PDF generation tests, March 2008

I've got copies of a few Wikipedia articles on my workstation's local test wiki, so decided to throw some at it and see....

Test set

This is not an exhaustive test of features; I used a couple articles I had copies of from previous bug testing, and today's featured article, just to do an initial tryout.

I took copies of the following articles from English Wikipedia, with all their templates and images:

• en:Borbera - A short article with an infobox and a non-printing stub template
• en:Flag of Armenia - Today's Featured Article; has lots of floating SVG images and references, some infoboxes.
• en:Experimental Assembly of Structures in EVA and Assembly Concept for Construction of Erectable Space Structures - Has a really long title, some awkwardly-sized floating images, and uses unit conversion templates and references extensively

Test procedure

From my test wiki, running MediaWiki current trunk, lots of extensions, plus PediaPress's Collation extension and mwlib / mwlib.rl Python utilities (with ReportLab installed from SVN):

1. Printed each article to PDF from Firefox 3.0b4 on my iMac
2. Did an individual 'Download as PDF' via mwlib for each page
3. Added all three articles to a collection and downloaded it as a combined PDF

I'm only considering the PDF output here, not the generation procedure, server issues, or UI.

Results

• en:Borbera
  • Media:Borbera (Firefox).pdf
  • Media:Borbera (mwlib).pdf
• en:Flag of Armenia
  • Media:Flag of Armenia (Firefox).pdf
  • Media:Flag of Armenia (mwlib).pdf
• en:Experimental Assembly of Structures in EVA and Assembly Concept for Construction of Erectable Space Structures
  • Media:Assembly (Firefox).pdf
  • Media:Assembly (mwlib).pdf
• Collection
  • Media:Collection test via mwlib.pdf

Notes

General issues

• mwlib does some nicer formatting in places; spacing and font sizing seems a little cleaner
• TOC listings are not shown in mwlib
• Licensing info also seems missing, but that may be a local config issue

Borbera

• The infobox is expanded to full page length, and loses formatting of the heading
• An inline interwiki link vanishes...
• Inline URL links don't list the URLs

Flag of Armenia

• We see the expanded infobox issue again.
• Inline URL links don't list the URLs
• A block quotation is shrunken very small.
• Some mystery extra page breaks occur with both Firefox and mwlib versions.
• Most worrisome is that the article simply cuts off about halfway through, at 'Early Soviet Armenia and the Transcaucasian SFSR'...

Experimental Assembly ....

• The very long title doesn't wrap or trim cleanly in the page header
• Inline URL links don't list the URLs
• A reference entry is missing
• Some mystery extra page breaks occur with both Firefox and mwlib versions.
• Images seem to be laid out nicer in mwlib than in Firefox, which is nice! :)
• Unit conversion templates fail to render correctly in mwlib, leaving text like "thirty-three

4.5 ft ( m) struts"

Collection

• Some of the page breaks seem to be different from the individual downloads