"Dan McFadyen" <danm@cryptocard.com> writes:
> You're right, the second one does use an index, one that is used to
> enforce a unique constraint on the column. I wasn't able to turn it off
> as the database is currently in use, and disabling a unique constraint
> probably isn't a good idea.
> Sorry for not mentioning the PG version, it's 8.3.
> Now, as for the LC_CTYPE, I didn't even know about such a thing till you
> mentioned it, and I found the lovely disclaimers in the postgresql docs
> about the consequences of it not being c or posix.
> It's Finnish Finland 1252. If that's the probable cause, I'm guessing
> the only option is to export, re-initdb, and import.
Oh, this is Windows huh? In principle, UTF8 encoding should work in all
locales for Windows --- at least, the code thinks so, which is why it
let you create that combination to start with. I'm not very familiar
with the possible pitfalls though. If you don't actually have any use
for Finnish-specific sorting or case-folding, a re-initdb in C locale
would probably be a good idea.
In the meantime it seems there might be a bug here. Can you give more
information about the contents of the column involved, in particular
what non-7-bit-ASCII characters it contains?
regards, tom lane
PS: please cc the list on replies.