Search in Django is painful and existing libraries/apps are spotty. The original Djangosearch works well for exact what it’s built for, but extension is hard/painful. So my goals are:

• Enable searching on third-party apps without touching that app’s code.
• Revise the way searching is done to be friendlier to the app developer.
• Provide ample room to extend/override any part with relative ease and minimal changes.
• Provide a means to have multiple indexes or limit the indexes registered.
• Full working backends for all supported engines (listed in the docs).
• Backends:

• Solr
• Lucene
• Whoosh

• Good test coverage.
• Supply ample documentation for app developers.
• New features (where supported by the backend):

• Term Boost
• More Like This
• Faceting
• Stored (non-indexed) fields
• Highlighting

• (Lofty ambition kicks in) Acceptance into django.contrib for Django 1.2.

There’s been a fair amount of progress. The current codebase has the following things complete/nearly complete:

• Enable searching on third-party apps without touching that app’s code.
• Revise the way searching is done to be friendlier to the app developer.
• Provide ample room to extend/override any part with relative ease and minimal changes.
• Provide a means to have multiple indexes or limit the indexes registered.
• Decent test coverage of the core API.
• Some documentation.
• Refactored ModelIndexes.
• Backends:

• Solr
• Whoosh

• New features (where supported by the backend):

• Term Boost
• More Like This
• Faceting
• Stored (non-indexed) fields
• Highlighting

The Whoosh backend could use a little further love to complete its API. It also suffers from locking issues and query parsing problems. The long term solution may be writing a standalone Whoosh server (much like Solr to Lucene) that handles locking intelligently and speaks HTTP.

In addition, I’d like to implement a reasonably complete Lucene backend.

A 1.0 release is blocking predominantly on the Lucene/Whoosh backends. However, there are a number of features & fixes I’d like to work in before that time. As of writing, Haystack is stable enough for production use with the understanding that things may change through time.