Re: tsearch_core for inclusion - Mailing list pgsql-hackers

On Thu, 15 Mar 2007, Robert Treat wrote:

> On Thursday 15 March 2007 12:17, Teodor Sigaev wrote:
>> Last try there was a fight about syntax of introduced commands. And we
>> (Oleg and me) developed variant of patch with another syntax. We will not
>> change docs until agreement will be reached, current version
>> http://mira.sai.msu.su/~megera/pgsql/ftsdoc/
>>
>> Following demonstrates subset of FTS syntax using example from
>> http://mira.sai.msu.su/~megera/pgsql/ftsdoc/fts-complete-tut.html.
>>
>
> This is nice.
>
> <snip>
>> Comparing that syntaxes with current tsearch2 is placed at
>> http://mira.sai.msu.su/~megera/pgsql/ftsdoc/fts-syntax-compare.html
>>
>> So, which is syntax more attractive?
>
> Honestly I don't find any of these syntax's to be head and shoulders above the
> others, but would probably lean toward the original syntax, since it has some
> level of familiarity among the existing user base.
>
>> And is there some another objections?
>
> Most people whom I talk to about tsearch who want the syntax changed to make
> it easier want something akin to just "CREATE INDEX fti1 on t1(c1) USING
> FULLTEXT" and then be done with it.  This patch isn't going to give people
> that.

Since we use standard postgresql-ish CREATE INDEX command, I assume 
people want to skip creation of tsvector column ? How they could manage
complex document indexing, when document is a combination (with different weights)
of many text attributes from several tables, for example ? There are several
other issues with that approach, for example, we need to store positional
information somewhere for ranking information. It's awkward to parse document
every time to get this information. 
>
> I'm also concerned about the stability of the tsearch api in general wrt
> including it in core.  Currently the recommended upgrade practice is to
> dump/reload without tsearch, installing the new servers version of tsearch
> instead.  IMHO this is not an acceptable solution for core-included features.
> So is this actually going to be improved in a core tsearch?  system tables
> are not dumped by default, so that seems easier, until you consider that your
> custom tsearch install will then be lost on upgrade... oops!

This is exact reason why we want to include tsearch into core, it was discussed
several times.
    Regards,        Oleg
_____________________________________________________________
Oleg Bartunov, Research Scientist, Head of AstroNet (www.astronet.ru),
Sternberg Astronomical Institute, Moscow University, Russia
Internet: oleg@sai.msu.su, http://www.sai.msu.su/~megera/
phone: +007(495)939-16-83, +007(495)939-23-83