Re: extraer parte decimal

> 2011/6/23 Álvaro Hernández <aht(at)nosys(dot)es>

> > CREATE FUNCTION random_rows(table_name text, n_rnd_rows int, n_rows bigint)
> > RETURNS SETOF RECORD AS $$
> > DECLARE
> > curs refcursor;
> > ret record;
> > rnd_row_pos int[];
> > BEGIN
> > FOR i IN 1..n_rnd_rows LOOP
> > rnd_row_pos[i] = floor(random() * n_rows);
> > END LOOP;
> >
> > -- Sort the array
> > SELECT INTO rnd_row_pos array_agg(i) FROM (SELECT * FROM
> > unnest(rnd_row_pos) i ORDER BY i) AS i;
> >
> > OPEN curs FOR EXECUTE 'SELECT * FROM ' || table_name; --
> > ORDER is not relevant
> >
> > MOVE ABSOLUTE rnd_row_pos[1] - 1 FROM curs;
> > FETCH NEXT FROM curs INTO ret;
> > RETURN NEXT ret;
> >
> > FOR i IN 2..n_rnd_rows LOOP
> > MOVE RELATIVE rnd_row_pos[i] - rnd_row_pos[i - 1] -
> > 1 FROM curs;
> > FETCH NEXT FROM curs INTO ret;
> > RETURN NEXT ret;
> > END LOOP;

El problema de esta idea es que cada MOVE y FETCH requieren leer todas
las tuplas intermedias para contarlas. Capturar una tupla que está
hacia el final de la tabla es muy costoso.

--
Álvaro Herrera <alvherre(at)alvh(dot)no-ip(dot)org>