ZF-1565: Best practice section for Zend_Search_Lucene module

Issue Type:
Docs: Task
Created:
2007-06-14T06:58:16.000+0000
Last Updated:
2008-03-21T16:25:15.000+0000
Status:
Resolved
Fix version(s):
  • 1.5.0 (17/Mar/08)
Reporter:
Alexander Veremyev (alexander)
Assignee:
Alexander Veremyev (alexander)
Tags:
  • Zend_Search_Lucene
Related issues:
Attachments:

Description

Some 'best practice' recommendations may already be added to Zend_Search_Lucene module documentation.

  1. Don't use 'id' for stored field names.

  2. Some recommendations for MaxBufferedDocs, MaxMergeDocs and MergeFactor settings for different usage modes (indexing, searching, batch indexing, indexing small/large documents).

  3. Shut down index before the end of script (unset index object), so auto-commit will be done before exception mechanism is shut down.

Comments

Posted by Alexander Veremyev (alexander) on 2007-06-14T15:28:52.000+0000

  1. Do not use index on NFS (see flock() documentation).

Posted by Alexander Veremyev (alexander) on 2007-06-19T16:45:59.000+0000

  1. If documents have any unique-id field, then documents should be retrieved through $index->termDocs() method instead of $index->find()

  2. If index is not optimized than each segment has his own term dictionary and preloaded term dictionary index. It also increases search time (each search is actually several mini-searches through segments)

Posted by Alexander Veremyev (alexander) on 2007-06-25T12:15:31.000+0000

  1. Encoding parameters usage.

Posted by Alexander Veremyev (alexander) on 2007-06-28T11:43:52.000+0000

  1. Index maintenance (backup and restoring).

Posted by Alexander Veremyev (alexander) on 2007-07-03T10:15:29.000+0000

Done.