Last modified: 2014-05-28 19:31:56 UTC
If you parse wikitext like [[File:Ulmer M?nster-1024409.jpeg|frame]] , Parsoid will produce a <figure> structure with href="./File:Ulmer_M%3Fnster-1024409.jpeg" (which is correct, the question mark is urlencoded) and resource="./File:Ulmer_M%3Fnster-1024409.jpeg" (not sure whether this is correct). When VisualEditor takes this information to build an inline image equivalent (taking the href and the resource, and dropping everything else including data-parsoid) we get [[File:Ulmer_M%3Fnster-1024409.jpeg|link=]] back. Note the space changed to an underscore and ? changed to %3F. There was information in data-parsoid to keep this from happening, but in this transformation VE discards it. $ echo '[[File:Ulmer M?nster-1024409.jpeg|frame]]' | node tests/parse.js --apiURL=http://localhost/w/api.php <!DOCTYPE html> <html prefix="dc: http://purl.org/dc/terms/ mw: http://mediawiki.org/rdf/"><head prefix="mwr: http://localhost/wiki/Special:Redirect/"><meta property="mw:parsoidVersion" content="0"/><link rel="dc:isVersionOf" href="//localhost/wiki/Main_Page"/><title></title><base href="//localhost/wiki/Main_Page"/></head><body data-parsoid='{"dsr":[0,42,0,0]}'><figure class="mw-default-size" typeof="mw:Image/Frame" data-parsoid='{"optList":[{"ck":"framed","ak":"frame"}],"dsr":[0,41,2,2]}'><a href="./File:Ulmer_M%3Fnster-1024409.jpeg" data-parsoid='{"a":{"href":"./File:Ulmer_M%3Fnster-1024409.jpeg"},"sa":{},"dsr":[2,39,null,null]}'><img resource="./File:Ulmer_M%3Fnster-1024409.jpeg" src="//localhost/w/images/f/f1/Ulmer_M%3Fnster-1024409.jpeg" height="1024" width="764" data-parsoid='{"a":{"resource":"./File:Ulmer_M%3Fnster-1024409.jpeg","height":"1024","width":"764"},"sa":{"resource":"File:Ulmer M?nster-1024409.jpeg"}}'/></a></figure> </body></html> $ echo '<p><span typeof="mw:Image" class="mw-default-size"><span><img src="//localhost/w/images/f/f1/Ulmer_M%3Fnster-1024409.jpeg" resource="./File:Ulmer_M%3Fnster-1024409.jpeg" width="764" height="1024"></span></span></p>' | node tests/parse.js --html2wt [[File:Ulmer_M%3Fnster-1024409.jpeg|link=]]
It looks like we aren't percent-decoding & underscore-removing the resource properly in the serializer. This also seems to be the case for block images.
