Re: Optimising queries involving unions

Sam Mason <sam(at)samason(dot)me(dot)uk> writes:
> Here's a bad example:

> SELECT u.txt
> FROM smalltable t, (
> SELECT id, txt FROM largetable1
> UNION ALL
> SELECT id, txt FROM largetable2) u
> WHERE t.id = u.id
> AND t.foo = 'bar';

> I was hoping that "smalltable" would get moved up into the union,
> but it doesn't at the moment and the database does a LOT of extra
> work.

I'm afraid we're a long way away from being able to do that; the
parse/plan representation of UNION wasn't chosen with an eye to
being able to optimize it at all :-(. We can push restriction
clauses down into a union, but we can't do much with join clauses,
because they necessarily refer to tables that don't even exist
within the sub-query formed by the UNION.

It'd be nice to fix this someday, but don't hold your breath ...

regards, tom lane