Finally I create a function like:
CREATE OR REPLACE FUNCTION nonsensible (text) RETURNS text AS $$
DECLARE
var1 varchar;
BEGIN
var1=replace($1, 'á', 'a');
var1=replace(var1, 'é', 'e');
var1=replace(var1, 'í', 'i');
var1=replace(var1, 'ó', 'o');
var1=replace(var1, 'ú', 'u');
var1=replace(var1, 'Á', 'A');
var1=replace(var1, 'É', 'E');
var1=replace(var1, 'Í', 'I');
var1=replace(var1, 'Ó', 'O');
var1=replace(var1, 'Ú', 'U');
return var1;
END
$$LANGUAGE plpgsql immutable;
Then, create text indexes, one for sensible queries and other for
unsensible ones:
CREATE INDEX textindex ON document USING
gin(to_tsvector('spanish',text));
CREATE INDEX textindexn ON document USING
gin(to_tsvector('spanish',nonsensible(text)));
And then make a query sensible or unsensible to accents doing:
SELECT id FROM document WHERE to_tsvector('spanish',text) @@
to_tsquery('spanish','word_with_accent');
or:
SELECT id FROM document WHERE to_tsvector('spanish',nonsensible(text))
@@ to_tsquery('spanish',nonsensible('word_with_accent'));
respectively.
I think postgreSQL uses both indexes as necessary. I believe to remember
reading something about it in the documentation.
Thank you very much,
Mario Barcala
> Here is an example
>
> CREATE FUNCTION dropatsymbol(text) RETURNS text
> AS 'select replace($1, ''@'', '' '');'
> LANGUAGE SQL;
>
> arxiv=# select to_tsvector('english',dropatsymbol('oleg@sai.msu.su'));
> to_tsvector
> -------------------------
> 'oleg':1 'sai.msu.su':2
