Page MenuHomePhabricator

Add uploaded file text and metadata from files to fulltext search set
Closed, ResolvedPublic


We're starting to integrate text extraction for djvu and pdf files -- currently used for ProofreadPage extension -- but it's not currently exposed to the search indexing.

This is also something frequently desired for text document types like .doc and .odf, and there are some extensions for doing that but there's not a clean interface to plug it in to that can be supported for all search backends.

Note that supporting the Lucene search which updates separately might require some additional attention.

Related bugs:

  • bug 6421 - search djvu file text
  • bug 6422 - search pdf file text
  • bug 13370 - search file metadata

Also interesting idea:

  • bug 18045 - search text of linked files (but if these are remote, that's much harder to handle!)

Things we need:

  • clear interface on File for things that need to be fetched (exif metadata, page text)
  • clear interface on the SearchEngine class for plugging additional info in to updates
  • a way to expose additional searchable info to the Lucene search's updaters (plugin to oai interface maybe to toss in extra data fields?)

Version: unspecified
Severity: enhancement



Event Timeline

bzimport raised the priority of this task from to Low.Nov 21 2014, 10:51 PM
bzimport added a project: MediaWiki-Search.
bzimport set Reference to bz21061.
bzimport added a subscriber: Unknown Object (MLST).

test5555 wrote:


*Bug 21795 "camera categories" (proposal c would allow searching metadata through categories they generate)

dr.trigon wrote:

bug 6421 could finally be closed - thanks to everybody involved there!

Updates to [search file metadata] could possibly be done with the Structured Data on Commons work, we already have some, but probably more would could be done. We might need more information first though, to really get a sense of what is wanted here. (It's an old ticket.)

Deskana claimed this task.
Deskana added subscribers: EBernhardson, Deskana.

This task talks about adding indexing of content of djvu and pdf files, which, according to @EBernhardson, is now done for those file types. Accordingly, this is resolved. For more complex metadata, waiting for Structured Data on Commons as @debt suggests is the best course of action.