Tomasz,
> I think if you really need to remove your 0.5 values, you need another
> table with all possible values of id1. This table will look like:
> create table t3
> (
> id1 integer;
> );
T1 contains values of id's,
The T2 matrix is a matrix of some sort of "distances" between id's
from T1, so original T2 is a T1 x T1.
(that's what you are proposing right?)
So there is no id1, id2 in T2 that don't exists in T1,
I checked my test tables and this is indeed the case.
In other words T1 is dense, I make a left join on it and I get wrong results!
And even wrong number of tuples :-(
cheers,
Mila