Re: Wich hardware suits best for large full-text indexed

Teodor Sigaev wrote:

> What is distribution of records by codgrupousuario field?

There are lots of documents in common between diferent user groups 
(codgrupousuario = user group id), but each user groups can access a 
subset of documents based on the types of document and data sources 
(fontes) they are allowed to use.

> You can use multi-column index (with contrib/btree_gist):
> create index fti on qrydocumentos using gist (codgrupousuario, 
> conteudo_stem_ix);
> or partial index
> create index fti on qrydocumentos using gist (conteudo_stem_ix) where 
> codgrupousuario = 1;

I´d rather not create partial index (I would have to create a trigger 
that creates an index each time a row is inserted on codgrupousuario). 
But I'll try it to see how it performs.
The multi-column index seems a good idea.

> One more. Let you use ispell dictionary ( I suppose, for Portuguese 
> language, 
> http://fmg-www.cs.ucla.edu/geoff/ispell-dictionaries.html#Portuguese-dicts 
> )

I'll make some tests with ispell, thanks for the advices.

My Regards,

-- 
Diogo Biazus
diogo@ikono.com.br
http://www.ikono.com.br