Re: A DISTINCT problem removing duplicates

Richard Huxton <dev@archonet.com> writes:
> Tom Lane wrote:
>> Richard Huxton <dev@archonet.com> writes:
>>> Anyone got anything more elegant?
>> 
>> Seems to me that no document should have an empty dup_set.  If it's not
>> a match to any existing document, then immediately assign a new dup_set
>> number to it.

> That was my initial thought too, but it means when I actually find a
> duplicate I have to decide which "direction" to renumber them in.

Hmm, so you mean you might decide that two docs are duplicates sometime
after initially putting them both in the database?  Seems like you have
issues with that anyway.  If you already know A,B are dups and
separately that C,D are dups, and you later decide B and C are dups,
what do you do?
        regards, tom lane