Markus Schaber <schabi(at)logix-tt(dot)com> writes:
> Don't forget that, in the current implementation, the query planner has
> no choice but planning the query without any actual parameter values,
> which is likely to give equal or worse results in most cases.
No, I think you are oversimplifying. What the planner is supposed to do
when given a parameterized query (and no parameter value info) is to
generate a plan that won't be too terribly awful regardless of the
parameter values. With a set of sample values, it may generate a plan
that is great for those values and utterly unusable for anything else.
Here's a simple example:
select ... from a join b on (a.key = b.key) where a.val = ?
Given a parameter value that looks like it will match only one A row,
you will likely get a plan like this:
Index Scan on A
Index Cond: val = ?
Index Scan on B
Index Cond: b.key = a.key
which is about as fast as you can get if indeed there's only one match.
However if the parameter value matches very many A rows, this plan is
horrid. With no value for the parameter, the planner should pick a
compromise plan (perhaps a mergejoin) that may not be as fast for the
single-match case, but will finish before doomsday in the other case.
regards, tom lane
In response to
pgsql-jdbc by date
|Next:||From: Markus Schaber||Date: 2006-09-26 13:20:03|
|Subject: Re: Bind message|
|Previous:||From: till toenges||Date: 2006-09-26 13:07:12|
|Subject: Re: bytea memory improvement|