1202 lines (932 sloc) 37.8 KB


This tutorial will walk you through basic but complete CherryPy applications that will show you common concepts as well as slightly more advanced ones.

Tutorial 1: A basic web application

The following example demonstrates the most basic application you could write with CherryPy. It starts a server and hosts an application that will be served at request reaching

Store this code snippet into a file named and execute it as follows:

$ python

This will display something along the following:

This tells you several things. The first three lines indicate the server will handle :mod:`signal` for you. The next line tells you the current state of the server, as that point it is in STARTING stage. Then, you are notified your application has no specific configuration set to it. Next, the server starts a couple of internal utilities that we will explain later. Finally, the server indicates it is now ready to accept incoming communications as it listens on the address In other words, at that stage your application is ready to be used.

Before moving on, let's discuss the message regarding the lack of configuration. By default, CherryPy has a feature which will review the syntax correctness of settings you could provide to configure the application. When none are provided, a warning message is thus displayed in the logs. That log is harmless and will not prevent CherryPy from working. You can refer to :ref:`the documentation above <perappconf>` to understand how to set the configuration.

Tutorial 2: Different URLs lead to different functions

Your applications will obviously handle more than a single URL. Let's imagine you have an application that generates a random string each time it is called:

Save this into a file named and run it as follows:

$ python

Go now to http://localhost:8080/generate and your browser will display a random string.

Let's take a minute to decompose what's happening here. This is the URL that you have typed into your browser: http://localhost:8080/generate

This URL contains various parts:

  • http:// which roughly indicates it's a URL using the HTTP protocol (see RFC 2616).
  • localhost:8080 is the server's address. It's made of a hostname and a port.
  • /generate which is the path segment of the URL. This is what CherryPy uses to locate an :term:`exposed` function or method to respond.

Here CherryPy uses the index() method to handle / and the generate() method to handle /generate

Tutorial 3: My URLs have parameters

In the previous tutorial, we have seen how to create an application that could generate a random string. Let's now assume you wish to indicate the length of that string dynamically.

Save this into a file named and run it as follows:

$ python

Go now to http://localhost:8080/generate?length=16 and your browser will display a generated string of length 16. Notice how we benefit from Python's default arguments' values to support URLs such as http://localhost:8080/generate still.

In a URL such as this one, the section after ? is called a query-string. Traditionally, the query-string is used to contextualize the URL by passing a set of (key, value) pairs. The format for those pairs is key=value. Each pair being separated by a & character.

Notice how we have to convert the given length value to an integer. Indeed, values are sent out from the client to our server as strings.

Much like CherryPy maps URL path segments to exposed functions, query-string keys are mapped to those exposed function parameters.

Tutorial 4: Submit this form

CherryPy is a web framework upon which you build web applications. The most traditional shape taken by applications is through an HTML user-interface speaking to your CherryPy server.

Let's see how to handle HTML forms via the following example.

Save this into a file named and run it as follows:

$ python

Go now to http://localhost:8080/ and your browser and this will display a simple input field to indicate the length of the string you want to generate.

Notice that in this example, the form uses the GET method and when you pressed the Give it now! button, the form is sent using the same URL as in the :ref:`previous <tut03>` tutorial. HTML forms also support the POST method, in that case the query-string is not appended to the URL but it sent as the body of the client's request to the server. However, this would not change your application's exposed method because CherryPy handles both the same way and uses the exposed's handler parameters to deal with the query-string (key, value) pairs.

Tutorial 5: Track my end-user's activity

It's not uncommon that an application needs to follow the user's activity for a while. The usual mechanism is to use a session identifier that is carried during the conversation between the user and your application.

Save this into a file named and run it as follows:

$ python

In this example, we generate the string as in the :ref:`previous <tut04>` tutorial but also store it in the current session. If you go to http://localhost:8080/, generate a random string, then go to http://localhost:8080/display, you will see the string you just generated.

The lines 30-34 show you how to enable the session support in your CherryPy application. By default, CherryPy will save sessions in the process's memory. It supports more persistent :ref:`backends <basicsession>` as well.

Tutorial 6: What about my javascripts, CSS and images?

Web applications are usually also made of static content such as javascript, CSS files or images. CherryPy provides support to serve static content to end-users.

Let's assume, you want to associate a stylesheet with your application to display a blue background color (why not?).

First, save the following stylesheet into a file named style.css and stored into a local directory public/css.

Now let's update the HTML code so that we link to the stylesheet using the http://localhost:8080/static/css/style.css URL.

Save this into a file named and run it as follows:

$ python

Going to http://localhost:8080/, you should be greeted by a flashy blue color.

CherryPy provides support to serve a single file or a complete directory structure. Most of the time, this is what you'll end up doing so this is what the code above demonstrates. First, we indicate the root directory of all of our static content. This must be an absolute path for security reason. CherryPy will complain if you provide only relative paths when looking for a match to your URLs.

Then we indicate that all URLs which path segment starts with /static will be served as static content. We map that URL to the public directory, a direct child of the root directory. The entire sub-tree of the public directory will be served as static content. CherryPy will map URLs to path within that directory. This is why /static/css/style.css is found in public/css/style.css.

Tutorial 7: Give us a REST

It's not unusual nowadays that web applications expose some sort of datamodel or computation functions. Without going into its details, one strategy is to follow the REST principles edicted by Roy T. Fielding.

Roughly speaking, it assumes that you can identify a resource and that you can address that resource through that identifier.

"What for?" you may ask. Well, mostly, these principles are there to ensure that you decouple, as best as you can, the entities your application expose from the way they are manipulated or consumed. To embrace this point of view, developers will usually design a web API that expose pairs of (URL, HTTP method, data, constraints).


You will often hear REST and web API together. The former is one strategy to provide the latter. This tutorial will not go deeper in that whole web API concept as it's a much more engaging subject, but you ought to read more about it online.

Lets go through a small example of a very basic web API mildly following REST principles.

Save this into a file named and run it as follows:

$ python

Before we see it in action, let's explain a few things. Until now, CherryPy was creating a tree of exposed methods that were used to match URLs. In the case of our web API, we want to stress the role played by the actual requests' HTTP methods. So we created methods that are named after them and they are all exposed at once by decorating the class itself with cherrypy.expose.

However, we must then switch from the default mechanism of matching URLs to method for one that is aware of the whole HTTP method shenanigan. This is what goes on line 27 where we create a :class:`~cherrypy.dispatch.MethodDispatcher` instance.

Then we force the responses content-type to be text/plain and we finally ensure that GET requests will only be responded to clients that accept that content-type by having a Accept: text/plain header set in their request. However, we do this only for that HTTP method as it wouldn't have much meaning on the other methods.

For the purpose of this tutorial, we will be using a Python client rather than your browser as we wouldn't be able to actually try our web API otherwise.

Please install requests through the following command:

$ pip install requests

Then fire up a Python terminal and try the following commands:

The first and last 500 responses stem from the fact that, in the first case, we haven't yet generated a string through POST and, on the latter case, that it doesn't exist after we've deleted it.

Lines 12-14 show you how the application reacted when our client requested the generated string as a JSON format. Since we configured the web API to only support plain text, it returns the appropriate HTTP error code.


We use the Session interface of requests so that it takes care of carrying the session id stored in the request cookie in each subsequent request. That is handy.


It's all about RESTful URLs these days, isn't it?

It is likely your URL will be made of dynamic parts that you will not be able to match to page handlers. For example, /library/12/book/15 cannot be directly handled by the default CherryPy dispatcher since the segments 12 and 15 will not be matched to any Python callable.

This can be easily workaround with two handy CherryPy features explained in the :ref:`advanced section <restful>`.

Tutorial 8: Make it smoother with Ajax

In the recent years, web applications have moved away from the simple pattern of "HTML forms + refresh the whole page". This traditional scheme still works very well but users have become used to web applications that don't refresh the entire page. Broadly speaking, web applications carry code performed client-side that can speak with the backend without having to refresh the whole page.

This tutorial will involve a little more code this time around. First, let's see our CSS stylesheet located in public/css/style.css.

We're adding a simple rule about the element that will display the generated string. By default, let's not show it up. Save the following HTML code into a file named index.html.

We'll be using the jQuery framework out of simplicity but feel free to replace it with your favourite tool. The page is composed of simple HTML elements to get user input and display the generated string. It also contains client-side code to talk to the backend API that actually performs the hard work.

Finally, here's the application's code that serves the HTML page above and responds to requests to generate strings. Both are hosted by the same application server.

Save this into a file named and run it as follows:

$ python

Go to and play with the input and buttons to generate, replace or delete the strings. Notice how the page isn't refreshed, simply part of its content.

Notice as well how your frontend converses with the backend using a straightfoward, yet clean, web service API. That same API could easily be used by non-HTML clients.

Tutorial 9: Data is all my life

Until now, all the generated strings were saved in the session, which by default is stored in the process memory. Though, you can persist sessions on disk or in a distributed memory store, this is not the right way of keeping your data on the long run. Sessions are there to identify your user and carry as little amount of data as necessary for the operation carried by the user.

To store, persist and query data you need a proper database server. There exist many to choose from with various paradigm support:

  • relational: PostgreSQL, SQLite, MariaDB, Firebird
  • column-oriented: HBase, Cassandra
  • key-store: redis, memcached
  • document oriented: Couchdb, MongoDB
  • graph-oriented: neo4j

Let's focus on the relational ones since they are the most common and probably what you will want to learn first.

For the sake of reducing the number of dependencies for these tutorials, we will go for the :mod:`sqlite` database which is directly supported by Python.

Our application will replace the storage of the generated string from the session to a SQLite database. The application will have the same HTML code as :ref:`tutorial 08 <tut08>`. So let's simply focus on the application code itself:

Save this into a file named and run it as follows:

$ python

Let's first see how we create two functions that create and destroy the table within our database. These functions are registered to the CherryPy's server on lines 85-86, so that they are called when the server starts and stops.

Next, notice how we replaced all the session code with calls to the database. We use the session id to identify the user's string within our database. Since the session will go away after a while, it's probably not the right approach. A better idea would be to associate the user's login or more resilient unique identifier. For the sake of our demo, this should do.


In this example, we must still set the session to a dummy value so that the session is not discarded on each request by CherryPy. Since we now use the database to store the generated string, we simply store a dummy timestamp inside the session.


Unfortunately, sqlite in Python forbids us to share a connection between threads. Since CherryPy is a multi-threaded server, this would be an issue. This is the reason why we open and close a connection to the database on each call. This is clearly not really production friendly, and it is probably advisable to either use a more capable database engine or a higher level library, such as SQLAlchemy, to better support your application's needs.

Tutorial 10: Make it a modern single-page application with React.js

In the recent years, client-side single-page applications (SPA) have gradually eaten server-side generated content web applications's lunch.

This tutorial demonstrates how to integrate with React.js, a Javascript library for SPA released by Facebook in 2013. Please refer to React.js documentation to learn more about it.

To demonstrate it, let's use the code from :ref:`tutorial 09 <tut09>`. However, we will be replacing the HTML and Javascript code.

First, let's see how our HTML code has changed:

Basically, we have removed the entire Javascript code that was using jQuery. Instead, we load the React.js library as well as a new, local, Javascript module, named gen.js and located in the public/js directory:

Wow! What a lot of code for something so simple, isn't it? The entry point is the last few lines where we indicate that we want to render the HTML code of the StringGeneratorBox React.js class inside the generator div.

When the page is rendered, so is that component. Notice how it is also made of another component that renders the form itself.

This might be a little over the top for such a simple example but hopefully will get you started with React.js in the process.

There is not much to say and, hopefully, the meaning of that code is rather clear. The component has an internal state in which we store the current string as generated/modified by the user.

When the user changes the content of the input boxes, the state is updated on the client side. Then, when a button is clicked, that state is sent out to the backend server using the API endpoint and the appropriate action takes places. Then, the state is updated and so is the view.

Tutorial 11: Organize my code

CherryPy comes with a powerful architecture that helps you organizing your code in a way that should make it easier to maintain and more flexible.

Several mechanisms are at your disposal, this tutorial will focus on the three main ones:

In order to understand them, let's imagine you are at a superstore:

  • You have several tills and people queuing for each of them (those are your requests)
  • You have various sections with food and other stuff (these are your data)
  • Finally you have the superstore people and their daily tasks to make sure sections are always in order (this is your backend)

In spite of being really simplistic, this is not far from how your application behaves. CherryPy helps you structure your application in a way that mirrors these high-level ideas.


Coming back to the superstore example, it is likely that you will want to perform operations based on the till:

  • Have a till for baskets with less than ten items
  • Have a till for disabled people
  • Have a till for pregnant women
  • Have a till where you can only using the store card

To support these use-cases, CherryPy provides a mechanism called a :ref:`dispatcher <dispatchers>`. A dispatcher is executed early during the request processing in order to determine which piece of code of your application will handle the incoming request. Or, to continue on the store analogy, a dispatcher will decide which till to lead a customer to.


Let's assume your store has decided to operate a discount spree but, only for a specific category of customers. CherryPy will deal with such use case via a mechanism called a :ref:`tool <tools>`.

A tool is a piece of code that runs on a per-request basis in order to perform additional work. Usually a tool is a simple Python function that is executed at a given point during the process of the request by CherryPy.


As we have seen, the store has a crew of people dedicated to manage the stock and deal with any customers' expectation.

In the CherryPy world, this translates into having functions that run outside of any request life-cycle. These functions should take care of background tasks, long lived connections (such as those to a database for instance), etc.

:ref:`Plugins <busplugins>` are called that way because they work along with the CherryPy :ref:`engine <cpengine>` and extend it with your operations.