Parsing XML sucks. This library provides a cleaner interface to get at the data in a Wordpress export XML file.

I'm using the built-in etree.ElementTree parser to parse the Wordpress XML file.

If you have a Wordpress export that breaks the parser I feel your pain. Try looking at the line that Expat is barfing on and manually fixing it.

from wp_export_parser import WPParser

with open('wp-export.xml') as export_file:
    parser = WPParser(export_file.read())
    print parser.get_domain() # outputs www.example.com
    for p in parser.get_post_and_pages():
        categories = p['categories']
        comments = p['comments']
        title = p['title']
        type = p['post_type'] #type can be 'page' or 'post'

wp_export_parser can extract the following features from a Wordpress export file:

• Posts
• Pages
• Comments (exposed as a generator returning dicts)
• Categories (exposed as list of strings)
• Postmeta (exposed as dict)

Wordpress export files often include shortcodes, which the Wordpress rendering engine replaces with HTML. Since you probably aren't going to want to reimplement Wordpress's shortcodes in your own blogging engine, I have ripped out the shortcode parsing regular expressions and provided implementations of the most commonly-used shortcodes inside wp_export_parser.

• [youtube]: wp_export_parser retrieves the correct embed code (using oEmbed) and replaces the shortcode transparently.
• [caption]: wp_export_parser attempts to generate the same HTML Wordpress will generate (and assumes UTF-8 encoding)

Feel free to fork and contribute more shortcode support with a pull request

• wp_export_parser attempts to emulate the same behavior Wordpress uses to add <p> and <br> tags. I did this by attempting a 1-to-1 translation of the giant regular expression Wordpress uses to render posts.

• wp_eport_parser will parse files iteratively so it should be able to handle really large exports. get_pages() returns a generator.
• wp_export_parser sometimes will return unicode strings for the blog contents.
• Tested with CPython 2.6 and 2.7, as well as Pypy.

Copyright (c) 2012-2013 Kevin McCarthy. Released under the terms of the MIT license.