Thanks for the rapid response.
I can understand the way the planner makes its guess, but as a matter of
fact, he'll be nearly always wrong, just becausethe most commented articles
have only around 5000 or so comments. I ran the explain analyze tonight
and got this results :
EXPLAIN ANALYZE SELE
On Tue, Jan 20, 2009 at 10:45 AM, Yannick Le Guédart
wrote:
>
> The second query scans the whole comment table which is very dangerous for
> production servers.
That's not quite true. The second does an index scan- the planner
seems to be guessing that it'll fulfill the required limit early in
t
Greetings,
I'm experiencing a strange query plan change on a "simple" request, based on
the LIMIT parameter. I present here two tables, named _article and _comment.
they are part of a much larger database.
It's a tree-like database (making use of the ltree to keep the coherence),
and in our case,
