ChangeLog
v0.3
- public release
- extensive debug, all known bugs solved
v0.3-beta1
- parameters consistency checked
- exact match start working
- AND words are now debug (or so I hope)
- v0.2 indexing is removed by default, can be reactivated via a define
- documentation checking
v0.3-alpha1
-
index is stored into 4 sqlite3 files instead of one
- motivations: this is due to locking limitation inside
sqlite3. The recommendation is to use multiple files.
- incremental indexation is in place
- when you modify your ft3 indexed datas, a journal is
generated. After the indexation is finished, results are immediatly
visibles.
- you later needs to merge incremental index and the full index
- deferred or immediate indexing is possible
- indexation can be time (CPU) consumming especially for large
documents. If you are storing books in the database, you will be
intrested by this features.
- you are responsible for scheduling indexation process. it can
be immediate, or batched every 10 minutes for example for a
better performance
- a http+xml server is in place
- Current XML is describe by a schema. I think I will move to
opensearch format. Any ideas?
- stopwords works again
- after language detection as been done, a per language stopword
lists can be used to restrict index size.
- stemming correlated to current language
- wikipedia and dmoz demos in PHP5 are available in the demo
directory
- The Dmoz demo has been used to index the 2.3GB of XML text
dumped by the dmoz.org project.
- The Wikimedia demo shows that ft3 can index the english part
of Wikimedia quite well on a medium end server (currently done
on a PIV 2Ghz with 1GB or RAM and 4 SATA disks.
v0.2
- first published version
- one shot indexer
- fast, stable