Question

Note: This question originally applied to Xapian, but due to cross-platform issues and poor understanding of Xapian I (our team) chose Solr instead.

I’m looking for snippets, tricks, tips, links, and anything to watch out for (gotchas). My technology stack includes:

• MySQL 5.1 (Not really pertinent)
• Red Hat and Windows configurations with final deployment to Linux
• Development primarily done on windows machines on my team
• No PHP or Java support in our configurations, ergo no Solr or Django-Sphinx Went with Java after all!

Thank you all for the help and insight!

Answer 1

A few notes and resources. My advice is mostly related to Haystack in general since I don’t have experience with Xapian as a backend.

1. Installing Xapian (from the Haystack
   docs) – note that Haystack doesn’t
   support Xapian on its own:
   http://haystacksearch.org/docs/installing_search_engines.html#xapian
2. It may be helpful to use Whoosh
   during development or for testing
   certain things, but keep in mind
   that it doesn’t support all the
   features Xapian does. Haystack does
   a good job of failing gracefully (a
   warning in your console) if you try
   to use Whoosh with a feature it
   doesn’t support, so switching between
   them is painless:
   http://haystacksearch.org/docs/installing_search_engines.html#whoosh

3. A snippet from my own code of
   switching between Whoosh and Solr
   easily:

   # Haystack search settings
   HAYSTACK_SITECONF = 'project.search_sites'
   HAYSTACK_INCLUDE_SPELLING = True
   # Haystack backend settings
   HAYSTACK_SEARCH_ENGINE = 'solr' # Switch this to 'whoosh' to use that backend instead
   if DEBUG:
       HAYSTACK_SOLR_URL = 'solr.development.url'
   else:
       HAYSTACK_SOLR_URL = 'solr.production.url'
   HAYSTACK_WHOOSH_PATH = os.path.join(PROJECT_ROOT, 'search_index', 'whoosh')
   

4. As far as I’m aware your choice of
   database doesn’t make a difference
   as long as Django supports it since Haystack uses the ORM.
5. If you run into any trouble,
   Haystack’s developer (Daniel
   Lindsley) is incredibly helpful and
   quick to respond. You can get help
   from him and others in the
   django-haystack Google group or
   the #haystack IRC channel (that is,
   if you don’t find an answer in the
   official docs).