Re: Searching GIN-index (FTS) and sort by timestamp-column

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: Andreas Joseph Krogh <andreas(at)visena(dot)com>
Cc: Evgeniy Shishkin <itparanoia(at)gmail(dot)com>, pgsql-performance <pgsql-performance(at)postgresql(dot)org>
Subject: Re: Searching GIN-index (FTS) and sort by timestamp-column
Date: 2016-03-17 17:57:46
Message-ID: 4024.1458237466@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

Andreas Joseph Krogh <andreas(at)visena(dot)com> writes:
> P torsdag 17. mars 2016 kl. 18:20:23, skrev Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us
> FWIW, the reason for that is that the int8_ops operator class that
> btree_gin creates doesn't contain any cross-type operators. Probably
> wouldn't be that hard to fix if somebody wanted to put in the work.

> Can you explain why it works when using prepared statement without casting?

If you mean the example

prepare fish AS SELECT del.id , del.received_timestamp FROM delivery del
WHERE 1= 1 AND del.fts_all @@ to_tsquery('simple', $1) AND
del.folder_id = $2 ORDER BY del.received_timestamp DESC LIMIT 101 OFFSET 0;

you didn't provide any type for the parameter $2, so the parser had to
infer a type, and the applicable heuristic here is "same type that's on
the other side of the operator". So you ended up with "bigint = bigint"
which is in the btree_gin operator class. If you'd specified the
parameter's type as integer, it would've worked the same as Evgeniy's
example.

regards, tom lane

In response to

Browse pgsql-performance by date

  From Date Subject
Next Message Dave Stibrany 2016-03-17 20:45:10 Disk Benchmarking Question
Previous Message Andreas Joseph Krogh 2016-03-17 17:30:12 Re: Searching GIN-index (FTS) and sort by timestamp-column