Re: Seq. scan when using comparison operators, why?

On Mon, 2002-04-08 at 07:45, Martijn van Oosterhout wrote:
> "Properly" in your opinion. It's more likely that postgres has a better idea
> of which one is faster...

This seems to be the standard response to any message questioning the
query planner's strategy.

In my opinion, such a response is condescending, discourages user
feedback about postgres performance in real database applications, and
fundamentally misses the point.

The point is that postgres performance frequently sucks on queries that
should be fast.  Any technical explanation about how postgres knows more
than a particular user about its own guts doesn't really bear on that
issue.  People are trying to be helpful by submitting query plans that
look suspicious... but they aren't just EXPLAINing in order to poke
holes in the query planner as entertainment, they are EXPLAINing to try
to figure out why it takes so long to get results from a simple query
that another DBMS can do in a fraction of the time.

You seem to be trying to put an end to the discussion by saying
"postgres knows how to plan queries, so keep your opinions to yourself".
Why is this user even concerned about the way Postgres is executing the
query?  Because he has a reasonable expectation that the query should be
faster than sequential scan of all records, and it's not.  Unless you
can either explain why that expectation is not reasonable, or explain
why postgres fails to meet reasonable expectations, you're just sneering
at a user reporting a problem.  Not good public relations.

Thanks,
Bill Gribble