The u32 page is excellent, but I wonder if documenting the procedure here
would be worthwhile. Who knows, the remote site might disappear. But also
there are some variations on the approach that might be useful:

 * using a python script and the dom library to extract the page names from
   Special:Allpages (such as
   <http://www.staff.ncl.ac.uk/jon.dowland/unix/docs/get_pagenames.py>)
 * Or, querying the mysql back-end to get the names
 * using WWW::MediaWiki for importing/exporting pages from the wiki, instead
   of Special::Export

Also, some detail on converting mediawiki transclusion to ikiwiki inlines...

-- [[users/Jon]]

> "Who knows, the remote site might disappear.". Right now, it appears to
> have done just that. -- [[users/Jon]]


The iki-fast-load ruby script from the u32 page is given below:

        #!/usr/bin/env ruby

        # This script is called on the final sorted, de-spammed revision
        # XML file.
        #
        # It doesn't currently check for no-op revisions...  I believe
        # that git-fast-load will dutifully load them even though nothing
        # happened.  I don't care to solve this by adding a file cache
        # to this script.  You can run iki-diff-next.rb to highlight any
        # empty revisions that need to be removed.
        #
        # This turns each node into an equivalent file.
        #    It does not convert spaces to underscores in file names.
        #       This would break wikilinks.
        #       I suppose you could fix this with mod_speling or mod_rewrite.
        #
        # It replaces nodes in the Image: namespace with the files themselves.


        require 'rubygems'
        require 'node-callback'
        require 'time'
        require 'ostruct'


        # pipe is the stream to receive the git-fast-import commands
        # putfrom is true if this branch has existing commits on it, false if not.
        def format_git_commit(pipe, f)
           # Need to escape backslashes and double-quotes for git?
           # No, git breaks when I do this. 
           # For the filename "path with \\", git sez: bad default revision 'HEAD'
           # filename = '"' + filename.gsub('\\', '\\\\\\\\').gsub('"', '\\"') + '"'

           # In the calls below, length must be the size in bytes!!
           # TODO: I haven't figured out how this works in the land of UTF8 and Ruby 1.9.
           pipe.puts "commit #{f.branch}"
           pipe.puts "committer #{f.username} <#{f.email}> #{f.timestamp.rfc2822}"
           pipe.puts "data #{f.message.length}\n#{f.message}\n"
           pipe.puts "from #{f.branch}^0" if f.putfrom
           pipe.puts "M 644 inline #{f.filename}"
           pipe.puts "data #{f.content.length}\n#{f.content}\n"
           pipe.puts
        end