On 9/22/07, Guillaume Smet <guillaume.smet@gmail.com> wrote:
> Anyway, I'm starting with a clean tree containing your fix and what
> Tom commited but I have to import the data again due to the catalog
> version bump :).
I have some good news. After Andrew's and Greg's patches, CVS HEAD is
as fast as 8.2 with latin1 encoding:
cityvox_latin1=# SELECT e.numeve FROM evenement e WHERE e.libgeseve
LIKE '%hocus pocus%';numeve
--------
(0 rows)
Time: 102.731 ms
cityvox_latin1=# SELECT e.numeve FROM evenement e WHERE e.libgeseve
ILIKE '%hocus pocus%'; numeve
-----------900024298 87578
(2 rows)
Time: 120.399 ms
So the only regression left is that from 8.2, ILIKE with UTF-8
encoding is really slower than before but it doesn't seem easy to
solve (if possible).
Regards,
--
Guillaume