4.3 Overview of a Filter Handler

 

A filter handler is a function that can alter the input or the output of the server. There are two kinds of filters - input and output that apply to input from the client and output to the client respectively.

At this time mod_python supports only request-level filters, meaning that only the body of HTTP request or response can be filtered. Apache provides support for connection-level filters, which will be supported in the future.

A filter handler receives a filter object as its argument. The request object is available as well via filter.req, but all writing and reading should be done via the filter's object read and write methods.

Filters need to be closed when a read operation returns None (indicating End-Of-Stream).

The return value of a filter is ignored. Filters cannot decline processing like handlers, but the same effect can be achieved by using the filter.pass_on() method.

Filters must first be registered using PythonInputFilter or PythonOutputFilter, then added using the Apache Add/SetInputFilter or Add/SetOutputFilter directives.

Here is an example of how to specify an output filter, it tells the server that all .py files should processed by CAPITALIZE filter:

  PythonOutputFilter capitalize CAPITALIZE
  AddOutputFilter CAPITALIZE .py

And here is what the code for the capitalize.py might look like:

from mod_python import apache

def outputfilter(filter):

    s = filter.read()
    while s:
        filter.write(s.upper())
        s = filter.read()

    if s is None:
        filter.close()

When writing filters, keep in mind that a filter will be called any time anything upstream requests an IO operation, and the filter has no control over the amount of data passed through it and no notion of where in the request processing it is called. For example, within a single request, a filter may be called once or five times, and there is no way for the filter to know beforehand that the request is over and which of calls is last or first for this request, thought encounter of an EOS (None returned from a read operation) is a fairly strong indiciation of an end of a request.

Also note that filters may end up being called recursively in subrequests. To avoid the data being altered more than once, always make sure you are not in a subrequest by examining the req.main value.

For more information on filters, see http://httpd.apache.org/docs-2.0/developer/filters.html.

What is this????