Confirmed users
764
edits
No edit summary |
(→Status: Status update) |
||
| Line 17: | Line 17: | ||
TAKING OFF | TAKING OFF | ||
==== Week of 2010/03/15 ==== | |||
* Progress in hooking up the awesomebar to FTS for the purpose of performance testing. | |||
** Exact matches are quite fast, i.e., matching whole words. Finding the records in a places database with 34000 history entries that match a whole word is on the order of 10 ms. Noticably faster than the current awesomebar. | |||
** However, prefix searches, i.e., searching for words that start with a given string, can be quite slow. For common prefixes, they can be even slower than a full table scan. (See also [http://www.mail-archive.com/sqlite-users@sqlite.org/msg29729.html this mail].) Prefixing searching is of course the real-world case. | |||
** If prefix searches are really slower than full table scans, then this project is done. | |||
** So I'm currently trying ways of optimizing prefix searches. Partitioning the FTS tables and doing multiple queries is showing some promise. | |||
==== Week of 2010/03/08 ==== | |||
* Gecko has some facilities for i18n word boundary analysis under the [http://mxr.mozilla.org/mozilla-central/source/intl/lwbrk/ intl/lwbrk/] directory. | * Gecko has some facilities for i18n word boundary analysis under the [http://mxr.mozilla.org/mozilla-central/source/intl/lwbrk/ intl/lwbrk/] directory. | ||