Markus Schiltknecht <markus(at)bluegap(dot)ch> writes:
> UNIQUE (t1, lower(t2)));
> I can easily create an index for my needs [1], why can I not add such a
> unique constraint? Thanks for clarification.
Because the SQL spec defines this syntax, and it only allows column
names there.
Extending the spec in this particular direction is not as easy as it
might look, either. What will you do with the information_schema
description of the unique constraint?
regards, tom lane