Re: How to speed up pg_trgm / gin index scan

On 28/07/15 16:42, Merlin Moncure wrote:
>>
>> Great stuff! Sorry Oleg I don't have your original message anymore and
>> can't reply into the right place in the thread, so I took the liberty to
>> CC: you.
>
> There are some more big optimizations (via Jeff Janes) coming down the
> pike for trigram searching. See thread:
> http://www.postgresql.org/message-id/CAMkU=1woR_Pdmie6d-zj6sDOPiHd_iUe3vZSXFGe_i4-AQYsJQ@mail.gmail.com.
>
> I think it should be possible to patch the 9.4 pg_trgm module with
> Jeff's stuff -- it might be worthwhile to do that and run some tests
> and report back. I don't know if they address your particular case
> but in some situations the speedups are really dramatic.
>
> merlin
>

On 28/07/15 16:45, Arthur Silva wrote:>
> You could experiment recompiling pg_trgm commenting out the
> KEEPONLYALNUM and/or IGNORECASE definitions if you are looking for exact
> matches, this will increase the index size but will make it more
selective.
>
> Also, there's a thread around for pg_trgrm 1.2 which will get you even
> more boost.

Thanks for the hints and all the hard work you guys are putting into
this. I'll follow the further development closely and report back if we
get any new breakthroughs with this rather big data set.

Christian