> i'd bet that it's likely that something could be done with a
> subsidiary relation -- instead of
you're absolutely right, that would make the indexing work (with
postgres). but using arrays is a design decision (good or bad). and
it's a decision we make often (with another db product) when we have
tables with millions of records. performance starts to suffer when you
join one 8.3 million record table with another 1.5 million record table.
Esp. when you have many fields that need to be "arrays".
plus, as an aside, since postgres supports arrays it just kinda makes
sense that you can be able to index 'em too. on the surface it doesn't
sound too difficult to implement, but I don't understand the sources
enough to do it... yet.
what I really need is full text indexing. the contrib/fti package seems
good, but the queries you have to generate are complicated.
eric