XML Processing with Python

December 6, 1999

Sean McGrath

As part of our XML'99 coverage, we are pleased to bring you this taster from the "Working with XML in Python" tutorial led by Sean McGrath.


A century ago, when HTML and CGI ruled the waves, Perl dominated the Web programming scene. As the transition to XML on the Web gathers pace, competition for the hearts and minds of Web developers is heating up. One language attracting a lot of attention at the moment is Python.

In this article we will take a high level look at Python. We will use the time honored "Hello world" example program to illustrate the principle features of the language. We will then examine the XML processing capabilities of Python.

Python is free

Python is free. You will find downloadable source code plus pre-compiled executables on As you know, "free" is one of those words that is often heavily loaded on the Internet. Fear not. Python is free with a capital "F". You are free to do essentially anything you like with Python, including make commercial use of it or derivatives created from it.

Python is interpreted

Python is an interpreted language. Programs can execute directly from the plain text files that house them. Typically Python files have a .py extension. There is no compilation phase as far as the programmer is concerned. Just edit and run!

Python is portable

Python is portable. It runs on basically every computing platform of note, from mainframes to Palm Pilots and everything in between. Python uses a virtual machine architecture, similar in concept to Java's virtual machine. The Python interpreter "compiles" programs to virtual machine code on-the-fly. These compiled files (typically having a .pyc extension) are also portable. That is to say, if you wish to keep your source files hidden from your end-users you can simply ship the compiled .pyc files.

Python is easy to understand

Python is very easy to understand. Here is a Python program that prints the string "Hello world":

print "Hello world"

I think you will agree that programming a "Hello world" application cannot get much simpler than that! To execute this program, you put it in a text file, say, and feed it to the Python interpreter like this:


The output is, surprise, surprise:

Hello world

Note the complete lack of syntactic baggage in the program. There are no mandatory keywords or semi-colons required to get this simple job done. This spartan, no-nonsense approach to syntax is one of the hallmarks of Python and applies equally well to large Python programs.

Python is interactive

By invoking the Python interpreter (typically by typing python on a UNIX/Linux system, or running the "IDLE" application on Windows), you will find yourself in an environment where you can execute Python statements interactively. As an example, here is the "Hello world" application again:

>>> print "Hello world"

This will output:

Hello world

Note that the ">>>" above is Python's command prompt. The interactive mode is an excellent environment for playing around with Python. It is also indispensable as a fully programmable calculator!

Python is WYSIWYG

Python is sometimes referred to as a WYSIWYG programming language. This is because the indentation of Python code controls how the code is executed. Python does not have begin/end keywords or braces for grouping code statements. It simply does not need them. Take a look at the following Python fragment:

if x > y:

 print x

 if y > z:

  print y

  print z


  print z

The indentation of the code is used to control how statements are grouped for execution purposes. There can be no ambiguity as to which if clause is associated with the else clause in the above code because both statements have same level of indentation.

Functions in Python

We can turn the "Hello world" program into a Python function like this:

def Hello():

 print "Hello world"

Note that statements within the body of a function are indented beneath the def Hello() line which introduces the function. The parenthesis are a place holder for function parameters. Here is a function that prints its parameters x and y as well as the string "Hello world":

def Hello(x,y):

 print "Hello world",x,y

Python modules

A Python program typically consists of a number of modules. Any Python source file can serve as a module and be imported into another Python program. For example, assuming the Hello function above is housed in the file we can import the function into a Python program and call it as follows:

# Import the Hello function from the Greeting module

from Greeting import Hello

# Call the Hello function


Programs as modules to larger programs

Python makes it easy to write programs that can be used both as stand-alone programs and as modules to other programs.

Here is a modified version of which will print "Hello world" but can also still be imported into other programs:

def Hello():

 print "Hello world"

if __name__ == "__main__":

 # Test Hello Function if running as

 # main program


Note the special __name__ variable above. This variable is automatically set to "__main__" when a program is being executed directly. If it is being imported into another program, __name__ is set to the name of the module, which in this case would be "Greeting".

Python is object-oriented

Python is a very object-oriented language. Here is an extended version of the "Hello world" program, called, that can print any message via MessageHolder objects:

#Create a class called MessageHolder

class MessageHolder:

 # Constructor - called automatically

 # when an object of this class is created 

 def __init__(self,msg):

  self.msg = msg

 # Function to return the stored message string

 def getMsg(self):

  return self.msg

Note how indentation is used to structure the source code. the getMsg function is associated with objects of the MessageHolder class because it is indented beneath the class MessageHolder. Functions associated with objects are more generally known as methods.

Suppose now that I need a variation on the MessageHolder class in which all messages are returned in uppercase. I can do that by subclassing MessageHolder, specifying the class I wish to inherit from in parentheses after the class name:

# Import existing MessageHolder class from

from Message import MessageHolder

# Create a sub-class of MessageHolder called MessageUpper

class MessageUpper(MessageHolder):

 # Constructor

 def __init__(self,msg):

  # Call constructor of superclass


 # Over-ride getMsg with new

 # functionality

 def getMsg(self):

  return string.upper(self.msg)

Python is extensible

The Python language consists of a small core and a large collection of modules. Some of these modules are written in Python and some are written in C. As a user of Python modules, you cannot tell the difference. For example:

import xmlproc

import pyexpat

The first statement imports Lars Marius Garshol's implementation of an XML parser that is written purely in Python. The second statement imports the Python wrapping of James Clark's expat XML parser which is written in C.

Python programs using these modules cannot tell what language they have been implemented in. As you would expect, programs based on expat are typically faster owing to the speed advantages of a pure C implementation of an XML parser.

It is remarkably easy to write a Python module in C. This facility is very useful for speed-critical parts of large Python systems. It is also easy to "wrap" existing C libraries as Python modules, as has been done with expat. Many technologies exposing a C API have been wrapped as Python modules, for example Oracle, the Win32 API, and the wxWindows GUI toolkit, to name a few.

XML programming support

The core Python distribution (currently at version 1.5.2) has a simple non-validating XML parser module called xmllib. The vast bulk of Python's XML support is in the form of an add-on module under active development by the SIG for XML Processing in Python (known as XML-SIG). To illustrate Python's XML support, we will switch to an XML 1.0 version of the "Hello world" program processing the following file:

<?xml version = "1.0"?>


Hello world



SAX is a simple API for XML, spearheaded by David Megginson and developed as a collaborative effort on the XML-dev mail list. The Python implementation was developed by Lars Marius Garshol.

A Python SAX application to count the words in Greeting.xml looks like this:

from xml.sax import saxexts, saxlib, saxutils

import string

# Create a class to handle document events

class docHandler(saxlib.DocumentHandler):

 # Start of document handler

 def startDocument(self):

  # Initialize storage for character data

  self.Storage = ""


 # end of document handler

 def endDocument(self):

  # Print approximate number of words

  # by counting the number of elements in

  # the list of words returned by the

  # string.split function

  print len(string.split(self.Storage))


 def characters(self,str,start,end):

  # Accumulate character data

  self.Storage = self.Storage + str[start:end]


# Create a parser

parser = saxexts.make_parser()

# Provide the parser with a document handler


# Parse the Greeting.xml file



The DOM is a W3C initiative to standardize an API to XML (and HTML) documents. Python has two DOM implementations. The one in the XML-SIG modules is the work of Andrew Kuchling and Stéfane Fermigier. The other is called 4DOM and is the work of Fourthought, who have also created XSLT and XPath implementations in Python.

Here is a sample DOM application to count the words in Greeting.xml:

from xml.dom import utils,core

import string

# Read an XML document into a DOM object

reader = utils.FileReader('Greeting.xml')

# Retrieve top level DOM document object

doc = reader.document

Storage = ""

# Walk over the nodes

for n in doc.documentElement.childNodes:

 if n.nodeType == core.TEXT_NODE:

  # Accumulate contents of text nodes

  Storage = Storage + n.nodeValue

print len(string.split(Storage))

Native Python APIs

As well as industry standard APIs, there is a native Python XML processing library known as Pyxie.

Pyxie is an open source XML processing library for Python which will be made publicly available in January 2000. Pyxie tries to make the best of Python's features to simplify XML processing.

Here is the word counting application developed using Pyxie:

from pyxie import *

# Load XML into tree structure

t = File2xTree("Greeting.xml")

Storage = ""

# Iterate over list of data nodes

for n in Data(t):

 Storage = Storage + t.Data

print len(string.split(Storage))

In conclusion

We have looked at some of the main features of Python in a high level way. Also, we have glimpsed at some of the XML processing facilities available. For further information on programming with Python, I suggest you start with

Future articles on will taker a closer look at implementing XML applications in Python.