Re: Grouping by initial letter

[Donal Lynch]

So if I understand correctly, the initial characters need to be converted into their unaccented forms? I'm not even sure if the collation machinery knows that "Ä" is a modified form of "A" any more than "A" is a modified form of "Ä", it's just aware of a symmetric equivalence. There are upper() and lower() string functions but there doesn't seem to be anything like an unaccented() function that would take advantage of knowing that one form of the letter is the "normal" one.

One workaround, if nobody comes up with something better, would be to join the result with an alphabet table that you've populated with the letters you're interested in for each language.