Re: unique constraint on 2 columns

[Jonathan Vanasco <postgres(at)2xlp(dot)com>]

On Apr 20, 2007, at 6:13 PM, Jeff Davis wrote:

> This is more correct structure, and yes, it would involve a join.

I know thats the 'more correct' way -- but I can't do the join ,  
which is why I posted about a 2 column unique index.
I tested with a join before posting - i have an already large table  
that is growing quickly.  in order to use the join and keep current  
performance I'd need to scale out in hardware - which is just not an  
option right now.  searching 100M records vs searching 100M records +  
a join is a huge difference.  when you try to do analytics, its just  
not appropriate in my situation.

> No, it does not mean you need to rewrite anything. Use a view; that's
> the great benefit you get from using a relational database like
> PostgreSQL.

i'd have to rewrite everything that reads from that table to use the  
view instead of the current query, and then worry about inserts.