Re: FTS performance with the Polish config

2009/11/15 Oleg Bartunov <oleg(at)sai(dot)msu(dot)su>:
> On Sun, 15 Nov 2009, Pavel Stehule wrote:
>
>> 2009/11/15 Oleg Bartunov <oleg(at)sai(dot)msu(dot)su>:
>>>
>>> Yes, as stated original author use polish ispell dictionary.
>>> Ispell dictionary is slow to load first time. In real life it should be
>>> no
>>> problem.
>>>
>>
>> it is a problem. People who needs fast access uses english without
>> czech. It drop some features, but it is significaly faster.
>
> just don't use ispell dictionary, czech snowball stemmer is as fast as
> english.

czech stemmer doesn't exist :(

>
> Ispell dictionary (doesn't matter english, or other language) is slow for
> the first load and then it caches, so there is no problem if use persistent
> database connection, which is de facto standard for any serious projects.
>

I agree so connection pooling should be a solution. But it is good?
Cannot we share dictionary better?

>>
>> Pavel
>>
>>> Oleg
>>> On Sat, 14 Nov 2009, Pavel Stehule wrote:
>>>
>>>> 2009/11/14 Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>:
>>>>>
>>>>> Kenneth Marshall <ktm(at)rice(dot)edu> writes:
>>>>>>
>>>>>> On Sat, Nov 14, 2009 at 12:25:05PM +0100, Wojciech Knapik wrote:
>>>>>>>
>>>>>>> I just finished implementing a "search engine" for my site and found
>>>>>>> ts_headline extremely slow when used with a Polish tsearch
>>>>>>> configuratio=
>>>>
>>>> n,
>>>>>>>
>>>>>>> while fast with English.
>>>>>
>>>>>> The documentation for ts_headline() states:
>>>>>> ts_headline uses the original document, not a tsvector summary, so it
>>>>>> can be slow and should be used with care.
>>>>>
>>>>> That's true but the argument in the docs would apply just as well to
>>>>> english or any other config. =C2=A0So while Wojciech would be well
>>>>> advised
>>>>> to try to avoid making a lot of calls to ts_headline, it's still
>>>>> curious
>>>>> that it's so much slower in polish than english. =C2=A0Could we see a
>>>>> self-contained test case?
>>>>
>>>> is it dictionary based or stem based?
>>>>
>>>> Dictionary based FTS is very slow (first load). Minimally czech FTS is
>>>> slow.
>>>>
>>>> regards
>>>> Pavel Stehule
>>>>
>>>>>
>>>>> =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0
>>>>> =C2=
>>>>
>>>> =A0 =C2=A0regards, tom lane
>>>>>
>>>>> --
>>>>> Sent via pgsql-performance mailing list
>>>>> (pgsql-performance(at)postgresql(dot)org)
>>>>> To make changes to your subscription:
>>>>> http://www.postgresql.org/mailpref/pgsql-performance
>>>>>
>>>>
>>>> --=20
>>>> Sent via pgsql-performance mailing list
>>>> (pgsql-performance(at)postgresql(dot)org)
>>>> To make changes to your subscription:
>>>> http://www.postgresql.org/mailpref/pgsql-performance
>>>>
>>>
>>>        Regards,
>>>                Oleg
>>> _____________________________________________________________
>>> Oleg Bartunov, Research Scientist, Head of AstroNet (www.astronet.ru),
>>> Sternberg Astronomical Institute, Moscow University, Russia
>>> Internet: oleg(at)sai(dot)msu(dot)su, http://www.sai.msu.su/~megera/
>>> phone: +007(495)939-16-83, +007(495)939-23-83
>>>
>>
>
>        Regards,
>                Oleg
> _____________________________________________________________
> Oleg Bartunov, Research Scientist, Head of AstroNet (www.astronet.ru),
> Sternberg Astronomical Institute, Moscow University, Russia
> Internet: oleg(at)sai(dot)msu(dot)su, http://www.sai.msu.su/~megera/
> phone: +007(495)939-16-83, +007(495)939-23-83