Steve Rogerson <steve.pg@yewtc.demon.co.uk> writes:
> On 05/01/16 19:47, Tom Lane wrote:
>> That's operating as designed. A unique constraint needs an index,
>> but not vice versa.
> I can see that might be plausible , hence the question but as a "unique index"
> imposes as constraint they seem equivalent. What's the functional difference
> between the two situations?
There is none so far as uniqueness-enforcement is concerned, because the
index is the same either way, and that's what enforces it.
The main reason we don't automatically create a constraint for every
unique index is that not all index declarations can be represented
by SQL-standard constraints.
regards, tom lane