Word2Mediawiki

This project contains some "fairly simple" scripts to deal with automated best-effort conversion from Word (or OpenOffice/LibreOffice) to Mediawiki. Each file converted will create a page on the wiki, where all the textual content will go, and also automatically upload all related images to the wiki as attachments.

Copyright

Author: Magnus Hagander <[email protected]>
Based on work from Lars Strand <[email protected]>
Work commissioned by Oslo Lufthavn AS, 2011 [email protected]

(C) 2011

Licensed under the GNU LGPL v2.1 - http://www.gnu.org/licenses/lgpl-2.1.html
- or any later version.

Credits

This project is based off work by Lars Strand, cleaned up and re-implemented in a single language, with proper error checking etc. For document conversion, it uses DocumentConverter.py by Mirko Nasato. Finally, the uploading to mediawiki is made using the Python Wikipedia Robot Framework.

Requirements

The system has so far been tested only on Ubuntu, but should work on any system that can provide the required software. The following Ubuntu packages are required (on top of the base system, and of course including any dependencies they required):

• python-magic
• python-uno
• openoffice.org-java-common
• openoffice.org-wiki-publisher
• openoffice.org-writer

Other than this, a checkout of http://svn.wikimedia.org/svnroot/pywikipedia/trunk/pywikipedia needs to be made into the subdirectory pywikipedia.

Operations

The basic script is word2mediawiki.py, and is executed once per file, specifying the file name on the command line. The script performs several actions in a flow:

1. The file is converted, using functions from DocumentConverter.py, to both XML and wiki markup, in a temporary working directory.
2. The returned XML file is then parsed for all image contents, since the wiki file does not include them. The images are stored in the XML file in base64 encoded format. The images are given sequential numbers that are later used for references and file names.
3. The returned wiki file is then parsed, and all instances of images (which at this point have no names) are replaced with explicit references to the image names generated in the previous step. The wiki text is also assigned to the category word2mediawiki, for easier tracking.
4. A login attempt is made to the wiki (this is an uncertain step, it sometimes returns successful login even when it doesn't work, but it's a necessary step).
5. The wiki is checked for presence of a document with this name. It checks both for a page with this name, and for uploaded files conflicting with the names of our generated images. If a file exists, a simple suffix counter is added to the name and it's tried again, to attempt to get around the naming conflict.
6. Once we have all this data, the page is uploaded. First the page itself, then all the images, and finally a copy of the original word document for reference.

Setting up

The first step of setting up, is to create a family for pywikipedia to use for logging in. This is done by editing the file pywikipedia/families/test_family.py (replace test with whatever is reasonable in your setup, for example osl). This file should contain something like

# -*- coding: utf-8  -*-
import family

class Family(family.Family):
      def __init__(self):
          family.Family.__init__(self)
          self.name = 'test' #Set the family name; this should be the same as in the file name.

          self.langs = {
               'en': None,
          }

      def hostname(self, code):
          return 'wiki.somewhere.com'

      def version(self, code):
          return "1.12.0"

      def scriptpath(self, code):
          return '/mediawiki'

(Obviously changed to suit the environment).

When this file is in place, run the login script from pywikipedia once (seems to sometimes be necessary, sometimes not, so just do it..)

cd pywikipedia
python login.py -force

In the order of the questions returned, enter:

1. The id of the new family you created earlier
2. The language (normally just pick the default)
3. The user name (must be entered, don't use the default)
4. Small configuration
5. Your password

Once this is done, update the word2mediawiki.ini file with family, password and language.

Cron job

If you want the system to run automatically, just put the convert_cron.sh script in cron at a suitable interval. You may need to edit this script to change paths etc.

When running this from cron, make sure that the user who's running it (which should never be root, of course) has permissions on at least the following directories (rooted in the installation directory):

• log
• share
• tmp
• pywikipedia/logs
• pywikipedia/login-data