Editing API:EPrints/Plugin/Search/Xapian (section)

==USAGE==
Install the [http://search.cpan.org/search?query=xapian&mode=dist Search::Xapian] extension. Note: there are two Perl bindings available for Xapian. The CPAN version is older and based on Perl-XS. xapian-bindings-perl available from xapian.org is based on SWIG and has better coverage of the API. Regardless, for the best feature support/performance it is highly recommended to have the latest stable version of the Xapian library.

Xapian uses a separate (from MySQL) index that is stored in <tt>archives/''[archiveid]''/var/xapian</tt>. To build the Xapian index you will need to reindex:

<pre>  ./bin/epadmin reindex [archiveid] eprint</pre>

(Repeat for any other datasets you expect to use Xapian with.)

The <tt>var/xapian/</tt> directory should contain something like:

<pre>  flintlock  position.baseA  position.DB     postlist.baseB  
       record.baseA  record.DB       termlist.baseB
  iamchert   position.baseB  postlist.baseA  postlist.DB     
       record.baseB  termlist.baseA  termlist.DB</pre>

The indexing process for Xapian is in <tt>lib/cfg.d/search_xapian.pl</tt>. This can be overridden by dropping the same-named file into your repository <tt>archives/''[archiveid]''/cfg.d/</tt>. If the Xapian search is not matching what you might expect it to, you probably need to fix the indexing process (and re-index!). Terms indexed by Xapian can also be weighted to e.g. give names a higher weighting than abstract text.

You will need to restart your Apache server to enable the Xapian search plugin and dependencies.

If the Xapian search is working correctly you will have a "by relevance" option available in the ordering of simple search results.

<!-- Edit below this comment -->


<!-- Pod2Wiki= -->
<!-- Pod2Wiki=head_lock_files -->
===Lock Files===
Xapian maintains a lock file in <tt>archives/''[archiveid]''/var/xapian</tt>. If you see indexing errors about not being able to lock the database ensure you aren't running multiple copies of the EPrints [http://wiki.eprints.org/w/API:bin/indexer indexer]. If no other processes are running you may need to manually remove the lock file from the <tt>var/xapian</tt> directory. While only one process may modify the Xapian index at a time, any number of processes may concurrently read.

<!-- Edit below this comment -->


<!-- Pod2Wiki= -->
<!-- Pod2Wiki=head_parameters -->