Re: Distinct oddity

Maximilian Tyrtania <maximilian(dot)tyrtania(at)onlinehome(dot)de> writes:
> FAKDB=# select distinct f.bezeichnung from firmen f order by 1 limit 5
> FAKDB-# ;
> bezeichnung
> -----------------------------------------

> sterreichisches Verkehrsbro AG
> \x01Assistenz
> \x10Frohstoff Design & Textilveredelung
> "1. Mittelschule ""Am Kupferberg"""
> (5 rows)

> FAKDB=# select distinct f.bezeichnung||'e' from firmen f order by 1 limit 5
> ;
> ?column?
> ------------------------------------------
> sterreich/Welt (Ltg.)e
> sterreichisches Verkehrsbro AGe
> \x01Assistenze
> \x10Frohstoff Design & Textilveredelunge
> "1. Mittelschule ""Am Kupferberg"""e
> (5 rows)

> Aha, the "sterreich/Welt (Ltg.)"-entry is missing in the 1st query. So that
> does smell like a locale problem.

That only proves that adding the 'e' changes the sort order, which is
completely unsurprising for any non-C locale. What you need to do is
dump out the *entire* results of the DISTINCT queries and look for the
unmatched lines. I'd try dumping to two files, stripping the 'e' with
sed, and then sort/diff.

regards, tom lane