version 0. Defaults to JSONEncoder. New in version 0.7. make_config(instance_relative=False) Used to create the config attribute by the Flask constructor.json_decoder The JSON decoder class to use.

name The name of the application. New in version 0. mode=’rb’) Opens a resource from the application’s resource folder. This runs all the registered shell context processors. default is ‘rb’. To access resources within subfolders use forward slashes as separator. To see how this works. consider the following folder structure: 200 . New in version 0. • mode – resource file opening mode. open_instance_resource(resource. New in version 0. make_shell_context() Returns the shell context for an interactive shell for this application. Instance resources can also be opened for writing. The following types are allowed for rv: response_class the object is returned unchanged str a response object is created with the string as body unicode a response object is created with the string encoded to utf-8 as body a WSGI function the function is called as WSGI application and buffered as response object tuple A tuple in the form (response. open_resource(resource.9: Previously a tuple was interpreted as the arguments for the response object.11.8.7. headers) where response is any of the types defined here. Parameters rv – the return value from the view function Changed in version 0. Parameters • resource – the name of the resource. status. make_response(rv) Converts the return value from a view function to a real response object that is an instance of response_class. headers) or (response. This is usually the import name with the dif- ference that it’s guessed from the run file if the import name is main. This name is used as a display name when Flask needs the name of the applica- tion. It can be set and overridden to change the value. mode=’rb’) Opens a resource from the application’s instance folder (instance_path). Otherwise works like open_resource(). status is a string or an inte- ger and headers is a list or a dictionary with header values.

css /templates /layout.sql') as f: contents = f.py /schema. This also triggers the url_value_preprocessor() functions before the actual before_request() functions are called. The default is 31 days which makes a permanent session survive for roughly one month. If any of these functions returns a value.open_resource('schema.html If you want to open the schema. New in version 0. Parameters request – an instance of request_class. To access resources within subfolders use forward slashes as separator.7.sql /static /style. passing no arguments. permanent_session_lifetime A timedelta which is used to set the expiration date of a permanent session. This requires that the secret_key is set. This attribute can also be configured from the config with the PERMANENT_SESSION_LIFETIME configuration key. /myapplication. open_session(request) Creates or opens a new session. it’s handled as if it was the return value from the view and further request handling is stopped. Default implementation stores all session data in a signed cookie. Defaults to timedelta(days=31) preprocess_request() Called before the actual request dispatching and will call each before_request() decorated function. preserve_context_on_exception Returns the value of the PRESERVE_CONTEXT_ON_EXCEPTION configuration value in case it’s set. default is ‘rb’.read() do_something_with(contents) Parameters • resource – the name of the resource.sql file you would do the following: with app. Instead of overriding this method we recommend replacing the session_interface. otherwise a sensible default is returned.html /index. • mode – resource file opening mode. 201 .

This must be used in combination with the with statement because the request is only bound to the current context for the duration of the with block. register_error_handler(code_or_exception.request_context(environ): do_something_with(request) The object returned can also be used without the with statement which is useful for working in the shell.request_context(environ) ctx. **options) Registers a blueprint on the application. See Request for more information. otherwise a sensible default is returned. Returns a new response object or the same. request_class The class that is used for request objects. alias of Request request_context(environ) Creates a RequestContext from the given environment and binds it to the current context. New in version 0. By default this will call all the after_request() decorated functions. The example above is doing exactly the same as this code: ctx = app. Example usage: with app.7. Parameters response – a response_class object.push() try: do_something_with(request) 202 .5: As of Flask 0. register_blueprint(blueprint. New in version 0. Changed in version 0. f ) Alternative error attach function to the errorhandler() decorator that is more straightforward to use for non decorator usage. has to be an instance of response_class. propagate_exceptions Returns the value of the PROPAGATE_EXCEPTIONS configuration value in case it’s set.5 the functions registered for after request execution are called in reverse order of registration.7.7. process_response(response) Can be overridden in order to modify the response object before it’s sent to the WSGI server. New in version 0.

This will keep the debugger’s traceback screen active. Parameters • rule – the URL rule as string • endpoint – the endpoint for the registered URL rule. It is not intended to meet security and performance requirements for a production server. see Deploy- ment Options for WSGI server recommendations. you can pass use_evalex=False as pa- rameter. OPTIONS is implicitly added and handled by the standard request handling. Flask itself assumes the name of the view function as endpoint • options – the options to be forwarded to the underlying Rule object.6. methods is a list of methods this rule should be limited to (GET. **options) Runs the application on a local development server.pop() Changed in version 0. Starting with Flask 0. but disable the code ex- ecution on the interactive debugger. Parameters environ – a WSGI environment response_class The class that is used for response objects. By default a rule just listens for GET (and implicitly HEAD). If you want to run the application in debug mode. **options) A decorator that is used to register a view function for a given URL rule. If the debug flag is set the server will automatically reload for code changes and show a debugger in case an exception happened. This does the same thing as add_url_rule() but is intended for decorator usage: @app. Do not use run() in a production setting. port=None. but disable code execution. alias of Response route(rule.).3: Added support for non-with statement usage and with statement is now passed the ctx object. debug=None. See Response for more informa- tion. Instead.route('/') def index(): return 'Hello World' For more information refer to URL Route Registrations. finally: ctx. 203 . POST etc. run(host=None. A change to Werkzeug is handling of method options.

Set this to a complex random value when you want to use the secure cookie for instance. save_session(session. Defaults to 5000 or the port defined in the SERVER_NAME config variable if present. select_jinja_autoescape(filename) Returns True if autoescaping should be active for the given template name. cryptographic components can use this to sign cookies and other things. Instead you should be using the flask command line script’s run support. Defaults to '127. Parameters • session – the session to be saved (a SecureCookie object) • response – an instance of response_class secret_key If a secret key is set. For the default implementation. Keep in Mind Flask will suppress any server error with a generic error page unless it is in debug mode. As such to enable just the interactive debugger with- out the code reloading.0' to have the server available externally as well. response) Saves the session if it needs updates. Changed in version 0. Setting use_debugger to True without being in debug mode won’t catch any exceptions because there won’t be any to catch. • options – the options to be forwarded to the underlying Werkzeug server.10: The default port is now picked from the SERVER_NAME variable. returns True. 204 .5. Instead of overriding this method we recommend replac- ing the session_interface. It is not recommended to use this function for development with automatic reloading as this is badly supported. enable or disable debug mode.0. you have to invoke run() with debug=True and use_reloader=False. See werkzeug. 1'. New in version 0. • port – the port of the webserver.run_simple() for more information. This attribute can also be configured from the config with the SECRET_KEY configuration key. Set this to '0. check open_session(). • debug – if given.0.serving.0. Parameters • host – the hostname to listen on. See debug. If no template name is given.0. Defaults to None.

The default is 12 hours. static_folder The absolute path to the configured static folder.5. shell_context_processors = None A list of shell context processor functions that should be run when a shell context is created. New in version 0. shell_context_processor(f ) Registers a shell context processor function. should_ignore_error(error) This is called to figure out if an error should be ignored or not as far as the teardown system is concerned. New in version 0.11.10. These functions are typically also called when the request context is popped.send_file_max_age_default A timedelta which is used as default cache_timeout for the send_file() functions.SecureCookieSessionInterface object> the session interface to use. If this function returns True then the teardown handlers will not be passed the error. session_cookie_name The secure cookie uses this for the name of the session cookie. New in version 0.11. Example: ctx = app. This attribute can also be configured from the config with the SESSION_COOKIE_NAME configuration key.sessions. This configuration vari- able can also be set with an integer value used as seconds. teardown_appcontext(f ) Registers a function to be called when the application context ends.push() 205 . Defaults to timedelta(hours=12) send_static_file(filename) Function used internally to send static files from the static folder to the browser. By default an instance of SecureCookieSessionInterface is used here. New in version 0.8. This attribute can also be configured from the config with the SEND_FILE_MAX_AGE_DEFAULT configuration key.app_context() ctx. New in version 0. Defaults to 'session' session_interface = <flask.

the teardown functions are called just before the app context moves from the stack of active con- texts. Debug Note In debug mode Flask will not tear down a request on an exception immediately..9. ctx. These functions are executed when the request context is popped. Instead it will keep it alive so that the interactive de- 206 . This becomes relevant if you are using such constructs in tests. teardown_request(f ) Register a function to be run at the end of each request.test_request_context() ctx. Since the application context is also torn down if the request ends this is the place to store code that disconnects from databases.pop() is executed in the above example.9. . Example: ctx = app.pop() When ctx. New in version 0.pop() is executed in the above example. the teardown functions are called just before the request context moves from the stack of active con- texts. This becomes relevant if you are using such constructs in tests. Generally teardown functions must take every necessary step to avoid that they will fail. Since a request context typically also manages an application context it would also be called when you pop a request context.. When a teardown function was called because of a exception it will be passed an error object. The return values of teardown functions are ignored. teardown_appcontext_funcs = None A list of functions that are called when the application context is destroyed.pop() When ctx.push() . When a teardown function was called because of an exception it will be passed an error object. New in version 0.. The return values of teardown functions are ignored.. regardless of whether there was an exception or not. If they do execute code that might fail they will have to sur- round the execution of these code by try/except statements and log occur- ring errors. ctx. even if not an actual request was performed.

template_global(name=None) A decorator that is used to register a custom template global function. otherwise the function name will be used. You can specify a name for the filter. Example: @app. These func- tions are not allowed to modify the request. To register a function here. Each returns a dictionary that the template context is updated with. You can specify a name for the test. If an exception occurred while processing the request.7. otherwise the function name will be used. New in version 0. use the teardown_request() decorator. template_context_processors = None A dictionary with list of functions that are called without argument to pop- ulate the template context. To register a function here. otherwise the function name will be used. Example: @app. and their return values are ig- nored. use the context_processor() decorator. it gets passed to each teardown_request function. Example: 207 . The key of the dictionary is the name of the blueprint this function is active for.template_global() def double(n): return 2 * n New in version 0. You can specify a name for the global function. Parameters name – the optional name of the global function. None for all requests. template_test(name=None) A decorator that is used to register custom template test.10. The key of the dictionary is the name of the blueprint this function is active for. teardown_request_funcs = None A dictionary with lists of functions that are called after each request.template_filter() def reverse(s): return s[::-1] Parameters name – the optional name of the filter. otherwise the function name will be used. bugger can still access it. even if an exception has occurred. None for all requests. template_filter(name=None) A decorator that is used to register custom template filter. This behavior can be controlled by the PRESERVE_CONTEXT_ON_EXCEPTION configuration variable. other- wise the function name will be used.

test_client(authentication='Basic .. For example: app. *args. See the testing attribute.10.test_client() as c: rv = c.test_client_class = CustomClient client = app. **kwargs): self. **kwargs) app. you must set app.testing = True client = app.get('/?vodka=42') assert request._authentication = kwargs. Changed in version 0.template_test() def is_prime(n): if n == 2: return True for i in range(2.. int(math.sqrt(n))) + 1): if n % i == 0: return False return True New in version 0.testing import FlaskClient class CustomClient(FlaskClient): def __init__(self. Parameters name – the optional name of the test.test_client() The test client can be used in a with block to defer the closing down of the context until the end of the with block. 208 . you may pass optional keyword arguments that will then be passed to the application’s test_client_class constructor.args['vodka'] == '42' Additionally.. Note that if you are testing for assertions or exceptions in your application code. test_client(use_cookies=True. For example: from flask.testing = True in order for the exceptions to prop- agate to the test client. This is useful if you want to access the context locals for testing: with app. @app.ceil(math. Otherwise. For information about unit testing head over to Testing Flask Applications.self). **kwargs) Creates a test client for this application.pop("authentication") super(CustomClient.4: added support for with block usage for the client.__init__( *args. the exception will be handled by the application (not visible to the test client) and the only indication of an As- sertionError or other exception will be a 500 status code response to the test client.') See FlaskClient for more information. otherwise the function name will be used.

This is helpful for debug- ging implicitly raised HTTP exceptions. Defaults to False. test_client_class = None the test client that is used with when test_client is used. For example this might activate unittest helpers that have an additional runtime cost which should not be enabled by default. **kwargs) Creates a WSGI environment from the given values (see werkzeug. session. New in version 0. Set this to True to enable the test mode of Flask extensions (and in the future probably also Flask itself). config and g into the template context as well as every- thing template context processors want to inject.6. This in- jects request. test_request_context(*args. testing The testing flag. this function accepts the same argu- ments). This is called for all HTTP exceptions raised by a view function. EnvironBuilder for more information. trap_http_exception(e) Checks if an HTTP exception should be trapped or not.11: Added **kwargs to support passing additional key- word arguments to the constructor of test_client_class. If it returns True for any exception the error handler for this exception is not called and it shows up as regular exception in the traceback. By default this will return False for all exceptions except for a bad request key er- ror if TRAP_BAD_REQUEST_ERRORS is set to True. If this is enabled and PROPAGATE_EXCEPTIONS is not changed from the default it’s implicitly enabled. New in version 0.7: The use_cookies parameter was added as well as the ability to override the client to be used by setting the test_client_class attribute. url_build_error_handlers = None A list of functions that are called when url_for() raises a BuildError.test. This attribute can also be configured from the config with the TESTING con- figuration key. update_template_context(context) Update the template context with some commonly used variables.7. Each 209 . Changed in version 0.8. Note that the as of Flask 0. New in version 0. the original values in the context will not be overridden if a context processor decides to return a value with the same key. Parameters context – the context as a dictionary that is updated in place to add extra variables. It also returns True if TRAP_HTTP_EXCEPTIONS is set to True.

7. Each of these functions has the chance to modify the dictionary of URL values before they are used as the key- word arguments of the view function. Whenever a URL is built these functions are called to modify 210 .routing.join(super(ListConverter. value): return value.url_map.converters['list'] = ListConverter url_rule_class The rule object to use for URL rules created.split('. url_defaults(f ) Callback function for URL defaults for all view functions of the applica- tion. This is used by add_url_rule(). endpoint and values.7. The key None here is used for application wide callbacks. If a function returns None or raises a BuildError the next function is tried. otherwise the key is the name of the blueprint.to_url(value) for value in values) app = Flask(__name__) app. New in version 0.') def to_url(self.'. values): return '. It’s called with the endpoint and values and should update the values passed in place. You can use this to change the routing converters after the class was created but before any routes are connected. function registered here is called with error.routing import BaseConverter class ListConverter(BaseConverter): def to_python(self. url_value_preprocessors = None A dictionary with lists of functions that can be used as URL value processor functions. New in version 0.Rule. Example: from werkzeug. It’s called before the view functions are called and can modify the url values provided. url_map = None The Map for this instance.9. url_default_functions = None A dictionary with lists of functions that can be used as URL value prepro- cessors. alias of Rule url_value_preprocessor(f ) Registers a function as URL value preprocessor for all view functions of the application. Defaults to werkzeug. New in version 0. self). For each function registered this one should also provide a url_defaults() function that adds the parameters automatically again that were removed that way.

otherwise the key is the name of the blueprint. the dictionary of values in place. view_functions = None A dictionary of all view functions registered. New in version 0.2. use_x_sendfile Enable this if you want to use the X-Sendfile feature. The key None here is used for application wide callbacks. independent on if an error occurred or not.7: The behavior of the before and after request call- backs was changed under error conditions and a new callback was added that will always execute at the end of the request. Changed in version 0. This only affects files sent with the send_file() method.wsgi_app = MyMiddleware(app. Defaults to False. wsgi_app(environ. So in- stead of doing this: app = MyMiddleware(app) It’s a better idea to do this instead: app. a list of headers and an optional exception context to start the re- sponse 211 . start_response) The actual WSGI application. use the route() decorator. See Callbacks and Errors. This is not implemented in __call__ so that middlewares can be applied without losing a reference to the class. To register a view function. This attribute can also be configured from the config with the USE_X_SENDFILE configuration key.7.wsgi_app) Then you still have the original application object around and can continue to call methods on it. Each of these functions has the chance to modify the dictionary New in version 0. The keys will be function names which are also used to generate URLs and the values are the function objects themselves. Parameters • environ – a WSGI environment • start_response – a callable accepting a status code. Keep in mind that the server has to support this.

See Modular Applications with Blueprints for more infor- mation. template_folder=None. import_name. url_defaults=None. add_app_template_test(f. static_url_path=None. A blueprint is an object that records functions that will be called with the BlueprintSetupState later to register functions or other things on the main application. Parameters name – the optional name of the global.after_request() but for a blueprint. **options) Like Flask. The endpoint for the url_for() function is prefixed with the name of the blueprint. subdomain=None. name=None) Register a custom template test. available application wide. available application wide. name=None) Register a custom template global. Parameters name – the optional name of the filter. after_request(f ) Like Flask. add_app_template_global(f. Such a function is executed after each request. view_func=None. Parameters name – the optional name of the test.23.7. Works exactly like the app_template_test() decorator.Blueprint(name. name=None) Register a custom template filter.after_request() but for a blueprint. Like Flask.2 Blueprint Objects class flask. add_template_filter() but for a blueprint. add_app_template_filter(f. Works exactly like the app_template_filter() decorator. even if outside of the blueprint. Like Flask. New in version 0. url_prefix=None. otherwise the function name will be used. root_path=None) Represents a blueprint. This function is only exe- cuted after each request that is handled by a function of that blueprint. New in version 0. add_url_rule(rule. Works exactly like the app_template_global() decorator. after_app_request(f ) Like Flask. New in version 0.add_url_rule() but for a blueprint.10. static_folder=None. otherwise the function name will be used. Like Flask. endpoint=None. available application wide.add_template_test() but for a blueprint.add_template_global() but for a blueprint. otherwise the function name will be used. 212 .10.

This function is only executed for requests handled by a blueprint. even if outside of a blueprint. before_request(f ) Like Flask. New in version 0.10. available application wide. context_processor(f ) Like Flask. template_global() but for a blueprint. Like Flask. app_template_global(name=None) Register a custom template global. before_app_request(f ) Like Flask.before_request() but for a blueprint. New in version 0.before_request(). template_filter() but for a blueprint. template_test() but for a blueprint. otherwise the function name will be used. Such a function is executed before the first request to the application.10.before_first_request(). Such a function is executed before each re- quest. before_app_first_request(f ) Like Flask. Parameters name – the optional name of the test. app_errorhandler(code) Like Flask. otherwise the function name will be used. app_template_test(name=None) Register a custom template test. This function is only exe- cuted before each request that is handled by a function of that blueprint. even if outside of the blueprint. app_url_defaults(f ) Same as url_defaults() but application wide. Like Flask. available application wide.context_processor() but for a blueprint. Parameters name – the optional name of the filter. Such a function is ex- ecuted each request. even if outside of the blueprint.errorhandler() but for a blueprint.app_context_processor(f ) Like Flask. app_template_filter(name=None) Register a custom template filter. app_url_value_preprocessor(f ) Same as url_value_preprocessor() but application wide. otherwise the function name will be used.context_processor() but for a blueprint. available application wide. 213 . Like Flask. This handler is used for all requests. Parameters name – the optional name of the global.

otherwise it’s an application independent endpoint. to set the cache timeout for . jinja_loader The Jinja loader for this package bound object. Another special case is the 500 internal server error which is always looked up from the application.9. this method is called. To see how this 214 .Flask. that time- out is used. make_setup_state(app. options. endpoint(endpoint) Like Flask. Otherwise works as the errorhandler() decorator of the Flask object.lower(). If the endpoint is prefixed with a . name): if name. get_send_file_max_age(filename) Provides default cache_timeout for the send_file() functions. Static file functions such as send_from_directory() use this function. This allows subclasses to change the behavior when sending files based on the filename.js files to 60 seconds: class MyFlask(flask. errorhandler(code_or_exception) Registers an error handler that becomes active for this blueprint only. name) New in version 0.Flask): def get_send_file_max_age(self. this function returns SEND_FILE_MAX_AGE_DEFAULT from the con- figuration of current_app. and send_file() calls this function on current_app when the given cache_timeout is None. For example. New in version 0. it will be registered to the current blueprint.5.5. mode=’rb’) Opens a resource from the application’s resource folder.endswith('.endpoint() but for a blueprint. If a cache_timeout is given in send_file(). this has to be done explicitly by the user of this method. This does not prefix the end- point with the blueprint name.js'): return 60 return flask. Please be aware that routing does not happen local to a blueprint so an error han- dler for 404 usually is not handled by a blueprint unless it is caused inside a view function. By default. open_resource(resource. has_static_folder This is True if the package bound object’s container has a folder for static files. otherwise. New in version 0. Subclasses can override this to return a sub- class of the setup state.get_send_file_max_age(self. first_registration=False) Creates an instance of BlueprintSetupState() object that is later passed to the register callback functions.

the function passed is not called. f ) Non-decorator version of the errorhandler() error attach function. akin to the register_error_handler() application-wide function of the Flask ob- ject but for error handlers limited to this blueprint.read() do_something_with(contents) Parameters • resource – the name of the resource. • mode – resource file opening mode.sql /static /style.html If you want to open the schema. This function is called with the state as argument as returned by the make_setup_state() method. If the blueprint is registered a second time on the application. **options) Like Flask.css /templates /layout.11.register_blueprint() to register a blueprint on the ap- plication. first_registration=False) Called by Flask. register(app. record_once(func) Works like record() but wraps the function in another function that will ensure the function is only called once.html /index. To access resources within subfolders use forward slashes as separator. consider the following folder structure: /myapplication. send_static_file(filename) Function used internally to send static files from the static folder to the browser.sql') as f: contents = f. default is ‘rb’. Key- word arguments from register_blueprint() are directly forwarded to this method in the options dictionary. works. record(func) Registers a function that is called when the blueprint is registered on the application.route() but for a blueprint. register_error_handler(code_or_exception. 215 .py /schema. route(rule. New in version 0. options. This can be overridden to customize the register behavior.open_resource('schema.sql file you would do the following: with app. The endpoint for the url_for() function is prefixed with the name of the blueprint.

Request(environ. static_folder The absolute path to the configured static folder.3 Incoming Request Data class flask. even when no actual request was performed. teardown_request(f ) Like Flask. This function is only executed when tearing down requests handled by a function of that blueprint. values A CombinedMultiDict with the contents of both form and args. New in version 0. url_defaults(f ) Callback function for URL defaults for this blueprint.5.teardown_request() but for a blueprint. If you want to replace the request object used you can subclass this and set request_class to your subclass. (The part in the URL after the question mark). but instead in the files attribute. teardown_app_request(f ) Like Flask. It is what ends up as request. Such a function is exe- cuted when tearing down each request. even if outside of the blueprint. stream If the incoming form data was not encoded with a known mimetype the 216 . form A MultiDict with the parsed form data from POST or PUT requests. Please keep in mind that file uploads will not end up here. cookies A dict with the contents of all cookies transmitted with the request. It’s called before the view functions are called and can modify the url values provided. populate_request=True. Remembers the matched endpoint and view arguments. It’s called with the endpoint and values and should update the values passed in place. shallow=False) The request object used by default in Flask. args A MultiDict with the parsed contents of the query string. Teardown request functions are executed when the request con- text is popped. The request object is a Request subclass and provides all of the attributes Werkzeug defines plus a few Flask specific ones.teardown_request() but for a blueprint. 23. url_value_preprocessor(f ) Registers a function as URL value preprocessor for this blueprint.

example.html? x=y' url_root u'http://www.com/myapplication/π/page.example.html' url u'http://www.com/myapplication And a user requests the following URI: http://www.example. with the difference that it also has a save() function that can store the file on the filesystem.com/myapplication/π/page. It basically behaves like a standard file object you know from Python. data is stored unmodified in this stream for consumption. Each file is stored as FileStorage object. data Contains the incoming request data as string in case it came with a mime- type Flask does not handle.html' full_path u'/π/page.com/myapplication/' 217 . files A MultiDict with files uploaded as part of a POST or PUT request.com/myapplication/%CF%80/page.html?x=y In this case the values of the above mentioned attributes would be the fol- lowing: path u'/π/page. The stream only returns the data once. headers The incoming request headers as a dictionary like object.example.html?x=y' script_root u'/myapplication' base_url u'http://www.) path full_path script_root url base_url url_root Provides different ways to look at the current IRI. method The current request method (POST. environ The underlying WSGI environment. Most of the time it is a better idea to use data which will give you that data as a string. GET etc.example. Imagine your application is listening on the following application root: http://www.

use blueprints instead. By default a request is considered to include JSON data if the mimetype is applicationhttps://www.scribd.com/json or application/ *+json. silent=False. Otherwise this will be None. This in combination with view_args can be used to reconstruct the same or a modified URL. json If the mimetype is applicationhttps://www.scribd.com/json this will contain the parsed JSON data. Libraries that do that are prototype. The return value of this method is used by get_json() when an error occurred. max_content_length Read-only view of the MAX_CONTENT_LENGTH config key. on_json_loading_failed(e) Called if decoding of the JSON data failed. The default implementation just raises a BadRequest exception. Parameters • force – if set to True the mimetype is ignored. If an exception happened when matching. The get_json() method should be used instead. blueprint The name of the current blueprint endpoint The endpoint that matched the request.11. jQuery and Mochikit and probably some more. 218 . If parsing fails the on_json_loading_failed() method on the request object will be invoked. By default this function will return None if the mimetype is not applicationhttps://www.scribd.com/json but this can be overridden by the force parameter. get_json(force=False. This only works with libraries that support the X-Requested-With header and set it to XMLHttpRequest. cache=True) Parses the incoming JSON request data and returns it. is_json Indicates if this request is JSON or not. New in version 0. • cache – if set to True the parsed JSON data is remembered on the request. module The name of the current module if the request was dispatched to an actual module. This is deprecated functionality. • silent – if set to True this method will fail silently and return None. is_xhr True if the request was triggered via a JavaScript XMLHttpRequest. this will be None.

Response(response=None.8. status A string with a response status. 23. headers=None. status=None. Works like the response object from Werkzeug but is set to have an HTML mimetype by default. This just shows a quick overview of the most im- portant ones.10: Removed buggy previous behavior of generating a random JSON response. This is a proxy. See Notes On Proxies for more information. view_args = None A dict of view arguments that matched the request. Changed in version 0. 219 . If you want to replace the response object used you can subclass this and set response_class to your subclass.url_rule. url_rule = None The internal URL rule that matched the request. If an exception hap- pened when matching. This can be useful to in- spect which methods are allowed for the URL from a before/after handler (request. content_type=None. direct_passthrough=False) The response object that is used by default in Flask. The request object is an instance of a Request subclass and provides all of the attributes Werkzeug defines.4 Response Objects class flask.6. headers A Headers object representing the response headers. Flask parses incoming request data for you and gives you access to it through that global object. class flask. New in version 0. This is usually a NotFound exception or something similar. routing_exception = None If matching the URL failed. Internally Flask makes sure that you always get the correct data for the active thread if you are in a multithreaded environment.methods) etc. If you want that behavior back you can trivially add it by subclassing. this is the exception that will be raised / was raised as part of the request handling. Quite often you don’t have to create this object yourself because make_response() will take care of that for you. you can use the global request object. this will be None.request To access incoming request data. New in version 0. mime- type=None.

• expires – should be a datetime object or UNIX timestamp. See Notes On Proxies for more information. value=’‘.com" will set a cookie that is readable by the domain www. max_age=None. This is a proxy. but not modify it unless they know the secret key. so make sure to set that to something complex and unguessable. path=’/’. foo. This should not be used and will eventually get deprecated. mimetype The mimetype (content type without charset etc. expires=None. too.example. The parameters are the same as in the cookie Morsel object in the Python standard library but it accepts unicode data. Oth- erwise.) set_cookie(key. with the difference that it keeps track on modifications. A session basically makes it possible to remember information from one request to an- other. Parameters • key – the key (name) of the cookie to be set. To access the current session you can use the session object: class flask.5 Sessions If you have the Flask. • path – limits the cookie to a given path. data A descriptor that calls get_data() and set_data(). For exam- ple. • domain – if you want to set a cross-domain cookie. domain=". The way Flask does this is by using a signed cookie.com. secure=None. httponly=False) Sets a cookie. or None (default) if the cookie should last only as long as the client’s browser session. 23.session The session object works pretty much like an ordinary dict. The following attributes are interesting: 220 . per default it will span the whole domain.example. • max_age – should be a number of seconds.secret_key set you can use sessions in Flask applications. a cookie will only be readable by the domain that set it. do- main=None. So the user can look at the session contents. • value – the value of the cookie. status_code The response status as integer.com etc.example.

new True if the session is new.6 Session Interface New in version 0.SessionInterface The basic interface you have to implement in order to replace the default session interface which uses werkzeug’s securecookie implementation. Here an example: # this change is not picked up because a mutable object (here # a list) is changed.session_interface: app = Flask(__name__) app. 23. The default NullSession class that is created will complain that the secret key was not set.modified = True permanent If set to True the session lives for permanent_session_lifetime seconds.sessions. The session object returned by the open_session() method has to provide a dic- tionary like interface plus the properties and methods from the SessionMixin. To replace the session interface on an application all you have to do is to assign flask. The session interface provides a simple way to replace the session implementation that Flask is using. The only meth- ods you have to implement are open_session() and save_session().Flask. in that situation you have to explicitly set the attribute to True yourself.session_interface = MySessionInterface() 221 . SessionMixin): pass If open_session() returns None Flask will call into make_null_session() to create a session that acts as replacement if the session support cannot work because some requirement is not fulfilled.8. False otherwise. If set to False (which is the default) the session will be deleted when the user closes the browser. session['objects']. modified True if the session object detected a modification. The default is 31 days.append(42) # so mark it as modified yourself session. class flask. the others have useful defaults which you don’t need to change. We recommend just subclassing a dict and adding that mixin: class Session(dict. Be advised that modifica- tions on mutable structures are not picked up automatically.

The default implementation returns now + the permanent session lifetime configured on the application. This checks if the object is an instance of null_session_class by default. get_cookie_path(app) Returns the path for which the cookie should be valid. get_expiration_time(app. and falls back to APPLICATION_ROOT or uses / if it’s None. Likewise the is_null_session() method will perform a typecheck against this type. get_cookie_secure(app) Returns True if the cookie should be secure. This currently just returns the value of the SESSION_COOKIE_SECURE setting. pickle_based = False A flag that indicates if the session interface is pickle based. get_cookie_domain(app) Helpful helper method that returns the cookie domain that should be used for the session cookie if session cookies are used. This can be used 222 . New in version 0. This creates an instance of null_session_class by default. alias of NullSession open_session(app. This currently just returns the value of the SESSION_COOKIE_HTTPONLY config var. null_session_class make_null_session() will look here for the class that should be created when a null session is requested. get_cookie_httponly(app) Returns True if the session cookie should be httponly. is_null_session(obj) Checks if a given object is a null session. request) This method has to be implemented and must either return None in case the loading failed because of a configuration error or an instance of a ses- sion object which implements a dictionary like interface + the methods and attributes on SessionMixin. Null sessions are not asked to be saved.8. session) A helper method that returns an expiration date for the session or None if the session is linked to the browser session. The default imple- mentation uses the value from the SESSION_COOKIE_PATH config var if it’s set. This mainly aids the user experience because the job of the null session is to still support lookup without complaining but modifications are answered with a helpful error message of what failed. make_null_session(app) Creates a null session which acts as a replacement object if the real session support could not be loaded due to a configuration error.

class flask. The default is a compact JSON derived serializer with support for some extra Python types such as datetime objects or tuples. This is still called during a request context so if you absolutely need access to the request you can do that. response) This is called for actual sessions returned by open_session() at the end of the request. Will still allow read-only access to the empty session but fail on setting. salt = ‘cookie-session’ the salt that should be applied on top of the secret key for the signing of cookie based sessions. The default is sha1 key_derivation = ‘hmac’ the name of the itsdangerous supported key derivation. serializer = <flask. by Flask extensions to make a decision in regards to how to deal with the session object. 223 . static digest_method() the hash function to use for the signature. session.SessionMixin Expands a basic dictionary with an accessors that are expected by Flask exten- sions and users for the session.SecureCookieSessionInterface The default session interface that stores sessions in signed cookies through the itsdangerous module. The default is hmac. session) Indicates whether a cookie should be set now or not.10. If it’s set to False then a cookie is only set if the session is modified. The default behavior is controlled by the SESSION_REFRESH_EACH_REQUEST config variable.sessions.sessions. session_class alias of SecureCookieSession class flask. class flask. This is used by session backends to figure out if they should emit a set-cookie header or not.SecureCookieSession(initial=None) Base class for sessions based on signed cookies. should_set_cookie(app.TaggedJSONSerializer object> A python serializer for the payload.sessions.sessions. if set to True it’s always set if the session is permanent. class flask.sessions.NullSession(initial=None) Class used to generate nicer error messages if sessions are not available. New in version 0. This check is usually skipped if sessions get deleted.11. New in version 0. save_session(app.

7 Test Client class flask. The default mixin implementation just hardcodes False in.test_client() includes preset default environ- ment. new = False some session backends can tell you if a session is new. flask. 224 . 23.test_client() object in client.sessions. datetime).FlaskClient(*args. The default mixin imple- mentation just hardcodes True in.Client. but some backends will default this to false and detect changes in the dictionary for as long as changes do not happen on mutable structures in the session. For general information about how to use this class refer to werkzeug.test. Basic usage is outlined in the Testing Flask Applications chapter. Changed in version 0.sessions.session_json_serializer = <flask. This object provides dumping and loading methods similar to simplejson but it also tags certain builtin Python objects that commonly appear in sessions. Either catch this down yourself or use the permanent_session_lifetime attribute on the app which converts the result to an integer automatically. **kwargs) Works like a regular Werkzeug test client but has some knowledge about how Flask works to defer the cleanup of the request context stack to the end of a with body when used in a with statement. Use with caution.8.environ_base. markup objects. but that is not nec- essarily guaranteed.TaggedJSONSerializer object> A customized JSON serializer that supports a few extra types that we take for granted when serializing (tuples. permanent this reflects the '_permanent' key in the dict. which can be set after instantiation of the app. Cur- rently the following extended values are supported in the JSON it dumps: •Markup objects •UUID objects •datetime objects •tuples Notice The PERMANENT_SESSION_LIFETIME config key can also be an integer starting with Flask 0.12: app.testing. modified = True for some backends this will always be True.

23.current_app Points to the application handling the request.g. These two usages are now equivalent: user = getattr(flask. Flask provides you with a special object that ensures it is only valid for the active request and that will return different values for each request. a global variable is not good enough because it would break in threaded environments. Additionally as of 0.get('user'. This is a proxy. In a nutshell: it does the right thing. like it does for request and session. 23.9 Useful Functions and Classes flask. flask. None) It’s now also possible to use the in operator on it to see if an attribute is defined and it yields all keys on iteration. This can be used to modify the session that the test client uses.10 you can use the get() method to get an attribute or None (or the second argument) if it’s not set.8 Application Globals To share data that is valid for one request only from one function to another. None) user = flask. session_transaction(*args. For example a database connection or the user that is currently logged in. Starting with Flask 0. See Notes On Proxies for more information. Once the with block is left the session is stored back.g Just store on this whatever you want.g. **kwargs) When used in combination with a with statement this opens a session trans- action.session_transaction() as session: session['value'] = 42 Internally this is implemented by going through a temporary test request context and since session handling could depend on request variables this function accepts the same arguments as test_request_context() which are directly passed through.10 this is stored on the application context and no longer on the request context which means it becomes available if only the application context is bound and not yet a request. As of 0.11 you can use pop() and setdefault() in the same way you would use them on a dictionary. This is especially useful when combined with the Faking Resources and Context pattern for testing. 'user'. with client. This is useful for extensions that 225 .

spawn(do_some_work) return 'Regular response' 226 .Model): def __init__(self. class User(db. username.remote_addr = remote_addr Alternatively you can also just test any of the context bound objects (such as request or g for truthness): class User(db.Model): def __init__(self. username.copy_current_request_context(f ) A helper function that decorates a function to retain the current request context.username = username if remote_addr is None and has_request_context(): remote_addr = request.has_request_context() If you have code that wants to test if a request context is there or not this function can be used. flask. For instance.route('/') def index(): @copy_current_request_context def do_some_work(): # do some work here. The moment the function is deco- rated a copy of the request context is created and then pushed when the function is called.remote_addr self. This is powered by the application context and not by the request context.remote_addr = remote_addr New in version 0.7. remote_addr=None): self.. See Notes On Proxies for more information. Example: import gevent from flask import copy_current_request_context @app. This is a proxy. you may want to take advantage of request informa- tion if the request object is available. This is useful when working with greenlets. gevent. remote_addr=None): self.remote_addr self. want to support multiple applications running side by side. it can access flask. so you can change the value of this proxy by using the app_context() method. but fail silently if it is unavailable.. .username = username if remote_addr is None and request: remote_addr = request. flask.request like you # would otherwise in the view function.

url_build_error_handlers. **values) if url is None: # External lookup did not have a URL. error is the instance of BuildError. When it does. exc_value." # This is an example of hooking the build_error_handler. instead of raising BuildError. New in version 0. values): "Looks up an external URL when `url_for` cannot build a URL. exc_value. You can also just do a boolean check on the current_app object instead.). New in version 0. and endpoint and values are the argu- ments passed into url_for. To integrate applications. the current_app calls its url_build_error_handlers if it is not None.10: The _scheme parameter was added. **values) Generates a URL to the given endpoint with the method provided. This will reference the index function local to the current blueprint: url_for('. head over to the Quickstart. and not for handling 404 NotFound errors. # Here. endpoint. the whole pair is skipped. flask. Flask has a hook to intercept URL build errors through Flask. which can return a string to use as the result of url_for (instead of url_for‘s default to raise the BuildError exception) or re-raise the exception. in context of original traceback. 227 . lookup_url is some utility function you've built # which looks up the endpoint in some external URL registry.has_app_context() Works like has_request_context() but for the application context. New in version 0. tb = sys.url_build_error_handlers. return url app. # Re-raise the BuildError. An example: def external_url_handler(error.9.append(external_url_handler) Here. exc_type.index') For more information.exc_info() if exc_value is error: raise exc_type. flask.url_for(endpoint. In case blueprints are active you can shortcut references to the same blueprint by prefixing the local endpoint with a dot (.10. Variable arguments that are unknown to the target endpoint are appended to the generated URL as query arguments. tb else: raise error # url_for will use this result. If the value of a query argument is None. url = lookup_url(endpoint. Note that this is for building URLs outside the current application. The url_for function results in a BuildError when the current app does not have a URL for the given endpoint and values.

an absolute URL is generated. The default is werkzeug. New in version 0.10. Server address can be changed via SERVER_NAME configuration variable which defaults to localhost. The default behavior uses the same scheme as the current re- quest. If the first argument to the callable is an integer it will be looked up in the mapping. Because views do not have to return response objects but can return a value that is converted 228 .redirect(location. • _scheme – a string specifying the desired URL scheme. 305. • code – the redirect status code. redirects the client to the target location. 300 is not supported because it’s not a real redirect and 304 because it’s the answer for a request with a request with defined If-Modified-Since headers. this also can be set to an empty string to build protocol-relative URLs.abort() When passed a dict of code -> exception items it can be used as callable that raises exceptions. The _ex- ternal parameter must be set to True or a ValueError is raised. if called. • _anchor – if provided this is added as anchor to the URL. Parameters • location – the location the response should redirect to. 303.9: The _anchor and _method parameters were added. The rest of the arguments are forwarded to the exception constructor. Parameters • endpoint – the endpoint of the URL (name of the function) • values – the variable arguments of the URL rule • _external – if set to True.wrappers. New in version 0. New in version 0. flask. code=302. As of Werkzeug 0. or PREFERRED_URL_SCHEME from the app configuration if no request context is available. if it’s a WSGI application it will be raised in a proxy exception.9: Calls Flask. 302. • Response (class) – a Response class to use when instantiating a response. flask. defaults to 302.make_response(*args) Sometimes it is necessary to set additional headers in a view. Response=None) Returns a response object (a WSGI application) that. Supported codes are 301. flask. • _method – if provided this explicitly specifies an HTTP method.6: The location can now be a unicode string that is encoded using the iri_to_uri() function.10: The class used for the Response object can now be passed in.Response if un- specified.handle_build_error() on BuildError. and 307. New in version 0.

Example: @app. 404) The other use case of this function is to force the return value of a view function into a response which is helpful with view decorators: response = make_response(view_function()) response. This is useful to modify response objects.make_response() is invoked with it. If view looked like this and you want to add a new header: def index(): return render_template('index. the arguments are passed to the flask.html'.headers['X-Foo'] = 'Parachute' return response return 'Hello World!' 229 . flask. into a response object by Flask itself. Flask. flask. New in version 0.headers['X-Parachutes'] = 'parachutes are cool' return response This function accepts the very same arguments you can return from a view func- tion.make_response() function as tuple. it becomes tricky to add headers to it.route('/') def index(): @after_this_request def add_header(response): response.html'.Flask. This for example creates a response with a 404 error code: response = make_response(render_template('not_found. The function is passed the response object and has to return the same or a new one.headers['X-Parachutes'] = 'parachutes are cool' Internally this function does the following things: •if no arguments are passed. foo=42) You can now do something like this: def index(): response = make_response(render_template('index.html').after_this_request(f ) Executes a function after this request. it creates a new response argument •if one argument is passed.6. foo=42)) response. •if more than one argument is passed. This function can be called instead of using a return and you will get a response object which you can use to attach headers.

5: The add_etags. The mimetype guessing requires a filename or an attachment_filename to be provided. last_modified=None) Sends the contents of a file to the client. Alternatively a file object might be provided in which case X-Sendfile might not work and fall back to the traditional 230 . but you can also explicitly provide one. This will use the most efficient method available and configured. Changed in version 0. attach- ment_filename=None. Alternatively you can set the application’s use_x_sendfile attribute to True to directly emit an X-Sendfile header.9. New in version 0. This however requires support of the underlying webserver for X-Sendfile.12: The attachment_filename is preferred over filename for MIME-type detection. mimetype=None. New in version 0. pass a filepath via filename_or_fp or attachment_filename. otherwise attach an etag yourself. add_etags=True. Pass a filename if you are able to. Parameters • filename_or_fp – the filename of the file to send in latin-1.12: The filename is no longer automatically inferred from file objects. as_attachment=False.2.9: cache_timeout pulls its default from application config. New in version 0. By default it will try to guess the mimetype for you. For extra security you probably want to send certain files as at- tachment (HTML for instance).7: mimetype guessing and etag support for file objects was deprecated because it was unreliable.0 Changed in version 0. For instance think of a decorator that wants to add some headers without converting the return value into a response object. Changed in version 0. The default behavior is now to attach etags. cache_timeout=None. If you want to use automatic mimetype and etag support. This will allow the request to be answered with partial content response. con- ditional=False. Please never pass filenames to this function from user sources. If conditional=True and filename is provided. this method will try to upgrade the response stream to support range requests. By default it will try to use the WSGI server’s file_wrapper support. flask. You can turn this off by setting add_etags=False. This functionality will be removed in Flask 1. Changed in version 0. This is relative to the root_path if a relative path is speci- fied. This is more useful if a function other than the view function wants to modify a response.send_file(filename_or_fp. cache_timeout and conditional parameters were added. ETags will also be attached automatically if a filename is provided. you should use send_from_directory() instead. when None.

• last_modified – set the Last-Modified header to this value. New in version 0. • options – optional keyword arguments that are directly for- warded to send_file(). **options) Send a file from a given directory with send_file(). • add_etags – set to False to disable attaching of etags. auto detection happens as fallback. • attachment_filename – the filename for the attachment if it dif- fers from the file’s filename. this value is set by get_send_file_max_age() of current_app. as_attachment=True) Sending files and Performance It is strongly recommended to activate either X-Sendfile support in your web- server or (if no authentication happens) to tell the webserver to serve files for the given path on its own without calling into the web application for improved performance. 231 . flask. • conditional – set to True to enable conditional responses. filename.route('/uploads/<path:filename>') def download_file(filename): return send_from_directory(app. filename. this overrides its mtime.5. method. If a file was passed. otherwise an error will be raised. • mimetype – the mimetype of the file if provided. • cache_timeout – the timeout in seconds for the headers. Example usage: @app. • filename – the filename relative to that directory to download.config['UPLOAD_FOLDER']. If a file path is given. • as_attachment – set to True if you want to send this file with a Content-Disposition: attachment header. Parameters • directory – the directory where all the files are stored. This is a secure way to quickly expose static files from an upload folder or something similar. a datetime or timestamp. Make sure that the file pointer is positioned at the start of data to send before calling send_file().send_from_directory(directory. When None (default).

return '<a href="#">foo</a>' . Marks return value as markup string.flask.route('/wiki/<path:filename>') def wiki_page(filename): filename = safe_join(app.... Raises NotFound if one or more passed paths fall out of its boundaries..config['WIKI_FOLDER']...Markup Marks a string as being safe for inclusion in HTML/XML output without need- ing to be escaped. The constructor of the Markup class can be used for three different things: When passed an unicode object it’s assumed to be safe.read() # Read and process the file content. when passed an object with an HTML representation (has an __html__ method) that representation is used. and ” in string s to HTML-safe sequences. class flask. Markup is a direct subclass of unicode and pro- vides all the methods of unicode just that it escapes arguments passed and always returns Markup. >. filename) with open(filename. flask. def __html__(self): . ‘. • pathnames – the untrusted pathnames relative to that directory. 'rb') as fd: content = fd. The escape function returns markup objects so that double escaping can’t happen. otherwise the object passed is converted into a unicode string and then assumed to be safe: >>> Markup("Hello <em>World</em>!") Markup(u'Hello <em>World</em>!') >>> class Foo(object): . Parameters • directory – the trusted base directory. *pathnames) Safely join directory and zero or more untrusted pathnames components.safe_join(directory. This implements the __html__ interface a couple of frame- works and web applications use.. >>> Markup(Foo()) Markup(u'<a href="#">foo</a>') If you want object passed being always treated as unsafe you can use the escape() classmethod to create a Markup object: 232 . Example usage: @app. Use this if you need to display text that might contain such characters in HTML. <..escape(s) → markup Convert the characters &.

>>> Markup.escape("Hello <em>World</em>!")
Markup(u'Hello &lt;em&gt;World&lt;/em&gt;!')

Operations on a markup string are markup aware which means that all argu-
ments are passed through the escape() function:

>>> em = Markup("<em>%s</em>")
>>> em % "foo & bar"
Markup(u'<em>foo &amp; bar</em>')
>>> strong = Markup("<strong>%(text)s</strong>")
>>> strong % {'text': '<blink>hacker here</blink>'}
Markup(u'<strong>&lt;blink&gt;hacker here&lt;/blink&gt;</strong>')
>>> Markup("<em>Hello</em> ") + "<foo>"
Markup(u'<em>Hello</em> &lt;foo&gt;')

classmethod escape(s)
Escape the string. Works like escape() with the difference that for sub-
classes of Markup this function would return the correct subclass.
striptags()
Unescape markup into an text_type string and strip all tags. This also re-
solves known HTML4 and XHTML entities. Whitespace is normalized to
one:

>>> Markup("Main &raquo; <em>About</em>").striptags()
u'Main \xbb About'

unescape()
Unescape markup again into an text_type string. This also resolves known
HTML4 and XHTML entities:

>>> Markup("Main &raquo; <em>About</em>").unescape()
u'Main \xbb <em>About</em>'

23.10 Message Flashing

flask.flash(message, category=’message’)
Flashes a message to the next request. In order to remove the flashed mes-
sage from the session and to display it to the user, the template has to call
get_flashed_messages().
Changed in version 0.3: category parameter added.
Parameters
• message – the message to be flashed.
• category – the category for the message. The following values
are recommended: 'message' for any kind of message, 'error'

233

for errors, 'info' for information messages and 'warning' for
warnings. However any kind of string can be used as category.
flask.get_flashed_messages(with_categories=False, category_filter=[])
Pulls all flashed messages from the session and returns them. Further calls in the
same request to the function will return the same messages. By default just the
messages are returned, but when with_categories is set to True, the return value
will be a list of tuples in the form (category, message) instead.
Filter the flashed messages to one or more categories by providing those cate-
gories in category_filter. This allows rendering categories in separate html blocks.
The with_categories and category_filter arguments are distinct:
•with_categories controls whether categories are returned with message text
(True gives a tuple, where False gives just the message text).
•category_filter filters the messages down to only those matching the pro-
vided categories.
See Message Flashing for examples.
Changed in version 0.3: with_categories parameter added.
Changed in version 0.9: category_filter parameter added.
Parameters
• with_categories – set to True to also receive categories.
• category_filter – whitelist of categories to limit return values

23.11 JSON Support

Flask uses simplejson for the JSON implementation. Since simplejson is provided by
both the standard library as well as extension, Flask will try simplejson first and then
fall back to the stdlib json module. On top of that it will delegate access to the current
application’s JSON encoders and decoders for easier customization.
So for starters instead of doing:

try:
import simplejson as json
except ImportError:
import json

You can instead just do this:

from flask import json

For usage examples, read the json documentation in the standard library. The follow-
ing extensions are by default applied to the stdlib’s JSON module:
1. datetime objects are serialized as RFC 822 strings.

234

2. Any object with an __html__ method (like Markup) will have that method called
and then the return value is serialized as string.
The htmlsafe_dumps() function of this json module is also available as filter called
|tojson in Jinja2. Note that inside script tags no escaping must take place, so make
sure to disable escaping with |safe if you intend to use it inside script tags unless
you are using Flask 0.10 which implies that:

<script type=text/javascript>
doSomethingWith({{ user.username|tojson|safe }});
</script>

Auto-Sort JSON Keys
The configuration variable JSON_SORT_KEYS (Configuration Handling) can be set to false
to stop Flask from auto-sorting keys. By default sorting is enabled and outside of the
app context sorting is turned on.
Notice that disabling key sorting can cause issues when using content based HTTP
caches and Python’s hash randomization feature.

flask.json.jsonify(*args, **kwargs)
This function wraps dumps() to add a few enhancements that make life easier. It
turns the JSON output into a Response object with the applicationhttps://www.scribd.com/json mime-
type. For convenience, it also converts multiple arguments into an array or mul-
tiple keyword arguments into a dict. This means that both jsonify(1,2,3) and
jsonify([1,2,3]) serialize to [1,2,3].
For clarity, the JSON serialization behavior has the following differences from
dumps():
1.Single argument: Passed straight through to dumps().
2.Multiple arguments: Converted to an array before being passed to dumps().
3.Multiple keyword arguments: Converted to a dict before being passed to
dumps().
4.Both args and kwargs: Behavior undefined and will throw an exception.
Example usage:

from flask import jsonify

@app.route('/_get_current_user')
def get_current_user():
return jsonify(username=g.user.username,
email=g.user.email,
id=g.user.id)

This will send a JSON response like this to the browser:

235

{
"username": "admin",
"email": "admin@localhost",
"id": 42
}

Changed in version 0.11: Added support for serializing top-level arrays. This
introduces a security risk in ancient browsers. See JSON Security for details.
This function’s response will be pretty printed if it was not requested
with X-Requested-With: XMLHttpRequest to simplify debugging unless the
JSONIFY_PRETTYPRINT_REGULAR config parameter is set to false. Compressed (not
pretty) formatting currently means no indents and no spaces after separators.
New in version 0.2.
flask.json.dumps(obj, **kwargs)
Serialize obj to a JSON formatted str by using the application’s configured en-
coder (json_encoder) if there is an application on the stack.
This function can return unicode strings or ascii-only bytestrings by default
which coerce into unicode strings automatically. That behavior by default is
controlled by the JSON_AS_ASCII configuration variable and can be overridden
by the simplejson ensure_ascii parameter.
flask.json.dump(obj, fp, **kwargs)
Like dumps() but writes into a file object.
flask.json.loads(s, **kwargs)
Unserialize a JSON object from a string s by using the application’s configured
decoder (json_decoder) if there is an application on the stack.
flask.json.load(fp, **kwargs)
Like loads() but reads from a file object.
class flask.json.JSONEncoder(skipkeys=False, ensure_ascii=True,
check_circular=True, allow_nan=True,
sort_keys=False, indent=None, separators=None,
encoding=’utf-8’, default=None)
The default Flask JSON encoder. This one extends the default simplejson encoder
by also supporting datetime objects, UUID as well as Markup objects which are
serialized as RFC 822 datetime strings (same as the HTTP date format). In order
to support more data types override the default() method.
default(o)
Implement this method in a subclass such that it returns a serializable object
for o, or calls the base implementation (to raise a TypeError).
For example, to support arbitrary iterators, you could implement default
like this:

def default(self, o):
try:

236

iterable = iter(o)
except TypeError:
pass
else:
return list(iterable)
return JSONEncoder.default(self, o)

class flask.json.JSONDecoder(encoding=None, object_hook=None,
parse_float=None, parse_int=None,
parse_constant=None, strict=True, ob-
ject_pairs_hook=None)
The default JSON decoder. This one does not change the behavior from the de-
fault simplejson decoder. Consult the json documentation for more informa-
tion. This decoder is not only used for the load functions of this module but also
Request.

23.12 Template Rendering

flask.render_template(template_name_or_list, **context)
Renders a template from the template folder with the given context.
Parameters
• template_name_or_list – the name of the template to be ren-
dered, or an iterable with template names the first one existing
will be rendered
• context – the variables that should be available in the context
of the template.
flask.render_template_string(source, **context)
Renders a template from the given template source string with the given context.
Template variables will be autoescaped.
Parameters
• source – the source code of the template to be rendered
• context – the variables that should be available in the context
of the template.
flask.get_template_attribute(template_name, attribute)
Loads a macro (or variable) a template exports. This can be used to invoke a
macro from within Python code. If you for example have a template named
_cider.html with the following contents:

{% macro hello(name) %}Hello {{ name }}!{% endmacro %}

You can access this from Python code like this:

237

hello = get_template_attribute('_cider.html', 'hello')
return hello('World')

New in version 0.2.
Parameters
• template_name – the name of the template
• attribute – the name of the variable of macro to access

23.13 Configuration

class flask.Config(root_path, defaults=None)
Works exactly like a dict but provides ways to fill it from files or special dictio-
naries. There are two common patterns to populate the config.
Either you can fill the config from a config file:

app.config.from_pyfile('yourconfig.cfg')

Or alternatively you can define the configuration options in the module that calls
from_object() or provide an import path to a module that should be loaded.
It is also possible to tell it to use the same module and with that provide the
configuration values just before the call:

DEBUG = True
SECRET_KEY = 'development key'
app.config.from_object(__name__)

In both cases (loading from any Python file or loading from modules), only up-
percase keys are added to the config. This makes it possible to use lowercase
values in the config file for temporary values that are not added to the config or
to define the config keys in the same file that implements the application.
Probably the most interesting way to load configurations is from an environment
variable pointing to a file:

app.config.from_envvar('YOURAPPLICATION_SETTINGS')

In this case before launching the application you have to set this environment
variable to the file you want to use. On Linux and OS X use the export statement:

export YOURAPPLICATION_SETTINGS='/path/to/config/file'

On windows use set instead.
Parameters

238

from_pyfile(os. False otherwise. from_object(obj) Updates the values from the given object. New in version 0. This can either be an absolute filename or a filename relative to the root path. A dict object will not work with from_object() because the keys of a dict are not attributes of the dict class.config.environ['YOURAPPLICATION_SETTINGS']) Parameters • variable_name – name of the environment variable • silent – set to True if you want silent failure for missing files. silent=False) Updates the values in the config from a JSON file. True if able to load config. this is the appli- cation’s root_path. This function behaves as if the JSON object was a dictionary and passed to the from_mapping() function. • defaults – an optional dictionary of default values from_envvar(variable_name. • root_path – path to which files are read relative from.11. from_mapping(*mapping. This is basically just a shortcut with nicer error messages for this line of code: app. Parameters • filename – the filename of the JSON file. An object can be of one of the following two types: •a string: in this case the object with that name will be imported •an actual object reference: that object is used directly Objects are usually either modules or classes.11. • silent – set to True if you want silent failure for missing files. silent=False) Loads a configuration from an environment variable pointing to a configu- ration file. Returns bool. from_object() loads only the uppercase attributes of the module/class. **kwargs) Updates the config like update() ignoring items with non-upper keys. Example of module-based configuration: 239 . When the config object is created by the application. from_json(filename. New in version 0.

website.com' image_store_config = app. silent=False) Updates the values in the config from a Python file.from_object('yourapplication.7: silent parameter. The actual config should be loaded with from_pyfile() and ideally from a location not within the package because the package might be installed system wide.config['IMAGE_STORE_TYPE'] = 'fs' app.from_object(default_config) You should not use this function to load the actual configuration but rather configuration defaults. See Development / Production for an example of class-based configuration using from_object().config['IMAGE_STORE_BASE_URL'] = 'http://img. trim_namespace=True) Returns a dictionary containing a subset of configuration options that match the specified namespace/prefix. This function behaves as if the file was imported as module with the from_object() function.config. New in version 0.config. get_namespace(namespace. Parameters • filename – the filename of the config.default_config') from yourapplication import default_config app.get_namespace('IMAGE_STORE_') The resulting dictionary image_store_config would look like: { 'type': 'fs'. Example usage: app. 'path': '/var/apphttps://www.scribd.com/images'.config['IMAGE_STORE_PATH'] = '/var/apphttps://www.scribd.com/images' app. Parameters • namespace – a configuration namespace • lowercase – a flag indicating if the keys of the resulting dic- tionary should be lowercase 240 .website. app. • silent – set to True if you want silent failure for missing files. lowercase=True. This can either be an absolute filename or a filename relative to the root path.config.com' } This is often useful when configuration options map directly to keyword arguments in functions or class constructors. 'base_url': 'http://img. Parameters obj – an import name or object from_pyfile(filename.

23.route('/stream') def streamed_response(): @stream_with_context def generate(): yield 'Hello ' yield request. This is done for efficiency reasons and to make it less likely to encounter memory leaks with badly written WSGI middlewares.args['name'] yield '!' return Response(generate()) Alternatively it can also be used around a specific generator: from flask import stream_with_context.ext import foo New in version 0.ext This module acts as redirect import module to Flask extensions.route('/stream') def streamed_response(): def generate(): 241 . This function however can help you keep the context around for longer: from flask import stream_with_context. 23. the generator cannot access request bound information any more. If you want to use an extension named “Flask-Foo” you would import it from ext as follows: from flask.15 Stream Helpers flask.8 as the canonical way to import Flask extensions and makes it possible for us to have more flexibility in how we distribute extensions.14 Extensions flask. request. The downside is that if you are using streamed responses.11. request. • trim_namespace – a flag indicating if the keys of the resulting dictionary should not include the namespace New in version 0. Response @app.8. It was added in 0. Response @app.stream_with_context(generator_or_function) Request contexts disappear when the response is started on the server.

23. pop(exc=<object object>) Pops the request context and unbinds it by doing that. New in version 0. Make sure to properly pop() the stack yourself in that situation. match_request() Can be overridden by a subclass to hook into the matching of the request. It will create the URL adapter and request object for the WSGI environment provided. environ.ctx.9: Added the exc argument. it will evaluate all the functions registered on the application for teardown execution (teardown_request()). Changed in version 0. When the request context is popped. copy() Creates a copy of this request context with the same request object. The request context is automatically popped at the end of the request for you. Do not attempt to use this class directly. It is created at the beginning of the request and pushed to the _request_ctx_stack and removed at the end of it. instead use test_request_context() and request_context() to create this object. This can be used to move a request context to a different greenlet. In debug mode the request context is kept around if exceptions happen so that interactive debuggers have a chance to introspect the data. This will also trigger the execution of functions registered by the teardown_request() decorator. Because the actual request object is the same this cannot be used to move a request context to a different thread unless access to the request object is locked.4 this can also be forced for requests that did not fail and outside of DEBUG mode. 242 .RequestContext(app.9. You might find this helpful for unittests where you need the information from the context local around for a little longer. request=None) The request context contains all request relevant information. By setting 'flask. This is used by the test_client() for example to implement the deferred cleanup functionality. With 0.args['name'] yield '!' return Response(stream_with_context(generate())) New in version 0. yield 'Hello ' yield request.16 Useful Internals class flask.10. otherwise your unittests will leak memory._preserve_context' to True on the WSGI environment the context will not pop itself at the end of the request.

first_registration) Temporary holder object for registering a blueprint with the application._app_ctx_stack Works similar to the request context but only binds the application. request the current request object. push() Binds the request context to the current context. An in- stance of this class is created by the make_setup_state() method and later passed to all register callback functions.ctx. 243 ._request_ctx_stack The internal LocalStack that is used to implement all the context local objects used in Flask. This is mainly there for extensions to store data. The application context is also implicitly created if a request context is created but the application is not on top of the individual application context. options.session class flask. This is a documented instance and can be used by extensions and application code but the use is discouraged in general. url_adapter the URL adapter that was used to match the request.g object.blueprints.AppContext(app) The application context binds an application object implicitly to the current thread or greenlet. flashes an internal cache for the flashed messages. flask.9. flask. pop(exc=<object object>) Pops the app context. push() Binds the app context to the current context.BlueprintSetupState(blueprint. Example usage: from flask import _request_ctx_stack def get_session(): ctx = _request_ctx_stack. app. class flask.top if ctx is not None: return ctx. g an object with all the attributes of the flask. session the active session object. New in version 0. similar to how the RequestContext binds request informa- tion. The following attributes are always present on each layer of the stack: app the active Flask application.

view_func=None. signals. This is the case when blinker is installed. **extra): sender.debug('Rendering template "%s" with context %s'. **options) A helper method to register a rule (and optionally a view function) to the application. The signal is in- voked with the instance of the template as template and the context as dictionary (named context). None otherwise.signals_available True if the signaling system is available. subdomain = None The subdomain that the blueprint should be active for. endpoint=None. url_defaults = None A dictionary with URL defaults that is added to each and every URL that was defined with the blueprint. The following signals exist in Flask: flask. url_prefix = None The prefix that should be used for all URLs defined on the blueprint. app = None a reference to the current application blueprint = None a reference to the blueprint that created this setup state. context. The endpoint is automatically prefixed with the blueprint’s name.template_rendered This signal is sent when a template was successfully rendered. template. 23. context) 244 . first_registration = None as blueprints can be registered multiple times with the application and not everything wants to be registered multiple times on it. options = None a dictionary with all options that were passed to the register_blueprint() method. add_url_rule(rule. Example subscriber: def log_template_renders(sender.logger. this attribute can be used to figure out if the blueprint was registered in the past already.17 Signals New in version 0. template.6.name or 'string template'.

app) flask. context. The signal is invoked with the instance of the template as template and the context as dictionary (named context). the subscriber can access the request with the standard global proxies such as request.connect(log_response. template.logger.connect(log_template_renders. Example subscriber: def log_template_renders(sender. before any request process- ing happens. Example subscriber: def log_response(sender. where no exception handling happens. Example subscriber: 245 .debug('Request context is set up') from flask import request_started request_started.logger. Because the request context is already bound. ' 'Response: %s'. app) flask.name or 'string template'. app) flask.request_finished This signal is sent right before the response is sent to the client.connect(log_request.debug('Rendering template "%s" with context %s'. **extra): sender. Example subscriber: def log_request(sender. It is sent before the standard exception handling kicks in and even in debug mode. from flask import template_rendered template_rendered. It is passed the response to be sent named response. response.got_request_exception This signal is sent when an exception happens during request processing. The exception itself is passed to the sub- scriber as exception.before_render_template This signal is sent before template rendering process. context) from flask import before_render_template before_render_template. **extra): sender. response) from flask import request_finished request_finished.request_started This signal is sent when the request context is set up. template. **extra): sender.logger.debug('Request context is about to close down.connect(log_template_renders. app) flask.

Currently functions listening to this signal are called after the regular teardown handlers. app) As of Flask 0.logger. Currently functions listening to this signal are called after the regular teardown handlers. exception. def log_exception(sender. **extra): session.close() from flask import request_tearing_down request_tearing_down.close() from flask import appcontext_tearing_down appcontext_tearing_down. This is always called. Example subscriber: def close_db_connection(sender. even if an exception is caused. Example usage: from contextlib import contextmanager from flask import appcontext_pushed @contextmanager def user_set(app. The sender is the application. app) This will also be passed an exc keyword argument that has a reference to the exception that caused the teardown if there was one.appcontext_pushed This signal is sent when an application context is pushed.request_tearing_down This signal is sent when the request is tearing down. app) flask.appcontext_tearing_down This signal is sent when the app context is tearing down. but this is not something you can rely on. Example subscriber: def close_db_connection(sender. but this is not something you can rely on.connect(close_db_connection.9. exception) from flask import got_request_exception got_request_exception. flask. **kwargs): 246 .debug('Got exception during processing: %s'. **extra): session. **extra): sender. even if an exception is caused. user): def handler(sender. This is usually useful for unittests in order to temporarily hook in information. this will also be passed an exc keyword argument that has a reference to the exception that caused the teardown if there was one. For instance it can be used to set a resource early onto the g object.connect(close_db_connection. This is always called.connect(log_exception. flask.

views. flask.test_client() resp = c. **extra): recorded. g. message.connect(record.10.7.View Alternative way to use view functions. including connecting. category)) from flask import message_flashed message_flashed.get('/users/me') assert resp.18 Class-Based Views New in version 0. app) New in version 0.10.Namespace An alias for blinker. signal(name. flask. 'john'): c = app. class signals. The sender is the application. A subclass has to implement 247 .connected_to(handler.data == 'username=john' New in version 0. This usually falls in line with the appcontext_tearing_down signal. category. app): yield And in the testcode: def test_user_me(self): with user_set(app. The messages is sent as message keyword argument and the category as category. otherwise returns a fake signal that has a send method that will do nothing but will fail with a RuntimeError for all other operations.user = user with appcontext_pushed. doc=None) Creates a new signal for this namespace if blinker is available.appcontext_popped This signal is sent when an application context is popped. Example subscriber: recorded = [] def record(sender.base. 23. This class is available for Flask extensions that want to provide the same fallback system as Flask itself. class flask.10.message_flashed This signal is sent when the application is flashing a message. New in version 0.Namespace if blinker is available.append((message. otherwise a dummy class that creates fake signals.

as_view('myview')) When you want to decorate a pluggable view you will have to either do that when the view function is created (by wrapping the return value of as_view()) or you can use the decorators attribute: class SecretView(View): methods = ['GET'] decorators = [superuser_required] def dispatch_request(self): . dispatch_request() Subclasses have to override this method to implement the actual view func- tion code. **class_kwargs) Converts the class into an actual view function that can be used with the routing system.8. If methods is provided the methods do not have to be passed to the add_url_rule() method explicitly: class MyView(View): methods = ['GET'] def dispatch_request(self.add_url_rule('/hello/<name>'. You can place one or more decorators in this list and whenever the view function is created the result is automatically decorated. 248 . However since this moves parts of the logic from the class declaration to the place where it’s hooked into the routing system. Internally this generates a function on the fly which will in- stantiate the View on each request and call the dispatch_request() method on it. methods = None A list of methods this view can handle. The arguments passed to as_view() are forwarded to the constructor of the class. The decorators stored in the decorators list are applied one after another when the view function is created. *class_args. decorators = () The canonical way to decorate class-based views is to decorate the return value of as_view(). dispatch_request() which is called with the view arguments from the URL rout- ing system.. name): return 'Hello %s!' % name app. Note that you can not use the class based decorators since those would decorate the view class and not the generated view function! classmethod as_view(name. New in version 0. This method is called with all the arguments from the URL rule. view_func=MyView..

You can use the flask.as_view('counter')) 23. You can directly access the underlying Werkzeug routing system which is ex- posed as flask. The following converters are available: string accepts any text without a slash (the default) int accepts integers float like int but for floating point values path like the default but also accepts slashes any matches one of the items provided uuid accepts UUID strings Custom converters can be defined using flask.route() decorator. By default a variable part in the URL accepts any string without a slash however a different converter can be specified as well by using <converter:name>.views.Flask.MethodView Like a regular class-based view but that dispatches requests to particular meth- ods. For instance if you implement a method called get() it means it will re- spond to 'GET' requests and the dispatch_request() implementation will auto- matically forward your request to that.19 URL Route Registrations Generally there are three ways to define rules for the routing system: 1.url_map.add_url_rule('/counter'. 0) + 1 return 'OK' app. 3.Flask.class flask. Variable parts in the route can be specified with angular brackets (/user/<username>). 0) def post(self): session['counter'] = session.get('counter'.url_map. 2. Also options is set for you automatically: class CounterAPI(MethodView): def get(self): return session.route('/') def index(): pass 249 . Here are some examples: @app.Flask. view_func=CounterAPI.get('counter'.add_url_rule() function. Variable parts are passed to the view function as keyword arguments. You can use the flask.Flask.

a 404 not found is raised.route('/users/'. defaults={'page': 1}) @app.route('/<username>') def show_user(username): pass @app. 2.route('/post/<int:post_id>') def show_post(post_id): pass An important detail to keep in mind is how Flask deals with trailing slashes. 250 . If a rule does not end with a trailing slash and the user requests the page with a trailing slash. the user is automatically redirected to the same page with a trailing slash attached. This is consistent with how web servers deal with static files. You can also define multiple rules for the same function. They have to be unique however. Here for example is a definition for a URL that accepts an optional page: @app.@app. Defaults can also be specified. The only difference is that with the route parameter the view function is defined with the decorator instead of the view_func parameter. Here are the parameters that route() and add_url_rule() accept. If a rule ends with a slash and is requested without a slash by the user.route('/users/page/<int:page>') def show_users(page): pass This specifies that /users/ will be the URL for page one and /users/page/N will be the URL for page N. This also makes it possi- ble to use relative link targets safely. The idea is to keep each URL unique so the following rules apply: 1.

de. If it does.. POST etc.. This can be useful when working with decorators that want to customize the OPTIONS response on a per-view basis. sub. If do. Flask will look on the view function object itself if a methods attribute exists. Starting with Flask 0. 23. view_func the function to call when serving a request to the provided endpoint. • methods: If methods are not provided when the URL rule is added. it will pull the information for the methods from there.). OPTIONS is implicitly added and handled by the standard request handling. not specified the default subdomain is assumed.specifies the rule for the subdomain in case subdomain matching is in use. If this is not provided one can specify the function later by storing it in the view_functions dictionary with the endpoint as key.method == 'OPTIONS': # custom options handling here . By default a rule just listens for GET (and implicitly HEAD).20 View Function Options For internal usage the view functions can have some attributes attached to customize behavior the view function would normally not have control over. The following at- tributes can be provided optionally to either override some defaults to add_url_rule() or general behavior: • __name__: The name of a function is by default used as endpoint. methods is a list of methods this rule should be limited to (GET. 251 . rule the URL rule as string end.the endpoint for the registered URL rule. Full example: def index(): if request. A dictionary with defaults for this rule. main **op. • provide_automatic_options: if this attribute is set Flask will either force enable or disable the automatic implementation of the HTTP OPTIONS response. • required_methods: if this attribute is set.the options to be forwarded to the underlying Rule object. A change to tions Werkzeug is handling of method options. Additionally this will be prefixed with the name of the blueprint by default which cannot be customized from the function itself. See the example above for how faults defaults work. Flask will always add these methods when registering a URL rule even if the methods were explicitly overridden in the route() call.6. They have to be specified as keyword arguments. Flask itself assumes that the name point of the view function is the name of the endpoint if not explicitly stated. If endpoint is provided explicitly this value is used.

Typically it’s created au- tomatically by the FlaskGroup but you can also manually create it and pass it onwards as click object. In future versions of Flask this object will most likely play a bigger role. class flask. Parameters • add_default_commands – if this is True then the default run and shell commands wil be added.FlaskGroup(add_default_commands=True. group(*args. **kwargs) This works exactly like the method of the same name on a regular click.AppGroup(name=None. • create_app – an optional callback that is passed the script info and returns the loaded app. 252 . **kwargs) This works exactly like the method of the same name on a regular click.provide_automatic_options = False index. commands=None.ScriptInfo(app_import_path=None. Group but it defaults the group class to AppGroup. create_app=None. index) New in version 0.add_url_rule('/'. class flask. Normally a developer does not have to interface with this class but there are some very advanced use cases for which it makes sense to create an instance of this.8: The provide_automatic_options functionality was added. command(*args. return 'Hello World!' index. Not to be confused with FlaskGroup.cli. 23. add_version_option=True. For information as of why this is useful see Custom Scripts. Group but it wraps callbacks in with_appcontext() unless it’s disabled by passing with_appcontext=False.21 Command Line Interface class flask. This is usually not necessary to in- terface with as it’s used internally in the dispatching to click. create_app=None) Help object to deal with Flask applications.cli.cli. **attrs) This works similar to a regular click Group but it changes the behavior of the command() decorator so that it automatically wraps the functions in with_appcontext().methods = ['GET'. 'OPTIONS'] app. • add_version_option – adds the --version option. **extra) Special subclass of the AppGroup group that supports loading more commands from the configured Flask app.

app_import_path = None Optionally the import path for the Flask application. load_app() Loads the Flask app (if not yet loaded) and returns it. flask.run_command = <click. create_app = None Optionally a function that is passed the script info to create the instance of the application.Command object> Runs an interactive Python shell in the context of a given Flask application. flask. data = None A dictionary with arbitrary data that can be associated with this script info. If callbacks are registered directly to the app.cli object then they are wrapped with this function by default unless it’s disabled. This local server is recommended for development purposes only but it can also be used for simple intranet deployments.shell_command = <click.cli. This can be changed with the –with- threads option which will enable basic multithreading. The reloader and debugger are by default enabled if the debug flag of Flask is enabled and disabled otherwise.cli. This is useful for executing small snippets of management code without having to manually configuring the application. flask.core.cli. 253 . Calling this multiple times will just result in the already loaded app to be returned. The application will populate the default namespace of this shell according to it’s configuration. flask.cli.pass_script_info(f ) Marks a function so that an instance of ScriptInfo is passed as first argument to the click callback. By default it will not support any sort of concurrency at all to simplify debugging.with_appcontext(f ) Wraps a callback so that it’s guaranteed to be executed with the script’s applica- tion context.core.Command object> Runs a local development server for the Flask application.

254 .

legal information and changelog are here for the interested. 255 .Part III ADDITIONAL NOTES Design notes.

256 .

When you want to test something 257 . this section is for you. but why can’t Flask do that itself? Without such an explicit application object the following code: from flask import Flask app = Flask(__name__) @app. The most important one is that implicit appli- cation objects require that there may only be one instance at the time. 24. In Flask this is an instance of the Flask class. Each Flask application has to create an instance of this class itself and pass it the name of the module.1 The Explicit Application Object A Python web application based on WSGI has to have one central callable object that implements the actual application. like maintaining a stack of applications. Now the question is: when does a microframework need more than one application at the same time? A good example for this is unittesting. This should give you an idea about some of the design decisions that may appear arbitrary and surprising at first. but this causes some problems I won’t outline here in detail. There are ways to fake multiple applications with a single application object.route('/') def index(): return 'Hello World!' Would look like this instead: from hypothetical_flask import route @route('/') def index(): return 'Hello World!' There are three major reasons for this. CHAPTER TWENTYFOUR DESIGN DECISIONS IN FLASK If you are curious why Flask does certain things the way it does and not differently. especially in direct comparison with other frameworks.

Werkzeug will go quite far with that in that it will automatically redirect to a canonical URL if a route is ambiguous. Worse: many web- servers do not set the working directory to the directory of your application but to the document root which does not have to be the same folder. 24. just wrap it and you’re done (though there are better ways to do that so that you do not lose the reference to the application object wsgi_app()). This would not be possible without hacks if the object were created ahead of time for you based on a class that is not exposed to you. If you want to apply a WSGI middleware. you don’t have to remember anything else.3 One Template Engine Flask decides on one template engine: Jinja2. Why doesn’t Flask have a pluggable tem- plate engine interface? You can obviously use a different template engine. This means that you can declare routes in arbitrary order and they will still work as expected. Now obviously there are frameworks around that do not need any configuration and will still be able to load templates relative to your application module. Whenever you create a Flask instance you usually pass it __name__ as package name. but Flask will still configure Jinja2 for you. When the application object is deleted everything it allocated will be freed again. But there is another very important reason why Flask depends on an explicit instantia- tion of that class: the package name. Flask depends on that information to properly load resources relative to your module. The third reason is “explicit is better than implicit”.it can be very helpful to create a minimal application to test specific behavior. The current working directory is process-wide and if you are running multiple applications in one process (which could happen in a webserver without you knowing) the paths will be off. While that limitation that Jinja2 is always configured 258 . That object is your WSGI ap- plication. With Python’s outstanding support for reflection it can then access the package to figure out where the templates and static files are stored (see open_resource()). Furthermore this design makes it possible to use a factory function to create the appli- cation which is very helpful for unittesting and similar things (Application Factories).2 The Routing System Flask uses the Werkzeug routing system which was designed to automatically order routes by complexity. which is a very unreliable way to determine where the application is. But they have to use the current working directory for that. This is a requirement if you want to properly imple- ment decorator based routing since decorators could be fired in undefined order when the application is split into multiple modules. Another design decision with the Werkzeug routing system is that routes in Werkzeug try to ensure that URLs are unique. 24. Another thing that becomes possible when you have an explicit object lying around in your code is that you can subclass the base class (Flask) to alter specific behavior.

Flask uses Jinja2’s extensive autoescaping support. When it comes to connecting a template engine with an application or framework there is more than just rendering templates. A template abstraction layer that would not take the unique features of the template engines away is a science on its own and a too large undertaking for a microframework like Flask. But that’s about where similarities end. supports iterative template rendering. Also it provides ways to access macros from Jinja2 templates. but an extension could still depend on Jinja itself. Why shouldn’t it? If we look over to the Ruby side of web development there we have a protocol very similar to WSGI. packages with dependencies are no longer an issue and there are very few reasons against having libraries that depend on others. Template engines are like programming languages and each of those engines has a certain understanding about how things work. Thanks to recent developments in the Python package infrastructure. This Rack library has two equivalents in Python: WebOb (formerly Paste) and Werkzeug. On the surface they all work the same: you tell the engine to evaluate a template with a set of variables and take the return value as string. but on top of a library with the same name. a certain way to do template inheritance. Flask is a framework that takes advantage of the work already done by Werkzeug to properly interface WSGI (which can be a complex task at times). On the other hand an engine like Genshi is based on XML stream evaluation. support for reusable blocks (macros) that can be used from inside templates and also from Python code. 259 .will probably go away. the decision to bundle one template engine and use that will not. Paste is still around but from my understanding it’s sort of deprecated in favour of WebOb. For instance. Furthermore extensions can then easily depend on one template language being present. Jinja2 for example has an extensive filter sys- tem. but besides that it looks very much like a WSGI rendition for Ruby. Mako on the other hand treats templates similar to Python modules. Just that it’s called Rack there. configurable syntax and more. 24. uses Unicode for all operations. But nearly all applications in Ruby land do not work with Rack directly. You can easily use your own templating language.4 Micro with Dependencies Why does Flask call itself a microframework and yet it depends on two libraries (namely Werkzeug and Jinja2). The development of WebOb and Werkzeug started side by side with similar ideas in mind: be a good implementation of WSGI for other applications to take advantage. template inheritance by taking the availability of XPath into account and more.

260 . Everything else is up for extensions.5 Thread Locals Flask uses thread local objects (context local objects in fact. Why is this the case? Because people have different preferences and requirements and Flask could not meet those if it would force any of this into the core. What Flask is Not Flask will never have a database layer. Flask itself just bridges to Werkzeug to implement a proper WSGI application and to Jinja2 to handle templating. Everything else is up to you or extensions. Also see the Becoming Big section of the documentation for some inspiration for larger applications based on Flask. session and an extra object you can put your own things on (g).6 What Flask is.24. However Flask is just not designed for large applications or asynchronous servers. they support greenlet con- texts as well) for request. Why is that and isn’t that a bad idea? Yes it is usually not such a bright idea to use thread locals. It will not have a form library or anything else in that direction. The majority of web applications will need a template engine in some sort. 24. It also binds to a few common standard library packages such as logging. However not every application needs a SQL database. They cause troubles for servers that are not based on the concept of threads and make large applications harder to maintain. The idea of Flask is to build a good foundation for all applications. Flask wants to make it quick and easy to write a traditional web application.

it appeared that HTML was about to be replaced by XHTML. assumed they were properly using XHTML. This is likely because properly using XHTML can be quite painful. There are XML based template en- gines. One of the most important causes of pain is XML’s draconian (strict and ruthless) er- ror handling. When an XML parsing error is encountered. when end tags are optional they are not used. However. the browser is supposed to show the user an ugly error message. which as said before is often not set properly. the one used in Flask) which do not protect you from accidentally creating invalid XHTML. but they often come with a larger runtime overhead and are not as straightforward to use because they have to obey XML rules. barely any websites on the Internet are actual XHTML (which is HTML processed us- ing XML rules). but Internet Ex- plorer refuses to read files with that MIME type. so that the HTML is cleaner and faster to load. There are a couple of major reasons why this is the case. even if the document properly validates as XHTML. They wrote an XHTML doctype at the top of the document and self-closed all the necessary tags (<br> becomes <br/> or <br></br> in XHTML). instead of attempting to recover from the error and display what it can. However. 25. To properly work with XHTML. however. few people do. pro- grammers have to use the namespaced DOM interface with the XHTML namespace 261 . such as Kid and the popular Genshi. this document tries to answer some of the major ques- tions. what really determines XHTML/HTML processing in browsers is the MIME type. The majority of users. While it is relatively easy to configure Web servers to serve XHTML properly. XHTML also changed the way JavaScript is used. CHAPTER TWENTYFIVE HTML/XHTML FAQ The Flask documentation and example applications are using HTML5. Because there is much confusion about HTML and XHTML among developers. Most of the (X)HTML generation on the web is based on non-XML template engines (such as Jinja. You may no- tice that in many situations. The XHTML spec states that XHTML must be served with the MIME type application/xhtml+xml.1 History of XHTML For a while. So the valid XHTML was being treated as invalid HTML. One of them is Internet Explorer’s lack of proper XHTML support.

For example. In 2007.to query for HTML elements. or WHATWG (which was formed by the major browser vendors Apple.01.1 and the barely-used XHTML5.3 HTML versus XHTML The following table gives you a quick overview of features available in HTML 4.2 History of HTML5 Development of the HTML5 specification was started in 2004 under the name “Web Applications 1. as the XHTML 2 working group has been disbanded and HTML5 is being implemented by all major browser vendors. known as HTML5. as it was superseded by XHTML 1. and Opera) with the goal of writing a new and improved HTML specification. Currently. However. the specification was adopted as the basis of a new HTML specification under the umbrella of the W3C.0 is not included. browser vendors implemented the XHTML syntax over the syntax defined by the specification. Mozilla.1 and HTML5. XHTML 1.) 262 . since people were using XHTML-like tags along the lines of <link />. 25. based on existing browser behavior instead of unrealistic and backwards-incompatible specifi- cations. 25.0” by the Web Hypertext Application Technology Working Group. (XHTML 1. it appears that XHTML is losing traction. in HTML4 <title/Hello/ theoretically parses exactly the same as <title>Hello</title>.

<tbody>. for reasons detailed above.1 HTML5 <tag/value/ == <tag>value</tag> 1 <br/> supported 2 <script/> supported should be served as text/html 3 should be served as application/xhtml+xml strict error handling inline SVG inline MathML <video> tag <audio> tag New semantic tags like <article> 25. <li>. It should not be used in new code. or "). This means the following page in HTML5 is perfectly valid: <!doctype html> <title>Hello HTML5</title> 1 This is an obscure feature inherited from SGML. Some of these work because of the lenient error handling most browsers use when they encounter a markup error. '. 263 . missing end tags or unquoted attribute values). <dd>. The following constructs are optional in HTML5 by standard. so long as they contain no whitespace or special characters (like <. 2 This is for compatibility with server code that generates XHTML for tags such as <br>.0 is the last XHTML standard that allows to be served as text/html for backwards com- patibility reasons.01 XHTML1. <thead>. others are actually specified. • Quoting attributes. 3 XHTML 1. • Requiring boolean attributes to have a value. >.unlike XHTML. HTML4.4 What does “strict” mean? HTML5 has strictly defined parsing rules. It is usually not supported by browsers. but it also specifies exactly how a browser should react to parsing errors . but have to be supported by browsers: • Wrapping the document in an <html> tag • Wrapping header elements in <head> or the body elements in <body> • Closing the <p>. or <tfoot> tags. <tr>. which simply states parsing should abort. Some people are confused by apparently invalid syntax that still generates the expected results (for example. <dt>. <td>. <th>.

A good guide to new features in HTML5 is Mark Pilgrim’s soon-to-be-published book. Web Sockets. as well. Not all of them are supported in browsers yet. HTML5 is hard to beat. however. Web Workers. • The <canvas> tag.5 New technologies in HTML5 HTML5 adds many new features that make Web applications easier to write and to use. and <time> that make con- tent easier to understand. and offline applications. • The <audio> and <video> tags provide a way to embed audio and video without complicated add-ons like QuickTime or Flash.<div class=header> <h1>Hello HTML5</h1> <p class=tagline>HTML5 is awesome </div> <ul class=nav> <li><a href=/index>Index</a> <li><a href=/downloads>Downloads</a> <li><a href=/about>About</a> </ul> <div class=body> <h2>HTML5 is probably the future</h2> <p> There might be some other things around but in terms of browser vendor support. <dl> <dt>Key 1 <dd>Value 1 <dt>Key 2 <dd>Value 2 </dl> </div> 25. Dive Into HTML5. <header>. • New form control types like <input type="date"> that allow user agents to make entering and validating values easier. geolo- cation. so use caution. reducing the need for server-generated images to present data graphically. <nav>. • Semantic elements like <article>. 264 . Many other features have been added. which supports a powerful drawing API. • Advanced JavaScript APIs like Web Storage.

6 What should be used? Currently. • HTML5 adds several new features. 265 . To summarize the reasons given above: • Internet Explorer (which.25. currently leads in market share) has poor sup- port for XHTML. • It has the support of most browser vendors behind it. sadly. including semantic tags and the long-awaited <audio> and <video> tags. due to the more compli- cated namespacing API it requires. • It is much easier to write. and more compact. the answer is HTML5. There are very few reasons to use XHTML consider- ing the latest developments in Web browsers. • Many JavaScript libraries also do not support XHTML. For most applications. it is undoubtedly better to use HTML5 than XHTML.

266 .

never do that. be sure to always quote your attributes with either double or single quotes when using Jinja expressions in them: <a href="{{ href }}">the text</a> Why is this necessary? Because if you would not be doing that. For example an attacker could inject this piece of HTML+JavaScript: 267 . Flask configures Jinja2 to automatically escape all values unless explicitly told other- wise.1 Cross-Site Scripting (XSS) Cross site scripting is the concept of injecting arbitrary HTML (and with it JavaScript) into the context of a website. • sending out textfiles from uploaded files. 26. Flask tries to solve a few of these things for you. To remedy this. Some browsers are using content-type guessing based on the first few bytes so users could trick a browser to execute HTML. use the Content-Disposition: attachment header to prevent that problem. Another thing that is very important are unquoted attributes. developers have to properly escape text so that it cannot include arbitrary HTML tags. This should rule out all XSS problems caused in templates. To counter this possible attack vector. but there are a couple more you have to take care of yourself. For more information on that have a look at the Wikipedia article on Cross-Site Scripting. While Jinja2 can protect you from XSS issues by escaping HTML. there is one thing it cannot protect you from: XSS by attribute injection. an attacker could easily inject custom JavaScript handlers. but there are still other places where you have to be careful: • generating HTML without the help of Jinja2 • calling Markup on data submitted by users • sending out HTML from uploaded files. CHAPTER TWENTYSIX SECURITY CONSIDERATIONS Web applications usually face all kinds of security problems and it’s very hard to get everything right.

10 and lower. their profiles would get deleted while they are looking at images of fluffy cats. 268 .2 Cross-Site Request Forgery (CSRF) Another big problem is CSRF. If you don’t keep that in mind. and that cookie is sent with each request to a page. 26. After receiving the data on the server again. which does not exist in Flask. If an attacker now creates a page that sends a post request to that page with some JavaScript they just have to trick some users to load that page and their profiles will end up being deleted. you have implicit state man- agement. you would then have to compare the two tokens and ensure they are equal. When users would go to that page.cookie) When the user would then move with the mouse over the link.3 JSON Security In Flask 0. so only extremely old browsers are still vul- nerable. This is a very complex topic and I won’t outline it here in detail just mention what it is and how to theoretically prevent it. Unfortunately that includes requests triggered by 3rd party sites. But instead of showing the cookie to the user. Say you have a specific URL that. In combination with CSS injections the attacker might even make the element fill out the entire page so that the user would just have to have the mouse anywhere on the page to trigger the attack. If your authentication information is stored in cookies. 26. ECMAScript 5 closed this vulnerability. How can you prevent that? Basically for each request that modifies content on the server you would have to either use a one-time token and store that in the cookie and also transmit it with the form data. a good attacker might also execute any other JavaScript code. Why does Flask not do that for you? The ideal place for this to happen is the form validation framework. the cookie would be presented to the user in an alert window. jsonify() did not serialize top-level arrays to JSON.onmouseover=alert(document. This was because of a security vulnerability in ECMAScript 4. when you sent POST requests to will delete a user’s profile (say http://example. The state of “being logged in” is controlled by a cookie.com/user/delete). so this behavior was changed and jsonify() now supports serializing arrays. some people might be able to trick your application’s users with social engineering to do stupid things without them knowing. Imagine you were to run Facebook with millions of concurrent users and someone would send out links to images of little kittens. All of these browsers have other more serious vulnerabilities.

Flask will do the encoding and setting of the appropriate headers for you. This part of the documentation just tries to cover the very basics so that you have a pleasant experience with Unicode related things. Not only these libraries. However HTML which is usually trans- mitted on top of HTTP supports a large variety of character sets and which ones are used. also the system used to address doc- uments on servers (so called URIs or URLs). also the majority of web related Python libraries that deal with text. is totally Unicode based when it comes to text.1 Automatic Conversion Flask has a few assumptions about your application (which you can change of course) that give you basic and painless Unicode support: • the encoding for text on your website is UTF-8 • internally you will always use Unicode exclusively for text except for literal strings with only ASCII character points. Not only the protocol. you should probably read The Absolute Min- imum Every Software Developer Absolutely. are transmitted in an HTTP header. • encoding and decoding happens whenever you are talking over a protocol that requires bytes to be transmitted. 27. SQLAlchemy or your other ORM should take care of that. CHAPTER TWENTYSEVEN UNICODE IN FLASK Flask. So what does this mean to you? HTTP is based on bytes. To not make this too complex Flask just assumes that if you are sending Unicode out you want it to be UTF-8 encoded. The same is true if you are talking to databases with the help of SQLAlchemy or a similar ORM system. Some databases have a protocol that already transmits Unicode and if they do not. If you don’t know Unicode so far. 269 . like Jinja2 and Werkzeug. Positively Must Know About Unicode and Character Sets.

27.3 Encoding and Decoding Yourself If you are talking with a filesystem or something that is not really based on Unicode you will have to ensure that you decode properly when working with Unicode inter- face.4 Configuring Editors Most editors save as UTF-8 by default nowadays but in case your editor is not config- ured to do this you have to change it. 27. some special characters of latin letters without umlauts or anything fancy) you can use regular string literals ('Hello World'). contents. Here some common ways to set your editor to 270 . So do yourself a favour and limit yourself to UTF-8 for text files as well. Anyways.encode(charset)) 27. • if you need anything else than ASCII in a string you have to mark this string as Unicode string by prefixing it with a lowercase u. work with Unicode.x mean? • as long as you are using ASCII charpoints only (basically numbers. To tell the interpreter your encoding you can put the # -*.read().decode() method: def read_file(filename. Here the old problem that text files do not specify their encoding comes into play.into the first or second line of your Python source file. So for example if you want to load a file on the filesystem and embed it into a Jinja2 template you will have to decode it from the encoding of that file. What does working with Unicode in Python 2. (like u'Hänsel und Gretel') • if you are using non-Unicode characters in your Python files you have to tell Python which encoding your file uses.write(contents.coding: utf-8 -*.2 The Golden Rule So the rule of thumb: if you are not dealing with binary data. encode() method: def write_file(filename. • Jinja is configured to decode the template files from UTF-8. To load such a file with Unicode you can use the built-in str. charset='utf-8'): with open(filename. Again. I recommend UTF-8 for this pur- pose. So make sure to tell your editor to save the file as UTF-8 there as well.decode(charset) To go from Unicode into a specific charset such as UTF-8 you can use the unicode. 'r') as f: return f. charset='utf-8'): with open(filename. 'w') as f: f.

• Emacs: either use an encoding cookie or put this into your . Go to Settings -> Preferences . 271 . Select the “New Document/Default Directory” tab 3. you can select it in the same panel but this is not a requirement..vimrc file.emacs file: (prefer-coding-system 'utf-8) (setq default-buffer-file-coding-system 'utf-8) • Notepad++: 1. Select “UTF-8 without BOM” as encoding It is also recommended to use the Unix newline format.store as UTF-8: • Vim: put set enc=utf-8 to your . 2..

272 .

ext. 28. So for example if you plan to add support for a library named simplexml to Flask. often requires some repetitive steps to get a third party library working.py files. Flask sets up a redirect package called flask. This is a requirement because many people will use patterns like the Application Factories pattern to create their application as needed to aid unittests and to support multiple configurations. But what do extensions look like themselves? An extension has to ensure that it works with multiple Flask application instances at once.ext where users should import the ex- tensions from. If you want to create your own Flask extension for something that does not exist yet. Make sure to include the name “Flask” somewhere in that name and that you check the capitalization. This is done to transition from the old namespace packages.1 Anatomy of an Extension Extensions are all located in a package called flask_something where “something” is the name of the library you want to bridge. Because of that it is crucial that your application supports that kind of behavior. The name of the actual extension (the human readable name) however would be some- thing like “Flask-SimpleXML”.py file and registered on PyPI. Most importantly the extension must be shipped with a setup.something. See Extension Import Transition for more details. Also the development checkout link should work so that people can easily install the development version into their virtualenv without having to download the library by hand. CHAPTER TWENTYEIGHT FLASK EXTENSION DEVELOPMENT Flask. this guide to extension development will help you get your extension running in no time and to feel like users would expect your extension to behave. Because very often these steps could be abstracted to support multi- ple projects the Flask Extension Registry was created. being a microframework. If you for instance have a package called flask_something users would import it as flask. This is how users can then register dependencies to your extension in their setup. you would name your extension’s package flask_simplexml. 273 .

py The next file that is absolutely required is the setup. version='1. author='Your Name'. license='BSD'. 274 . The following contents are something you can work with: """ Flask-SQLite3 ------------- This is the description for that library """ from setuptools import setup setup( name='Flask-SQLite3'. Keep in mind that the Flask Extension Registry is a moderated place and libraries will be reviewed upfront if they behave as required. url='http://example. description='Very short description'.py file which is used to install your Flask extension.py LICENSE README Here’s the contents of the most important files: 28. py_modules=['flask_sqlite3']. zip_safe=False. The extension we want to create here will provide very basic support for SQLite3. author_email='your-email@example. # if you would be using a package instead use packages instead # of py_modules: # packages=['flask_sqlite3'].2.com'. long_description=__doc__.1 setup. MIT or more liberal license to be able to be enlisted in the Flask Extension Registry.0'.Flask extensions must be licensed under a BSD. 28. First we create the following folder structure: flask-sqlite3/ flask_sqlite3. include_package_data=True.com/flask-sqlite3/'.2 “Hello Flaskext!” So let’s get started with creating such a Flask extension.

28. extra_args]) that initializes the extension for that application. For example. There are two recommended ways for an extension to initialize: initialization functions: If your extension is called helloworld you might have a function called init_helloworld(app[. 28.py Now this is where your extension code goes.3 Initializing Extensions Many extensions will need some kind of initialization step. platforms='any'. 'Topic :: Internet :: WWW/HTTP :: Dynamic Content'. 'Operating System :: OS Independent'. 'Intended Audience :: Developers'. 'License :: OSI Approved :: BSD License'. For an example look at how the OAuth exten- sion works: there is an OAuth object that provides some helper functions like OAuth.2. 275 . consider an application that’s currently connecting to SQLite like the documentation suggests (Us- ing SQLite 3 with Flask). But how exactly should such an exten- sion look like? What are the best practices? Continue reading for some insight. classifiers=[ 'Environment :: Web Environment'.remote_app to create a reference to a remote application that uses OAuth. 'Programming Language :: Python'.2 flask_sqlite3. 'Topic :: Software Development :: Libraries :: Python Modules' ] ) That’s a lot of code but you can really just copy/paste that from existing extensions and adapt. It could attach before / after handlers etc. So how does the extension know the name of the application object? Quite simple: you pass it to it. install_requires=[ 'Flask' ]. classes: Classes work mostly like initialization functions but can later be used to further change the behavior.

What’s important about classes is that they encourage to be shared around on mod- ule level.config.teardown_request(self. 'sqlite3_db'): ctx.connect(current_app. # otherwise fall back to the request context if hasattr(app. the _app_ctx_stack is the correct one. app): app. For the SQLite 3 extension we will use the class-based approach because it will provide users with an object that handles opening and closing database connections.top if hasattr(ctx.setdefault('SQLITE3_DATABASE'.9. # Starting with Flask 0.close() @property 276 . app=None): self. # before that we need to use the _request_ctx_stack.init_app(app) def init_app(self. In that case.What to use depends on what you have in mind.config['SQLITE3_DATABASE']) def teardown(self.py for copy/paste: import sqlite3 from flask import current_app # Find the stack on which we want to store the database connection. try: from flask import _app_ctx_stack as stack except ImportError: from flask import _request_ctx_stack as stack class SQLite3(object): def __init__(self.4 The Extension Code Here’s the contents of the flask_sqlite3.app = app if app is not None: self. exception): ctx = stack.teardown) else: app. the object itself must not under any circumstances store any application specific state and must be shareable between different application.teardown) def connect(self): return sqlite3. 'teardown_appcontext'): app. 28.teardown_appcontext(self.sqlite3_db. ':memory:') # Use the newstyle teardown_appcontext if it's available.

The init_app will set the configuration for the database.connect() return ctx. Note that we’re falling back to the _request_ctx_stack. 'sqlite3_db'): ctx. 2. Next. In addition. Extensions should use the top context for stor- ing their own information with a sufficiently complex name.cfg') db = SQLite3(app) You can then use the database from views like this: @app.connection.from_pyfile('the-config.9 or later with the app context support. It will try to use the new- style app context handler and if it does not exist. 3. de- faulting to an in memory database if no configuration is supplied.sqlite3_db So here’s what these lines of code do: 1.top if ctx is not None: if not hasattr(ctx.execute(. if supplied. we define a connect method that opens a database connection. This method supports the factory pattern for cre- ating applications.. The init_app method exists so that the SQLite3 object can be instantiated with- out requiring an app object. you can use the database in the same way: 277 . The __init__ method takes an optional app object and. So why did we decide on a class-based approach here? Because using our extension looks something like this: from flask import Flask from flask_sqlite3 import SQLite3 app = Flask(__name__) app.top if the application is using an older version of Flask that does not support it.route('/') def show_all(): cur = db. Finally.top.config. the init_app method attaches the teardown handler.cursor() cur. Note here that we’re attaching our database connection to the top application context via _app_ctx_stack. 4. will call init_app. This is also the recommended way to handling resources: fetch resources lazily the first time they are used. def connection(self): ctx = stack. we add a connection property that on first access opens the database connection and stores it on the context.) Likewise if you are outside of a request but you are using Flask 0.. falls back to the request context one.sqlite3_db = self.

28.6 and older Due to the change in Flask 0. init_app does not assign app to self. 28. If the _app_ctx_stack does not exist because the user uses an old version of Flask. the sqlite3_db connection is closed. it must either use the current_app context local or change the API in a way that you can pass the application explicitly.app_context(): cur = db. the same connection to the sqlite3 database is accessible to anything that needs it for the duration of the request. before every request.5 Using _app_ctx_stack In the example above.cfg') db. This tells the extension: I am not interested in using multiple applications.init_app(app) Keep in mind that supporting this factory pattern for creating apps is required for approved flask extensions (described below). This is intentional! Class based Flask extensions must only store the application on the object when the application was passed to the constructor..with app. this variable is accessible using the connection property of SQLite3.execute(. The following pattern is a good way to support both: 278 . During the teardown of a request. a sqlite3_db variable is assigned to _app_ctx_stack. app = create_app('the-config. the init_app method is used to support the factory pattern for creating apps: db = Sqlite3() # Then later on. In a view function. it is recommended to fall back to _request_ctx_stack which is bound to a request. By using this pattern.) At the end of the with block the teardown handles will be executed automatically.7 regarding functions that are run at the end of the request your extension will have to be extra careful there if it wants to continue to support older versions of Flask.6 Teardown Behavior This is only relevant if you want to support Flask 0.top.connection. Note on init_app As you noticed.. When the extension needs to find the current application and it does not have a refer- ence to it. Additionally.cursor() cur.

so introduce your project on the mailinglist. 'teardown_request'): app.top ctx. 28. but also to avoid having multiple developers working on pretty much the same side by side.8 Approved Extensions Flask also has the concept of approved extensions.teardown_request(close_connection) else: app. and let other developers give you a helping hand with designing the API. 1. the project should find a new maintainer including full source hosting transition and PyPI access. it’s a very good idea to check out existing extensions on the Flask Extension Registry. If no main- tainer is available. The best Flask extensions are extensions that share common idioms for the API. Approved extensions are tested as part of Flask itself to ensure extensions do not break on new releases. An approved Flask extension must provide exactly one package or module named flask_extensionname. If you want to learn more. And this can only work if collaboration happens early.after_request(close_connection) Strictly speaking the above code is wrong.close() return response if hasattr(app.sqlite3_db. give access to the Flask core team.7 Learn from Others This documentation only touches the bare minimum for extension development. These approved extensions are listed on the Flask Extension Registry and marked appropriately.def close_connection(response): ctx = _request_ctx_stack. This not only to get an idea about what people might want to have from an extension. Remember: good API design is hard. If you feel lost there is still the mailinglist and the IRC chan- nel to get some ideas for nice looking APIs. However because the return value is discarded this will just work assuming that the code in between does not touch the passed parameter. In the event an extension author would like to move beyond the project. 279 . An approved Flask extension requires a maintainer. it might be a very good idea to get some more input. If you want your own extension to be approved you have to follow these guidelines: 0. 28. Especially if you do something nobody before you did. because teardown functions are passed the exception and typically don’t return anything.

ext.py file. As of Flask 0.11.11 but is now deprecated – you should use flask_foo.foo compatibility alias is still in Flask 0. The setup. even if the extension would be safe for zipping. 9. 280 .py file un- less a dependency cannot be met because it is not available on PyPI. 3.py description (and thus the PyPI description) has to link to the doc- umentation. Consequently we have rec- ommended to name extensions flask_foo over flaskext. • it must be possible to use the factory pattern for creating applications.foo.foo in that order.8 introduced a redirect import system as a compatibility aid for app develop- ers: Importing flask. of the form flaskext. The naming scheme for official extensions is Flask-ExtensionName or ExtensionName-Flask. This turned out to be problematic in practice be- cause it meant that multiple flaskext packages coexist. 10. APIs of approved extensions will be checked for the following characteristics: • an approved extension has to support multiple applications running in the same Python process.9 Extension Import Transition In early versions of Flask we recommended using namespace packages for Flask ex- tensions.foo for a long time.7 28. Flask 0. The flask. The license must be BSD/MIT/WTFPL licensed. website (if there is one) and there must be a link to automatically install the development version (PackageName==dev). 6. 7. Approved extensions must define all their dependencies in the setup. An extension currently has to support Python 2. most Flask extensions have transitioned to the new naming schema.ext.foo would try flask_foo and flaskext.6 as well as Python 2. For test suites invoked with make test the extension has to ensure that all dependencies for the test are installed automatically. The test suite also has to be part of the distribution. 5. 8. 4. If tests are invoked with python setup.py test.py test. The extension must have documentation that uses one of the two Flask themes for Sphinx documentation. test dependencies can be specified in the setup. The zip_safe flag in the setup script must be set to False. 2. It must ship a testing suite that can either be invoked with make test or python setup.

No tabs. 42.desc()) \ .filter(MyModel. 'and even more') For lists or tuples with many items.1 General Layout Indentation: 4 real spaces. 'with many parameters') \ . 23. 'with many parameters'.query. break immediately after the opening brace: items = [ 'this is the first'. 'to come in this line'.order_by(MyModel. Maximum line length: 79 characters with a soft limit for 84 if absolutely necessary. Continuing long statements: To continue a statement you can use backslashes in which case you should align the next line with the last dot or equal sign. align to the braces: this_is_a_very_long(function_call. 'set of items'. no exceptions. 'with more items'.scalar > 120) \ . Try to avoid too nested code by cleverly placing break. CHAPTER TWENTYNINE POCOO STYLEGUIDE The Pocoo styleguide is the styleguide for all Pocoo Projects.that_returns_an_object_with_an_attribute MyModel. This styleguide is a requirement for Patches to Flask and a recommendation for Flask ex- tensions. In general the Pocoo Styleguide closely follows PEP 8 with some small differences and extensions.limit(10) If you break in a statement with parentheses or braces.name. including Flask. 29. or indent four spaces: this_is_a_very_long(function_call. 'like this' ] 281 . continue and return state- ments.

name): self. 1.' % name class MyClass(object): """This is a simple docstring""" def __init__(self.name = name def get_annoying_name(self): return self.) as well on the inner side of parentheses. Example: def hello(name): print 'Hello %s!' % name def goodbye(name): print 'See you %s.Blank lines: Top level functions and classes are separated by two lines. ~ etc.g. Do not use too many blank lines to separate logical segments in code. • Whitespace is placed between binary operators.05 value = (item_value / item_count) * offset / exp value = my_list[index] value = my_dict['key'] Bad: exp = .upper() + '!!!!111' 29.: -.name.05 value = ( item_value / item_count ) * offset / exp value = (item_value/item_count)*offset/exp value=( item_value/item_count ) * offset/exp value = my_list[ index ] value = my_dict ['key'] Yoda statements are a no-go: Never compare constant with variable. always variable with constant: 282 . Good: exp = -1.2 Expressions and Statements General whitespace rules: • No whitespace for unary operators that are not words (e. everything else by one.

3 Naming Conventions • Class names: CamelCase. Good: if method == 'md5': pass Bad: if 'md5' == method: pass Comparisons: • against arbitrary types: == and != • against singletons with is and is not (eg: foo is not None) • never compare something with True or False (for example never do foo == False. If the function needs to access a shadowed builtin.real_name or x. On classes with keywords. but try to avoid instance checks in general. 29. rebind the builtin to a different name instead. Double underscores are reserved for mixin classes. trailing underscores are appended.username) 283 . C) instead of type(A) is C. Clashes with builtins are allowed and must not be resolved by appending an underline to the variable name. with acronyms kept uppercase (HTTPWriter and not HttpWriter) • Variable names: lowercase_with_underscores • Method and function names: lowercase_with_underscores • Constants: UPPERCASE_WITH_UNDERSCORES • precompiled regular expressions: name_re Protected members are prefixed with a single underscore. do not foo instead) Negated containment checks: use foo not in bar instead of not foo in bar Instance checks: isinstance(a. Check for features. Function and method arguments: • class methods: cls as first parameter • instance methods: self as first parameter • lambdas for properties might have the first parameter replaced with x like in display_name = property(lambda x: x.

29. """ Please keep in mind that proper copyrights and license files are a requirement for approved Flask extensions. In this case the closing triple quote is on its own line. If a comment is used to document an attribute.4 Docstrings Docstring conventions: All docstrings are formatted with reStructuredText as under- stood by Sphinx. put a colon after the opening pound sign (#): class User(object): #: the name of the user as unicode string name = Column(String) #: the sha1 hash of the password + inline salt pw_hash = Column(String) 284 .5 Comments Rules for comments are similar to docstrings. Depending on the number of lines in the docstring. see LICENSE_FILE for more details. """ Module header: The module header consists of an utf-8 encoding declaration (if non ASCII letters are used.29. they are laid out differently. Both are formatted with reStructured- Text. If it’s just one line. the closing triple quote is on the same line as the opening. :copyright: (c) YEAR by AUTHOR. :license: LICENSE_NAME.coding: utf-8 -*- """ package. otherwise the text is on the same line as the opening quote and the triple quote that closes the string on its own line: def foo(): """This is a simple docstring""" def bar(): """This is a longer docstring with so much information in there that it spans three lines. but it is recommended all the time) and a standard doc- string: # -*.module ~~~~~~~~~~~~~~ A brief description goes here.

3 or higher. which complicates how low level code handles HTTP data.2 and older are not supported. Python 3 changed how unicode and bytes are handled. so encoding issues should not affect you. You should use the latest versions of all Flask-related packages. 3. CHAPTER THIRTY PYTHON 3 SUPPORT Flask. Werkzeug wraps that information in high-level helpers. The majority of the upgrade work is in the lower-level libraries like Flask and Werkzeug. all of the examples in the Flask repository work on both Python 2 and 3 and did not require a single line of code changed. For example.9 were the first versions to introduce Python 3 support. and most Flask extensions support Python 3. This mainly affects WSGI middleware interacting with the WSGI environ data. Flask 0. but there are a few things to be aware of. 285 . You should start using Python 3 for your next project. its dependencies. You need to use Python 3.10 and Werkzeug 0. not the high-level application code.

286 .

. CHAPTER THIRTYONE UPGRADING TO NEWER RELEASES Flask itself is changing like any software is changing over time. etag generation or MIME-type guessing: response = send_file(open('/path/to/file. This section of the documentation enumerates all the changes in Flask from release to release and how you can change your code to have a painless updating experience.1 Version 0.txt' # Just pass the filepath directly response = send_file(fname) # Set the MIME-type and ETag explicitly response = send_file(open(fname). Use the pip command to upgrade your existing Flask installation by providing the --upgrade parameter: $ pip install --upgrade Flask 31. However every once in a while there are changes that do require some changes in your code or there are changes that make it possible for you to improve your own code quality by taking advantage of new features in Flask.1.12 31.) 287 . Most of the changes are the nice kind.. mimetype='text/plain') response.txt')) Any of the following is functionally equivalent: fname = '/path/to/file.set_etag(. the kind where you don’t have to change anything in your code to profit from a new release. This means that the following code will no longer automatically have X-Sendfile support.1 Changes to send_file The filename is no longer automatically inferred from file-like objects.

Note: There used to be a logic error allowing you to register handlers only for excep- tion instances.11 is an odd release in the Flask release cycle because it was supposed to be the 1.) The reason for this is that some file-like objects have a invalid or even misleading name attribute. respectively. Silently swallowing errors in such cases was not a satisfying solution. This was unintended and plain wrong.2 Version 0. 31. Instead the new LOGGER_HANDLER_POLICY configuration can be used to disable the default log handlers and custom log handlers can be set up. However because there was such a long lead time up to the release we decided to push out a 0.2. Trying to register a handler on an instance now raises ValueError.11 release first with some changes removed to make the transition easier. Additionally the default of falling back to application/octet-stream has been re- stricted.2. and therefore was replaced with the intended behavior of registering handlers only using exception classes and HTTP error codes.11 0.set_etag(.11 removed the debug_log_format attribute from Flask applications. See Error handlers for specifics. If Flask can’t guess one or the user didn’t provide one. In case you did track the master branch you will notice that flask --app is removed now. 288 . 31.. If you have been tracking the master branch which was 1. Now the inheritance hierarchy takes precedence and handlers for more specific exception classes are executed instead of more general ones. attachment_filename=fname) response. the function fails if no filename information was provided. 31.0 release. You need to use the environment variable to specify an application..1 Debugging Flask 0.2 Error handling The behavior of error handlers was changed.# Set `attachment_filename` for MIME-type guessing # ETag still needs to be manually set response = send_file(open(fname). The precedence of handlers used to be based on the decoration/call order of errorhandler() and register_error_handler().0 you might see some unexpected changes.

9 The behavior of returning tuples from a function was simplified. We also provide a migration utility called flask-ext-migrate that is supposed to automatically rewrite your imports for this.2. If you return a tuple it no longer defines the arguments for the response object you’re creating.exthook. 31. rv): if isinstance(rv. If you depend on the old behavior. but Flask will issue a flask. Flask also started storing the flask. This change should be transparent for you but it means that you now can store things on the g object when there is no request context yet but an applica- tion context.31. The old form still works.4 Extension imports Extension imports of the form flask.2. This better matches the behavior of render_template(). tuple): return self.ExtDeprecationWarning for each extension you import the old way. This change has been done in order to avoid the damage an attacker can do if the secret key is leaked.response_class(*rv) return Flask.ext.10 is that the cookie serialization format changed from pickle to a specialized JSON format.make_response(self. headers) where at least one item has to be provided.app_ctx_globals_class.request_globals_class attribute was renamed to flask. you should use flask_foo. rv) 289 .9 to 0. In order to not break people’s sessions it is possible to continue using the old session system by using the Flask-OldSessions extension. 31. The new sessions are by design much more restricted to only allow JSON with a few small extensions for tuples and strings with HTML markup. The old flask.3 Templating The render_template_string() function has changed to autoescape template vari- ables by default.foo are deprecated. 31.4 Version 0. you can add it easily by subclassing Flask: class TraditionalFlask(Flask): def make_response(self.Flask.3 Version 0.g object on the application context instead of the request context.10 The biggest change going from 0.Flask. When you upgrade you will notice two major changes: all sessions that were issued before the upgrade are invalidated and you can only store a limited amount of types in the session. it’s now always a tuple in the form (response. status.

We strongly recommend that you hand review the generated patchfile and only apply the chunks that look good.session which is the global session object. the app context stack makes sense for extensions which connect to databases. If you used the previously undocumented session support we urge you to upgrade. If you are using git as version control system for your project we recommend applying the patch with path -p1 < patchfile. This was since fixed but might require some changes in your test suites if you relied on this behavior.session the module that implements sessions and flask. Using the app context stack instead of the request context stack will make extensions more readily handle use cases outside of requests. Because we want to make upgrading as easy as possible we tried to counter the prob- lems arising from these changes by providing a script that can ease the transition. For instance.7 did not trigger teardown handlers when the test client was used in a with statement.If you maintain an extension that was using _request_ctx_stack before. Due to a bug in the test client Flask 0. We internally spread a lot of deprecation warnings all over the place to make it easy to find pieces of code that it was unable to upgrade. However as this is an automated tool it won’t be able to find all use cases and it might miss some.diff and then using the interactive commit feature to only apply the chunks that look good. 31. If invalid JSON data was submitted Flask will now raise a BadRequest exception in- stead of letting the default ValueError bubble up. If you were catching this down explicitly in the past as ValueError you will need to change this. 31. With that introduction we moved the implementation details for the session system into a new module called flask. This has the advantage that you no longer have to handle that error to avoid an internal server error showing up for the user. We also noticed that there was a naming collision between flask. To apply the upgrade script do the following: 290 .7 In Flask 0.7 we cleaned up the code base internally a lot and did some backwards in- compatible changes that make it easier to implement larger applications with Flask. sessions. The script scans your whole application and generates an unified diff with changes it assumes are safe to apply.8 Flask introduced a new session interface system.6 Version 0. please con- sider changing to _app_ctx_stack if it makes sense for your extension.5 Version 0.

diff 3.py > patchfile. Now you need to name the template templates/admin/index. Also source control systems like svn.html for a template file called templates/index. make sure to update your application to work with either filenames there or disable etag attaching and attach them yourself. Run it in the directory of your application: python flask-07-upgrade. hg or git have builtin support for applying unified diffs as generated by the tool. Apply the patch: patch -p1 < patchfile. Additionally the send_file() function is now issuing deprecation warnings if you depend on functionality that will be removed in Flask 0. 4. If you are working with windows and you lack the patch command line utility you can get it as part of various Unix runtime environments for windows including cygwin. Old code: return send_file(my_file_object) return send_file(my_file_object) New code: 291 . msysgit or ming32. Check the manual of your version control system for more information.1 Bug in Request Locals Due to a bug in earlier implementations the request local proxies now raise a RuntimeError instead of an AttributeError when they are unbound.6. This no longer is the case. 1.py 2. Review the generated patchfile. If you get a deprecation warning. The tool will not detect this so you will have to do that on your own. you should catch them with RuntimeError now. If you were using per-module template folders you need to move some templates around. Previously it was possible to use etags and mimetypes when file objects were passed.html.diff 5. This was unreliable and caused issues for a few setups. Download the script: flask-07-upgrade. Please note that deprecation warnings are disabled by default starting with Python 2.7. 31. Previously if you had a folder named templates next to a blueprint named admin the implicit template path automatically was admin/index.html. If you caught these exceptions with AttributeError before. In order to see the deprecation warnings that might be emitted you have to enabled them with the warnings module.11.

close() return response You are now encouraged to use this instead: @app. This makes it now a lot easier to write unit tests as you can prevent closing down of database connections for a while. In general we no longer recommend custom error handler attaching via assignments to the underlying dictionary due to the more complex internal handling to support arbitrary exception classes and blueprints. Unfortunately that change also made after-request work differently under error conditions. If you have database connection code that looks like this: @app.db.errorhandler() for more information. It’s not consistently skipped if exceptions happen whereas previously it might have been called twice to ensure it is executed at the end of the request.close() On the upside this change greatly improves the internal code flow and makes it easier to customize the dispatching and error handling.2 Upgrading to new Teardown Handling We streamlined the behavior of the callbacks for request handling.teardown_request def after_request(exception): if hasattr(g.6.get('/') # g.db.test_client() as client: resp = client.error_handlers it’s discour- aged to do so and in fact deprecated. but for things that absolutely must happen at the end of request we introduced the new teardown_request() decorator.after_request def after_request(response): g. 'db'): g. See Flask. For things that modify the response the after_request() decorators continue to work as expected.6. 292 . You can take advantage of the fact that the teardown callbacks are called when the re- sponse context is removed from the stack so a test can query the database after request handling: with app.db is still bound if there is such a thing # and here it's gone 31.3 Manual Error Handler Attaching While it is still possible to attach error handlers to Flask.return send_file(my_file_object. add_etags=False) 31.

They provide better semantics for various features and work better with large applications. If you continue to use the Module object which is deprecated. • Blueprints do not automatically provide static folders.4 Blueprint Support Blueprints replace the previous concept of “Modules” in Flask.html as the template name. This means that you need to add another subfolder with the blueprint’s name into that folder if you want blueprintname/template.error_handlers[403] = handle_error Into this: app. blueprint specific error handlers and a lot more.6. The upgrade script tries to guess that name but it might fail as this information could change at runtime.7) 31. Now the blueprints can provide template folders which are added to a general template searchpath.The proper upgrade is to change this: app. • Blueprints have an inverse behavior for url_for(). However we strongly recommend upgrading to the new blueprints as they provide a lot of useful improvement such as the ability to attach a blueprint multiple times. • Rendering templates was simplified. The script will inverse all calls to url_for() automatically for you.. Flask will restore the previous behavior as good as possible. Previously . 293 .errorhandler(403) def handle_error(e): . Same with static files: if you want to continue serving static files you need to tell the constructor explic- itly the path to the static folder (which can be relative to the blueprint’s module path). What changed? • Blueprints need explicit names.foo told url_for() that it should look for the endpoint foo on the application. Now it means “relative to current module”.. It will do this in a very eager way so you might end up with some unnecessary leading dots in your code if you’re not using modules. but there might be some cases where it fails to upgrade. handle_error) Alternatively you should just attach the function with a decorator: @app. The update script provided should be able to upgrade your applications automatically. They will also no longer automatically export templates from a folder called templates next to their loca- tion however but it can be enabled from the constructor.register_error_handler(403. Modules had an automatic name guessing scheme where the shortname for the module was taken from the last part of the import module. (Note that register_error_handler() is new in Flask 0.

use the create_jinja_environment() method instead. . There were a couple of internal refactoring so if you depend on undocumented internal details you probably have to adapt the imports. the default context processor will not override it with the current request object.5 Flask 0. 31. 31. If you have templates with different extensions you should override the select_jinja_autoescape() method. • Flask no longer supports zipped applications in this release. • The create_jinja_loader function is gone.html. .xhtml.7 Version 0. This change was made so that Flask behaves more like people expected it to work and how other systems handle request pre.4 For application developers there are no changes that require changes in your code. Instead it is configured to only happen on files ending with .9 Version 0.5 is the first release that comes as a Python package instead of a single mod- ule. If you depend on the order of execution of post-request functions. 294 . If for example request is as variable passed directly to the template. Removing support for this makes the Flask internal code easier to understand and fixes a couple of small issues that make debugging harder than necessary. now they are called in reverse order.6 comes with a backwards incompatible change which affects the order of after- request handlers. This makes it easier to extend context processors later to inject additional variables without breaking existing template not expecting them. and that extension has a unittest-mode you might want to link the activation of that mode to the new TESTING flag.xml and .6 Flask 0.31. Another change that breaks backwards compatibility is that context processors will no longer override values passed directly to the template rendering function. In case you are developing on a Flask extension however.and post- processing. The following changes may be relevant to your application: • autoescaping no longer happens for all templates. Previously they were called in the order of the registration. If you want to customize the Jinja loader now.8 Version 0.htm. be sure to change the order. This functionality might come back in future releases if there is demand for this feature.

31. Read all about that in the Message Flashing pattern.10.1 Configuration Support The configuration support makes it easier to write any kind of application that requires some sort of configuration. This makes it possible to render errors. 31. Also you can start logging warnings and errors into the logger when appropriately. you might be interested in attaching a proper log handler. For more information on that. All these are features that are 100% backwards compatible but you might want to take advantage of them.3 introduces configuration support and logging as well as categories for flash- ing messages.3 Categories for Flash Messages Flash messages can now have categories attached.secret_key = SECRET_KEY You no longer have to do that. 295 .10 Version 0. This is an opt-in feature because it requires some rethinking in the code. (Which most likely is the case for any application out there).10.3 Flask 0. If you run your application in production and want to profit from automatic error logging.debug = DEBUG app.2 Logging Integration Flask now configures a logger for you with some basic and useful defaults. warnings or regular messages differently for example. 31. read Application Errors. How this works is outlined in Configuration Handling. 31. If you previously had code like this: app.10. instead you can just load a configuration into the config object.

296 .

• Disable logger propagation by default for the app logger.12 Released on December 21st 2016. 32. • Make flask. • Correctly invoke response handlers for both regular request dispatching as well as error handlers.test_client includes preset default environment. unreleased • Fix crash when running under PyPy3. as per issue #104.safe_join able to join multiple paths like os.2 Version 0.path. • Add support for range requests in send_file. • Mimetype guessing and ETag generation for file-like objects in send_file has been removed. • Mimetype guessing in send_file now fails loudly and doesn’t fall back to application/octet-stream. 32. • the cli command now responds to –version. See pull request #1988. codename Punsch.11.2 Bugfix release.join (pull re- quest #1730). CHAPTER THIRTYTWO FLASK CHANGELOG Here you can see the full list of changes between each Flask release. 297 . • app. which can now be di- rectly set.1 Version 0. • Revert a behavior change that made the dev server crash instead of returning a Internal Server Error (pull request #2006). instead of per client. See pull request #1849.get. see pull request #1814.

Test. • Added flask and the flask. if set to False it will only be modified if the session actually modifies.Config.3 Version 0.3’s namespace loader.Config.Flask. See pull request #1872. • Added flask.Flask. • Added support for explicit root paths when using Python 3.32. • Added support for returning tuples in the form (response. • Error handlers that match specific classes are now checked first.get_namespace().1 Bugfix release.py from working.test_client_class. This makes it possible for an extension author to create exceptions 298 . If set to True a permanent session will be refreshed each request and get their lifetime extended. This can be configured with the new TEMPLATES_AUTO_RELOAD config key. 32. released on June 7th 2016. See JSON Security for details.run() method as it works faster and more reliable due to a different design and also replaces Flask-Script.cli module to start the local debug server through the click CLI system.11 Released on May 29th 2016. headers) from a view function.11. This is recommended over the old flask. • Added a workaround for a limitation in Python 3. exceptions).test_client() to support passing additional keyword arguments to the constructor of flask.4 Version 0. Non permanent sessions are not affected by this and will al- ways expire if the browser window closes.3’s namespace pack- ages.config_class. This intro- duces a security risk in ancient browsers. • Added **kwargs to flask. codename Absinthe. thereby allow- ing catching exceptions that are subclasses of HTTP exceptions (in werkzeug.jsonify(). • Made Flask support custom JSON mimetypes for incoming data. • Added SESSION_REFRESH_EACH_REQUEST config key that controls the set-cookie behavior. • Added flask. • Added flask.from_json(). • Added support to serializing top-level arrays to flask. • Fixed a bug that prevented FLASK_APP=foobar/__init__. • Added before_render_template signal. • Templates are no longer automatically reloaded outside of debug mode.

that will by default result in the HTTP error of their choosing. This should help users debug when the wrong templates are loaded. • Enforce blueprint handling in the order they were registered for template load- ing.jsonify now supports the datetime. • Add “pretty” and “compressed” separators definitions in jsonify() method. See https://github. • Allow custom Jinja environment subclasses (pull request #1422). because it is a convention that UNIX text files end with a newline and some clients don’t deal well when this newline is missing.date type (pull request #1326). • Flask will now log by default even if debug is disabled.Config. • flask. • flask. • Don’t leak exception info of already catched exceptions to context teardown han- dlers (pull request #1393). • Added the JSONIFY_MIMETYPE configuration variable (pull request #1728). • send_from_directory now raises BadRequest if the filename is invalid on the server OS (pull request #1763).test. • Deprecated request. • Removed deprecated module functionality.render_template_string by default (pull request #1515).templating.get_json().g now has pop() and setdefault methods. Re- duces JSON response size when JSONIFY_PRETTYPRINT_REGULAR=False by removing unnecessary white space included by default after separators. • Added the EXPLAIN_TEMPLATE_LOADING config flag which when enabled will in- struct Flask to explain how it locates templates.json. • Added flask.ext is now deprecated (pull request #1484). but may be caught with a custom error handler if desired.com/pallets/flask/pull/ 1262 – this came up originally as a part of https://github.json in favour of request. • Exceptions during teardown handling will no longer leave bad application con- texts lingering around. The log format is now hardcoded but the default log handling can be disabled through the LOGGER_HANDLER_POLICY configuration key.from_mapping(). • Turn on autoescape for flask. • flask. • Ported test suite to py. 299 .com/kennethreitz/ httpbin/issues/168 • The automatically provided OPTIONS method is now correctly disabled if the user registered an overriding rule with the lowercase-version options (issue #1288). • JSON responses are now terminated with a newline character.

find_package() with a useful mes- sage explaining why it is raised when a PEP 302 import hook is used without an is_package() method. 300 . • Fixed an issue with query parameters getting removed from requests in the test client when absolute URLs were requested.url_build_error_handlers reraises the BuildError.1 (bugfix release.js easier.10 Released on June 13th 2013. 32. released on June 14th 2013) • Fixed an issue where |tojson was not quoting single quotes which made the filter not work properly in HTML attributes.6 Version 0.10.7 Version 0. • Raise an AttributeError in flask.2 (bugfix release. • Made @before_first_request into a decorator as intended. This broke compati- bility with the common case of people putting binary data for token verification into the session. This should make using that filter with angular. 32. This will allow some uses that are potentially dangerous but should probably be per- mitted. • Changed logic of before first request handlers to flip the flag after invoking. • Fixed an issue where registering the same method twice for the same endpoint would trigger an exception incorrectly.10. codename Limoncello. • Fixed send_from_directory not expanding to the application root path correctly.helpers. release date to be announced) • Fixed broken test_appcontext_signals() test case. Now it’s possible to use that filter in single quoted attributes. • Added support for byte strings back to the session system. • Fixed Python 3 bug when a handler from app.5 Version 0.32. • Fixed an issue causing exceptions raised before entering a request or app context to be passed to teardown handlers. • Fixed an etags bug when sending a file streams with a name.

session and g objects will not be available and blueprint’s context processors are not called. • Templates can now be rendered without request context. This allows much easier customization of how JSON is handled by Flask or any Flask extension. 301 . One has to be very careful with those though because usage outside of macros might cause caching.session the session proxy. app_ctx_globals_class which is a better name to what it does since 0.session module alias. The behavior is slightly different as the request. • tojson used in templates is now safe by default due.g now can be used with the in operator to see what’s defined and it now is iterable and will yield all attributes stored. • flask.Flask. • flask. It’s disabled by default to not cause confusion with existing libraries that might expect flask. • Added template_global methods in addition to the already existing template_filter method family. • request.dumps to return bytestrings by default. This is not to be confused with flask. • Set the content-length header for x-sendfile. • Added wrapper module around simplejson and added default serialization of datetime objects. • Removed deprecated internal flask. • tojson filter now does not escape script blocks in HTML5 parsers.g is now stored on the app context instead of the request context.10 for more infor- mation. • Flask will now raise an error if you attempt to register a new function on an already used endpoint. This was allowed due to the different escaping behavior.• Changed default cookie serialization format from pickle to JSON to limit the impact an attacker can do if the secret key leaks. See Version 0. session and g are now also added as proxies to the template context which makes them available in imported templates. • flask.json.sessions instead to get the session module. • The config object is now available to the template as a real global and not through a context processor which makes it available even in imported templates by de- fault.Flask.10. • Added template_test methods in addition to the already existing template_filter method family. • Added an option to generate non-ascii encoded JSON which should result in less bytes being transmitted over the network. Use flask.g now gained a get() method for not erroring out on non existing items.request_globals_class got renamed to flask. • flask.

Request. • The builtin run method now takes the SERVER_NAME into account when picking the default port to run on.url_for() function now can also explicitly generate URL rules spe- cific to a given HTTP method. • Python requirements changed: requiring Python 2. 302 .url_for() function now can generate anchors to the generated links. • Added the JSONIFY_PRETTYPRINT_REGULAR configuration variable.get_json() as a replacement for the old flask. • Flask will no longer invoke the wrong error handlers if a proxy exception is passed through.on_json_loading_failed() now returns a JSON formatted response by default.6 or 2.8 Version 0.request. • Added support for copying of request contexts for better working with greenlets. • Added flask. • Changed how the teardown system is informed about exceptions. • The flask.json prop- erty. • Logger now only returns the debug log setting if it was not set explicitly.9 Released on July 1st 2012. • Added message_flashed signal that simplifies flashing testing.3 port. • Flask now orders JSON keys by default to not trash HTTP caches due to different hash seeds between different workers. • Request context preservation in debug mode now keeps the exception informa- tion around which means that teardown handlers are able to distinguish error from success cases. • The flask. Using them however is strongly discouraged as the interface was flawed.request. codename Campari. • Added a workaround for chrome’s cookies in localhost not working as intended with domain names.7 now to prepare for Python 3. • Added appcontext_pushed and appcontext_popped signals. If you were relying on them you can reintroduce them again yourself trivially. • Changed logic for picking defaults for cookie values from sessions to work better with Google Chrome. This is now more reliable in case something handles an exception halfway through the error handling process. • Removed custom JSON HTTP exception subclasses. • The flask. 32.

• Unregister a circular dependency between the WSGI environment and the re-
quest object when shutting down the request. This means that environ werkzeug.
request will be None after the response was returned to the WSGI server but has
the advantage that the garbage collector is not needed on CPython to tear down
the request unless the user created circular dependencies themselves.
• Session is now stored after callbacks so that if the session payload is stored in the
session you can still modify it in an after request callback.
• The flask.Flask class will avoid importing the provided import name if it can
(the required first parameter), to benefit tools which build Flask instances pro-
grammatically. The Flask class will fall back to using import on systems with
custom module hooks, e.g. Google App Engine, or when the import name is
inside a zip archive (usually a .egg) prior to Python 2.7.
• Blueprints now have a decorator to add custom template filters application wide,
flask.Blueprint.app_template_filter().
• The Flask and Blueprint classes now have a non-decorator method for adding
custom template filters application wide, flask.Flask.add_template_filter()
and flask.Blueprint.add_app_template_filter().
• The flask.get_flashed_messages() function now allows rendering flashed mes-
sage categories in separate blocks, through a category_filter argument.
• The flask.Flask.run() method now accepts None for host and port arguments,
using default values when None. This allows for calling run using configuration
values, e.g. app.run(app.config.get('MYHOST'), app.config.get('MYPORT')),
with proper behavior whether or not a config file is provided.
• The flask.render_template() method now accepts a either an iterable of tem-
plate names or a single template name. Previously, it only accepted a single
template name. On an iterable, the first template found is rendered.
• Added flask.Flask.app_context() which works very similar to the request con-
text but only provides access to the current application. This also adds support
for URL generation without an active request context.
• View functions can now return a tuple with the first instance being an instance
of flask.Response. This allows for returning jsonify(error="error msg"), 400
from a view function.
• Flask and Blueprint now provide a get_send_file_max_age() hook for sub-
classes to override behavior of serving static files from Flask when using
flask.Flask.send_static_file() (used for the default static file handler) and
send_file(). This hook is provided a filename, which for example allows chang-
ing cache controls by file extension. The default max-age for send_file and static
files can be configured through a new SEND_FILE_MAX_AGE_DEFAULT configuration
variable, which is used in the default get_send_file_max_age implementation.
• Fixed an assumption in sessions implementation which could break message
flashing on sessions implementations which use external storage.

303

• Changed the behavior of tuple return values from functions. They are no longer
arguments to the response object, they now have a defined meaning.
• Added flask.Flask.request_globals_class to allow a specific class to be used
on creation of the g instance of each request.
• Added required_methods attribute to view functions to force-add methods on reg-
istration.
• Added flask.after_this_request().
• Added flask.stream_with_context() and the ability to push contexts multiple
times without producing unexpected behavior.

32.9 Version 0.8.1

Bugfix release, released on July 1st 2012
• Fixed an issue with the undocumented flask.session module to not work properly
on Python 2.5. It should not be used but did cause some problems for package
managers.

32.10 Version 0.8

Released on September 29th 2011, codename Rakija
• Refactored session support into a session interface so that the implementation of
the sessions can be changed without having to override the Flask class.
• Empty session cookies are now deleted properly automatically.
• View functions can now opt out of getting the automatic OPTIONS implementa-
tion.
• HTTP exceptions and Bad Request errors can now be trapped so that they show
up normally in the traceback.
• Flask in debug mode is now detecting some common problems and tries to warn
you about them.
• Flask in debug mode will now complain with an assertion error if a view was
attached after the first request was handled. This gives earlier feedback when
users forget to import view code ahead of time.
• Added the ability to register callbacks that are only triggered once at the begin-
ning of the first request. (Flask.before_first_request())
• Malformed JSON data will now trigger a bad request HTTP exception instead
of a value error which usually would result in a 500 internal server error if not
handled. This is a backwards incompatible change.

304

• Applications now not only have a root path where the resources and modules are
located but also an instance path which is the designated place to drop files that
are modified at runtime (uploads etc.). Also this is conceptually only instance de-
pending and outside version control so it’s the perfect place to put configuration
files etc. For more information see Instance Folders.
• Added the APPLICATION_ROOT configuration variable.
• Implemented session_transaction() to easily modify sessions from the test en-
vironment.
• Refactored test client internally. The APPLICATION_ROOT configuration variable as
well as SERVER_NAME are now properly used by the test client as defaults.
• Added flask.views.View.decorators to support simpler decorating of plug-
gable (class-based) views.
• Fixed an issue where the test client if used with the “with” statement did not
trigger the execution of the teardown handlers.
• Added finer control over the session cookie parameters.
• HEAD requests to a method view now automatically dispatch to the get method
if no handler was implemented.
• Implemented the virtual flask.ext package to import extensions from.
• The context preservation on exceptions is now an integral component of Flask
itself and no longer of the test client. This cleaned up some internal logic and
lowers the odds of runaway request contexts in unittests.

32.11 Version 0.7.3

Bugfix release, release date to be decided
• Fixed the Jinja2 environment’s list_templates method not returning the correct
names when blueprints or modules were involved.

32.12 Version 0.7.2

Bugfix release, released on July 6th 2011
• Fixed an issue with URL processors not properly working on blueprints.

32.13 Version 0.7.1

Bugfix release, released on June 29th 2011
• Added missing future import that broke 2.5 compatibility.

305

• Fixed an infinite redirect issue with blueprints.

32.14 Version 0.7

Released on June 28th 2011, codename Grappa
• Added make_default_options_response() which can be used by subclasses to
alter the default behavior for OPTIONS responses.
• Unbound locals now raise a proper RuntimeError instead of an AttributeError.
• Mimetype guessing and etag support based on file objects is now deprecated for
flask.send_file() because it was unreliable. Pass filenames instead or attach
your own etags and provide a proper mimetype by hand.
• Static file handling for modules now requires the name of the static folder to
be supplied explicitly. The previous autodetection was not reliable and caused
issues on Google’s App Engine. Until 1.0 the old behavior will continue to work
but issue dependency warnings.
• fixed a problem for Flask to run on jython.
• added a PROPAGATE_EXCEPTIONS configuration variable that can be used to flip
the setting of exception propagation which previously was linked to DEBUG alone
and is now linked to either DEBUG or TESTING.
• Flask no longer internally depends on rules being added through the add_url_rule
function and can now also accept regular werkzeug rules added to the url map.
• Added an endpoint method to the flask application object which allows one to
register a callback to an arbitrary endpoint with a decorator.
• Use Last-Modified for static file sending instead of Date which was incorrectly
introduced in 0.6.
• Added create_jinja_loader to override the loader creation process.
• Implemented a silent flag for config.from_pyfile.
• Added teardown_request decorator, for functions that should run at the end of
a request regardless of whether an exception occurred. Also the behavior for
after_request was changed. It’s now no longer executed when an exception is
raised. See Upgrading to new Teardown Handling
• Implemented flask.has_request_context()
• Deprecated init_jinja_globals. Override the create_jinja_environment() method
instead to achieve the same functionality.
• Added flask.safe_join()
• The automatic JSON request data unpacking now looks at the charset mimetype
parameter.

306

• Don’t modify the session on flask.get_flashed_messages() if there are no mes-
sages in the session.
• before_request handlers are now able to abort requests with errors.
• it is not possible to define user exception handlers. That way you can provide
custom error messages from a central hub for certain errors that might occur dur-
ing request processing (for instance database connection errors, timeouts from
remote resources etc.).
• Blueprints can provide blueprint specific error handlers.
• Implemented generic Pluggable Views (class-based views).

32.15 Version 0.6.1

Bugfix release, released on December 31st 2010
• Fixed an issue where the default OPTIONS response was not exposing all valid
methods in the Allow header.
• Jinja2 template loading syntax now allows ”./” in front of a template load path.
Previously this caused issues with module setups.
• Fixed an issue where the subdomain setting for modules was ignored for the
static folder.
• Fixed a security problem that allowed clients to download arbitrary files if the
host server was a windows based operating system and the client uses back-
slashes to escape the directory the files where exposed from.

32.16 Version 0.6

Released on July 27th 2010, codename Whisky
• after request functions are now called in reverse order of registration.
• OPTIONS is now automatically implemented by Flask unless the application
explicitly adds ‘OPTIONS’ as method to the URL rule. In this case no automatic
OPTIONS handling kicks in.
• static rules are now even in place if there is no static folder for the module. This
was implemented to aid GAE which will remove the static folder if it’s part of a
mapping in the .yml file.
• the config is now available in the templates as config.
• context processors will no longer override values passed directly to the render
function.
• added the ability to limit the incoming request data with the new
MAX_CONTENT_LENGTH configuration value.

307

• the endpoint for the flask.Module.add_url_rule() method is now optional to
be consistent with the function of the same name on the application object.
• added a flask.make_response() function that simplifies creating response object
instances in views.
• added signalling support based on blinker. This feature is currently optional and
supposed to be used by extensions and applications. If you want to use it, make
sure to have blinker installed.
• refactored the way URL adapters are created. This process is now fully customiz-
able with the create_url_adapter() method.
• modules can now register for a subdomain instead of just an URL prefix. This
makes it possible to bind a whole module to a configurable subdomain.

32.17 Version 0.5.2

Bugfix Release, released on July 15th 2010
• fixed another issue with loading templates from directories when modules were
used.

32.18 Version 0.5.1

Bugfix Release, released on July 6th 2010
• fixes an issue with template loading from directories when modules where used.

32.19 Version 0.5

Released on July 6th 2010, codename Calvados
• fixed a bug with subdomains that was caused by the inability to specify the
server name. The server name can now be set with the SERVER_NAME config key.
This key is now also used to set the session cookie cross-subdomain wide.
• autoescaping is no longer active for all templates. Instead it is only active for .
html, .htm, .xml and .xhtml. Inside templates this behavior can be changed with
the autoescape tag.
• refactored Flask internally. It now consists of more than a single file.
• flask.send_file() now emits etags and has the ability to do conditional re-
sponses builtin.
• (temporarily) dropped support for zipped applications. This was a rarely used
feature and led to some confusing behavior.

308

20 Version 0. • added support for per-package template and static-file directories.git folder for themes.from_envvar() • removed some unused code from flask • release does no longer include development leftover files (.Handler and will log request handling exceptions to that logger when not in debug mode. • added TESTING switch that can activate unittesting helpers.3.5 due to the improved module support. • the logger switches to DEBUG mode now if debug is enabled.21 Version 0.22 Version 0. • the application now configures a logging. This makes it possible to receive mails on server errors for example.4 Released on June 18th 2010.1 Bugfix release. • because the Python standard library caches loggers. codename Schnaps • added support for categories for flashed messages. released on May 28th 2010 • fixed a error reporting bug with flask. 32. 32. • removed support for create_jinja_loader which is no longer used in 0.pyc files) 32. 309 . • added a helper function to expose files from any directory. • after_request() handlers are now also invoked if the request dies with an ex- ception and an error handling page kicks in.3 Released on May 28th 2010. codename Rakia • added the ability to register application wide error handlers from modules. • added support for context binding that does not require the use of the with state- ment for playing in the console. This can also be used to trigger custom requests that do not pop the request stack for testing. the name of the logger is configurable now to better support unittests.Config. built documentation in zip and pdf file and some . • test client has not the ability to preserve the request context for a little longer.

• add_url_rule() can now also register a view function. 32.2 Released on May 12th 2010.1 by default now to fix issues with chrome. • the request context is now available within the with statement making it possible to further push the request context or pop it.24 Version 0. • added support for Google Appengine. • refactored internal request dispatching. • added support for send_file() • module support and internal request handling refactoring to better support pluggable applications. codename Jägermeister • various bugfixes • integrated JSON support • added get_template_attribute() helper function.1 First public preview release.23 Version 0. • added external URL support. • sessions can be set to be permanent now on a per-session basis. 310 . • server listens on 127. • better error reporting on missing secret keys. 32.0.0. • added support for configurations.

1 Development Lead • Armin Ronacher <armin. the conditions are not modified and the disclaimer is present.com> 33.ronacher@active-4.1.2 Patches and Suggestions • Adam Zapletal • Ali Afshar • Chris Edgemon • Chris Grindstaff • Christopher Grebs • Daniel Neuhäuser • Dan Sully • David Lord @davidism • Edmond Burnett • Florent Xicluna 311 . 33. Furthermore you must not use the names of the authors to promote derivatives of the software without written consent.1.1 Authors Flask is written and maintained by Armin Ronacher and various contributors: 33. CHAPTER THIRTYTHREE LICENSE Flask is licensed under a three clause BSD License. It basically means: do whatever you want with it as long as the copyright in Flask sticks around. The full license text can be found below (Flask License). For the documentation and artwork different licenses apply.

with or without modification. Some rights reserved. 33. • The “Flask Artwork License” applies to the project’s Horn-Logo. See AUTHORS for more details.2 General License Definitions The following section contains the full license texts for Flask and the documentation. Redistribution and use in source and binary forms of the software as well as docu- mentation. • Georg Brandl • Jeff Widman @jeffwidman • Justin Quick • Kenneth Reitz • Keyan Pishdadian • Marian Sigler • Martijn Pieters • Matt Campell • Matthew Frazier • Michael van Tellingen • Ron DuPlain • Sebastien Estienne • Simon Sapin • Stephane Wirtel • Thomas Schranz • Zhao Xiaohong 33. • “AUTHORS” hereby refers to all the authors listed in the Authors section. are permitted provided that the following conditions are met: 312 . • The “Flask License” applies to all the source code shipped as part of Flask (Flask itself as well as the examples and the unittests) as well as documentation.3 Flask License Copyright (c) 2015 by Armin Ronacher and contributors.

) of the image. • The names of the contributors may not be used to endorse or promote products derived from this software without specific prior written permission. PRO- CUREMENT OF SUBSTITUTE GOODS OR SERVICES. DATA. WHETHER IN CONTRACT. • Redistributions of source code must retain the above copyright notice. OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE AND DOCUMENTATION. 33. STRICT LIABILITY. INDIRECT. EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT. THIS SOFTWARE AND DOCUMENTATION IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” AND ANY EXPRESS OR IMPLIED WAR- RANTIES. 313 . OR CONSEQUENTIAL DAMAGES (INCLUDING. OR PROFITS. but does not indicate endorsement by the project. Some rights reserved. INCLUDING. SPECIAL. OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY. INCIDENTAL. are permitted provided that the fol- lowing conditions are met: • Redistributions of source code must retain the above copyright notice and this list of conditions. EXEMPLARY. with or without modification. Note: we would appreciate that you make the image a link to http://flask.4 Flask Artwork License Copyright (c) 2010 by Armin Ronacher. this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. this list of conditions and the following disclaimer.org/ if you use it on a web page. Redistribution and use in source (the SVG file) and binary forms (rendered PNG files etc.pocoo. BUT NOT LIMITED TO. • Redistributions in binary form must reproduce the above copyright notice. BUT NOT LIMITED TO. THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DIS- CLAIMED. This logo or a modified version may be used by anyone to refer to the Flask project. • The names of the contributors to the Flask software (see AUTHORS) may not be used to endorse or promote products derived from this software without specific prior written permission. LOSS OF USE.

246 212 appcontext_tearing_down (in module add_url_rule() flask).cli). 243 tribute). 213 add_template_filter() (flask.Blueprint method). 252 add_app_template_filter() app_template_filter() (flask.ctx).Blueprint method). 244 method). 193 212 after_request() (flask. 193 314 . 243 args (flask. INDEX Symbols app_ctx_globals_class (flask. 213 A app_import_path (flask.Blueprint (flask. 246 (flask.views. 247 add_url_rule() (flask.Flask method).BlueprintSetupState AppGroup (class in flask.Flask method).Blueprint (flask.Blueprint method). 213 add_template_test() (flask. 192 _request_ctx_stack (in module flask).Flask method). method).cli.Flask at- _app_ctx_stack (in module flask). 192 B after_request_funcs (flask. 193 method).Request attribute).Flask at- tribute). 212 auto_find_instance_path() (flask.Blueprint method). 213 add_app_template_global() app_template_global() (flask.Request attribute).BlueprintSetupState before_app_request() (flask. 228 tribute). 216 add_url_rule() (flask. base_url (flask. 213 app (flask. 217 192 before_app_first_request() after_this_request() (in module flask). 252 method). 212 method). 213 app_context() (flask. 212 before_first_request_funcs (flask.Flask app_url_value_preprocessor() method).Blueprint app_template_test() (flask. 191 (flask.blueprints. 212 method). 192 before_first_request() (flask. 248 method). 229 (flask. 213 add_template_global() (flask. 190 method). 213 add_app_template_test() (flask.Blueprint method).Blueprint method).Flask method). AppContext (class in flask.ScriptInfo at- abort() (in module flask). 191 as_view() (flask.Blueprint method).Flask app_url_defaults() (flask.View class after_app_request() (flask. 243 app_errorhandler() (flask. appcontext_pushed (in module flask).Blueprint attribute).Flask attribute).Blueprint method).blueprints.Blueprint method). 212 method).Blueprint method).Blueprint method).Flask app_context_processor() (flask.Flask after_request() (flask. 243 191 appcontext_popped (in module flask).

copy() (flask.BlueprintSetupState current_app (in module flask).Flask method).Flask attribute). 195 213 digest_method() before_request() (flask. 220 flask.blueprints.Markup class method).Flask flask.json). 32 194 YOURAPPLICATION_SETTINGS. 224 248 FlaskGroup (class in flask. 193 command() (flask.Blueprint method). 189 data (flask. 34 context_processor() (flask. 232 create_global_jinja_loader() (flask. 223 tribute).json.SecureCookieSessionInterface before_request_funcs (flask. 238 endpoint() (flask.Blueprint method).Flask blueprints (flask. 217 create_url_adapter() (flask. 218 Config (class in flask). 233 D Flask (class in flask).Flask F method). 213 config (flask. 244 method). 217 context_processor() (flask.blueprints). 196 create_app (flask.Request attribute). 212 195 blueprint (flask.Request attribute).View attribute).Request attribute).ext (in module flask).AppGroup method).Blueprint environment variable method).RequestContext method). 194 endpoint() (flask. 193 dispatch_request() (flask. 194 files (flask.Response attribute).Flask at. default_config (flask.cli.cli). 213 FLASK_DEBUG. 252 default() (flask. 214 ule flask).Flask method).cli. 196 method).views. 195 BlueprintSetupState (class in do_teardown_request() (flask. 68 cookies (flask. copy_current_request_context() (in mod.sessions.Request attribute). first_registration 194 (flask.Flask attribute).before_request() (flask. static method). FlaskClient (class in flask.BlueprintSetupState dispatch_request() (flask.Flask method). 243 method). 32 method). 216 315 .ScriptInfo attribute). 195 config_class (flask.Flask attribute). FLASKR_SETTINGS. 194 FLASK_DEBUG. 217 flask. 216 error_handler_spec (flask.Flask attribute).View attribute). E 252 endpoint (flask.Flask attribute).Request attribute). 194 environ (flask. 194 create_jinja_environment() (flask. 218 do_teardown_appcontext() (flask. 34 decorators (flask.Flask method).testing). 193 method).Flask method).ctx.views.Request attribute). 189 data (flask.json).json (module). 195 dump() (in module flask. 241 data (flask.blueprints.Flask attribute). 248 blueprint (flask.Flask method).Blueprint method). 244 flash() (in module flask). 236 cli (flask. 195 242 errorhandler() (flask.Flask attribute). 236 C dumps() (in module flask.Flask attribute). Blueprint (class in flask). 225 attribute).Request attribute). 253 flask (module). escape() (flask. 234 debug (flask.JSONEncoder FLASKR_SETTINGS.cli.Flask extensions (flask. 226 errorhandler() (flask. 236 form (flask. 233 253 escape() (in module flask).ScriptInfo attribute). 193 (flask.

239 method).Flask method).sessions.Flask from_json() (flask.Blueprint attribute). 219 get_cookie_httponly() (flask. 222 method).sessions. 222 headers (flask.SessionInterface headers (flask.json). 197 has_request_context() (in module flask). 240 method). 239 handle_user_exception() (flask.Request attribute).SecureCookieSessionInterface attribute).sessions.Config method).Request attribute). 222 is_xhr (flask.Flask attribute). get_send_file_max_age() (flask. 222 is_null_session() get_expiration_time() (flask. jinja_env (flask. 235 H K handle_exception() (flask. 222 init_jinja_globals() (flask. 197 jinja_loader (flask. get_cookie_path() 198 (flask.AppGroup method).sessions. 197 from_object() (flask.Flask attribute).Request attribute).Config method).Flask 239 method). iter_blueprints() (flask. get_cookie_domain() 198 (flask. 198 (flask. 217 226 has_static_folder (flask. 198 flask). 199 got_request_exception (in module flask).SessionInterface inject_url_defaults() (flask.Flask has_app_context() (in module flask). full_path (flask. 237 245 JSONEncoder (class in flask. 218 J get_namespace() (flask. 217 method). 239 handle_http_exception() (flask. 214 g (in module flask). 252 jsonify() (in module flask.SessionInterface I method).Flask method).Config method).Request attribute).Request method).sessions. 225 has_static_folder (flask.Blueprint at- G tribute).Flask method). key_derivation 197 (flask. 198 get_template_attribute() (in module jinja_options (flask. 223 316 .Blueprint 198 method).Flask attribute). 236 group() (flask. 218 got_first_request (flask.Config method).json). 227 method).Flask 214 method). get_send_file_max_age() (flask. 197 from_mapping() (flask.from_envvar() (flask. method).SessionInterface is_json (flask.Flask attribute). 214 jinja_loader (flask.SessionInterface method).sessions. 237 json (flask.Flask attribute). handle_url_build_error() (flask.Response attribute). 198 240 jinja_environment (flask.json). json_decoder (flask.Config method).Flask attribute). 198 full_dispatch_request() (flask.Flask from_pyfile() (flask.Flask method).SessionInterface (flask. 217 get_flashed_messages() (in module flask).Config method). 218 method). 198 234 get_json() (flask. 198 197 json_encoder (flask.Request attribute). JSONDecoder (class in flask.Flask attribute).Flask attribute).cli.Flask attribute).sessions. 222 198 get_cookie_secure() instance_path (flask.

Flask attribute). (flask. 242 new (flask.Flask method). 243 tribute).ctx.Flask Markup (class in flask).sessions.sessions. 248 method).Request attribute).AppContext method).Blueprint method).SessionInterface method). 224 PEP 8.ctx. 236 NullSession (class in flask.View attribute).Flask attribute). 222 max_content_length (flask.Response attribute). M 214 make_config() (flask.views. 199 open_instance_resource() (flask. 242 attribute). 201 match_request() pickle_based (flask.sessions. 221 process_response() (flask.RequestContext method). 248 preserve_context_on_exception mimetype (flask.views). 224 method).ctx. 201 modified (flask.Blueprint permanent (flask.ScriptInfo method). 199 open_resource() (flask. 199 open_session() make_null_session() (flask.sessions). message_flashed (in module flask). 232 attribute). 281 null_session_class (flask.Flask method). 221 method). 220 method). 223 load_app() (flask.Flask method).ctx. 244 method).SessionMixin make_shell_context() (flask. pass_script_info() (in module flask.cli). 200 open_resource() (flask. modified (flask. 201 MethodView (class in flask.Flask attribute).Flask method).Request attribute).Flask methods (flask.sessions. 228 path (flask.L attribute). 218 logger (flask.Request log_exception() (flask. 200 253 make_response() (in module flask). 218 tribute).blueprints.Flask method).Flask at- module (flask.BlueprintSetupState (flask. 243 name (flask.RequestContext new (flask. 199 method). 222 make_null_session() options (flask.Request at- pop() (flask.Flask attribute).SessionMixin at- 201 tribute).SessionMixin at- Python Enhancement Proposals tribute). 214 permanent (flask. 200 push() (flask.Request attribute).sessions. 253 O loads() (in module flask. 200 permanent_session_lifetime (flask.Flask method).json).ctx.RequestContext (flask.json).Flask logger_name (flask.SessionInterface 317 . 201 (flask. 222 P make_response() (flask.Flask method).sessions. 199 method). 218 pop() (flask.session attribute).session attribute). 200 make_default_options_response() open_session() (flask.SessionInterface attribute). 202 N push() (flask. 247 242 method (flask.AppContext method).Flask attribute).session attribute).sessions. 222 load() (in module flask.SessionInterface 199 method). 236 on_json_loading_failed() (flask. 217 preprocess_request() (flask. 223 propagate_exceptions (flask. 217 make_setup_state() (flask. 220 (flask.cli.Flask method).

245 attribute). 253 shell_command (in module flask. 245 session_cookie_name (flask.Flask at- request_tearing_down (in module flask). 247 signals. 237 205 Request (class in flask).Request at.sessions).sessions). send_file() (in module flask).sessions).Namespace ScriptInfo (class in flask. 224 RFC session_transaction() RFC 822.sessions. 219 attribute). 228 select_jinja_autoescape() (flask.testing.cli).SecureCookieSessionInterface tribute). 203 set_cookie() (flask. routing_exception (flask. 223 request_class (flask. 205 246 session_interface (flask.Blueprint method).cli).Flask attribute). RequestContext (class in flask. 220 run_command (in module flask.Blueprint method).Blueprint render_template() (in module flask).Blueprint method). 202 send_static_file() (flask.signals_available (in module 318 .Flask save_session() (flask.Blueprint method). flask).Flask register() (flask.Flask method).Flask S method). 223 method).Flask register_error_handler() (flask.Flask method). 204 signals.Flask attribute). 220 request_context() (flask. 205 safe_join() (in module flask).sessions.Flask method). 203 flask.SessionInterface method). 202 session (class in flask). 216 serializer (flask. 221 tribute). route() (flask. 215 render_template_string() (in module send_static_file() (flask.Flask attribute). 252 method). register_error_handler() (flask.sessions). 223 should_ignore_error() (flask.Flask attribute). 223 script_root (flask.Request attribute).Flask at- salt (flask. 204 register_blueprint() (flask. 205 save_session() should_set_cookie() (flask.SessionInterface (flask.Flask method). 215 method). session_class 202 (flask.sessions). SecureCookieSessionInterface (class in 215 flask. 223 record_once() (flask. 237 method). 253 shell_context_processor() (flask. 247 secret_key (flask.SecureCookieSessionInterface request (class in flask).Response method). 231 shell_context_processors (flask. 219 SessionMixin (class in flask. 242 205 Response (class in flask). 230 202 send_file_max_age_default (flask.sessions. 215 send_from_directory() (in module flask).Flask 231 method). 223 run() (flask. 223 redirect() (in module flask). 234 (flask.cli).Flask method). 219 session_json_serializer (in module response_class (flask.R SecureCookieSession (class in record() (flask.SecureCookieSessionInterface request_finished (in module flask). 205 attribute). 204 method). 223 request_started (in module flask).sessions.Blueprint attribute).signals. 217 signal() (flask.ctx).Flask method).Namespace (class in flask). 203 SessionInterface (class in flask. 215 flask.sessions. tribute).FlaskClient method). 204 method). 215 224 route() (flask.

216 tribute). stream (flask. 216 use_x_sendfile (flask.Request attribute). 210 241 url_for() (in module flask). 206 V teardown_request_funcs (flask.Response attribute). 233 url_map (flask.Request attribute).Flask attribute). 209 216 url_default_functions (flask.blueprints. 207 wsgi_app() (flask. 205 tribute). 210 method).Flask view_args (flask.Request attribute).Flask attribute).views). 206 url_value_preprocessors (flask.Flask at.Blueprint teardown_appcontext() (flask. 247 template_context_processors (flask. 207 template_global() (flask. 209 testing (flask.Request attribute).Flask method).Flask at- static_folder (flask. 244 url_root (flask.Flask teardown_appcontext_funcs (flask. 209 url (flask. 209 U unescape() (flask.Flask method).Flask method). 227 striptags() (flask.Flask attribute).values (flask.Flask method).Flask method). 244 url_build_error_handlers (flask. 211 test_client() (flask. 216 method). flask).Flask attribute). tribute).blueprints. 244 attribute). 68 test_request_context() (flask.Blueprint tribute). template_rendered (in module flask).Flask method).Flask method).Request attribute). Y 209 YOURAPPLICATION_SETTINGS.BlueprintSetupStateurl_prefix (flask.BlueprintSetupState status_code (flask. 210 status (flask. 244 219 url_defaults() (flask.Flask at- static_folder (flask. 211 template_filter() (flask.Blueprint url_rule_class (flask.Flask at- teardown_request() (flask.cli). 208 test_client_class (flask.Markup method).Flask method).Response attribute). 211 teardown_request() (flask.Flask attribute). 209 trap_http_exception() (flask.Flask method). 210 attribute). 219 url_defaults (flask. 217 T url_rule (flask.BlueprintSetupState attribute).Request attribute). 210 subdomain (flask. 219 attribute).blueprints. 216 216 stream_with_context() (in module flask). 219 teardown_app_request() (flask.Flask attribute). url_defaults() (flask. 207 View (class in flask. 207 view_functions (flask. 210 method). 233 update_template_context() (flask.Blueprint attribute).Blueprint method). W 207 with_appcontext() (in module flask. 205 url_value_preprocessor() (flask.Flask attribute). 217 319 .Flask method). attribute).Markup method). 244 253 template_test() (flask.Flask method).Flask method). 216 url_value_preprocessor() (flask.