Deployment

  1. CGI
  2. WSGI
  3. Twisted.Web
  4. Zomne
  5. Conclusion

Nevow includes two major phases for deciding what HTML to render. Object Traversal is the procedure by which a URL is mapped to a Python object which will perform the HTML generation. Page Rendering is the process by which data objects are combined with an HTML template to produce the final output.

Before any of this can take place, however, we must have an environment in which our Python code can run in response to an HTTP request, and HTML can be returned to the browser for rendering. This is called the Deployment Environment.

There are various deployment options for Nevow page code:

CGI

You can deploy Nevow on any webserver which uses the Common Gateway Interface. Using this method, your code is responsible for properly formatting and outputting the HTTP response headers, and Nevow is used only to generate the HTML body of your page. Here is the simplest possible CGI: 

#!/usr/bin/env python

print "Content-type: text/plain\r\n\r\n",

from nevow import rend, loaders

class HelloWorld(rend.Page):
    docFactory = loaders.stan("Hello, world!")

print HelloWorld().renderSynchronously()

With this simple CGI you can use the Nevow template loaders and standard nevow template interpolation techniques in your CGIs. However, you do not get any Object Traversal features, and you have to generate HTTP headers yourself. WSGI is a slightly higher-level deployment option which does not suffer these problems.

WSGI

WSGI is a python interface for plugging web applications into various HTTP server architectures. It is described in PEP 333, the Python Web Services Gateway Interface Python Enhancement Proposal. Nevow includes the nevow.wsgi module, which includes a createWSGIApplication function which takes a Page and returns a standard WSGI application callable. With the help of the run_with_cgi example gateway from the PEP (which I will omit here), our CGI example becomes shorter: 

#!/usr/bin/env python

from nevow import rend, loaders, wsgi

class HelloWorld(rend.Page):
    docFactory = loaders.stan("Hello, world!")

run_with_cgi(wsgi.createWSGIApplication(HelloWorld()))

Of course, you can use any available WSGI gateway to publish your application object, such as one of the gateways which comes with the PEAK toolkit. For example, here is a simple python module which creates a WSGI application which we will then deploy with PEAK's SimpleHTTPServer gateway:: 

## helloworld.py

from nevow import rend, loaders, wsgi

class HelloWorld(rend.Page):
    docFactory = loaders.stan("Hello, world!")

application = wsgi.createWSGIApplication(HelloWorld())

Save this file as "helloworld.py" somewhere on your PYTHONPATH and then run the following command: 

peak launch WSGI import:helloworld.application

This will bring up a SimpleHTTPServer running your Nevow code and launch a web browser to view the output. (TODO: I couldn't get this working immediately but I will seek assistance with PEAK and update the instructions once I do.)

Twisted.Web

A convenient and powerful way to deploy Nevow applications is inside a process running the twisted.web HTTP server. With Python, Twisted, and Nevow installed, you have all you need to run a Web Application, with no other dependencies or external HTTP servers such as Apache required. Running your Nevow applications under twisted.web also gives you access to some of the more advanced "Live" features of Nevow, such as nevow.livepage and nevow.canvas. Currently, these modules require more control over the HTTP socket than CGI or WSGI can provide. (This may change in the future.)

Deploying a Nevow application under twisted.web requires a little more boilerplate, but can be considerably easier to set up than other deployment options because there are no external dependencies. Note that normally you should declare your Page classes in modules external to the twisted configuration file, but everything is included in one file here for brevity. Here is the minimal configuration file required to use Nevow with twisted.web: 

from nevow import rend, loaders, appserver

class HelloWorld(rend.Page):
    docFactory = loaders.stan("Hello, world!")

from twisted.application import service, internet
application = service.Application("hello-world")
internet.TCPServer(8080, appserver.NevowSite(HelloWorld())).setServiceParent(application)

Save this file as "helloworld.tac" and start the server using the command: 

twistd -noy helloworld.tac

Then visit your twisted.web server by viewing the url "http://localhost:8080/" in your browser. See the twistd man page for more information about what twistd is capable of, including daemonizing the HTTP server.

Zomne

Warning Zomne is experimental. It may blow up your computer and require your first born son as a sacrifice. Zomne also only works in UNIX-like environments where unix domain sockets are available, and may not work on windows.

Zomne, or "Zombie Nevow", is a CGI written in C which can start up a long-running Application Server process if one is not already running. It then uses a simple custom protocol to transmit information about the HTTP request from the CGI process to the application server process.

Zomne combines the ease of deployment of the CGI environment with the speed and flexibility of the twisted.web long-running application server process model.

To use Zomne, you must first compile the CGI. cd into the directory created when unpacking the Nevow tarball, and compile the CGI: 

% gcc zomne.c

Move it into your cgi-bin: 

% mv a.out /Library/WebServer/CGI-Executables/nevow.cgi

Create a file which tells the cgi where to look for the application: 

% cat > /Library/WebServer/CGI-Executables/.nevow.cgi.dir
/Users/dp/zomne-test
^D

The CGI name can be anything, as long as there is a file with a prepended "." and a postfixed ".dir" in the same directory which contains the full path of a zomne application directory. Next, create the application directory: 

mkdir /Users/dp/zomne-test

Finally, create the zomne.tac file which the zomne.cgi will execute to start the long-running application server process: 

from nevow import rend, loaders, zomnesrv

class HelloWorld(rend.Page):
    docFactory = loaders.stan("Hello, world!")

from twisted.application import service, internet
application = service.Application('nevow-zomne-test')
internet.UNIXServer('zomne.socket', zomnesrv.ZomneFactory(HelloWorld())).setServiceParent(application)

Now, visiting the nevow.cgi URL through the web should render the Hello World page, after a pause while the server is starting up. Subsequent requests should be very fast, because the application server is already running, and the CGI merely has to forward the request to it.

Another useful capability of the zomne CGI process is the ability to control environment variables the CGI will use. Create a directory named "zomne_environ" in the application directory, and fill it with text files whose name will be the environment key and whose contents will be the environment value: 

% cd zomne-test
% mkdir zomne-environ
% cd zomne-environ
% cat > PYTHONPATH
/Users/dp/Projects/Nevow:/Users/dp/Projects/helloworld
^D

Conclusion

Nevow may be deployed in a number of environments, from the most restrictive to the most permissive. Writing a CGI can be an easy way to try out the Nevow templating mechanism, but can be slow. A long-running application server process can be a good way to get good performance as well as additional features such as in-memory server-side sessions, advanced automatic form handling with formless, and live page updating features such as nevow.livepage and nevow.canvas.

Which deployment option you choose will depend on the amount of control you have over your deployment environment, and what advanced features your application will require.

Index

Version: