Full text search - altering the default parser

Started by Richard Huxtonalmost 18 years ago1 messages
#1Richard Huxton
dev@archonet.com

The default parser doesn't allow commas in numbers (I can see why, I think).

SELECT ts_parse('default', '123,000');
ts_parse
----------
(22,123)
(12,",")
(22,000)

One option of course is to pre-process the text, but since we can
support custom parsers I thought I'd take a look at the code to teach it
some flexibility on numbers. I'm guessing this would be of interest to
anyone wanting to support European-style "," decimal indicators too.

My "C" is horribly rusty, so can I check I've got this right? Before I
start exploring compiler errors I've not seen for decades ;-)

The parser functions (prsd_xxx) are all defined in
backend/tsearch/wparser_def.c

The state machine is driven through the TParserStateActionItem
definitions on lines 644 - 1263. Changing one of these will change the
definition of the corresponding token-type.

To add a new token-type, I'd add it to the various lists line 30-194,
then add the relevant TParserStateActionItems.

Thanks

--
Richard Huxton
Archonet Ltd