For implementing incremental search in a GUI,
I'm looking how to write a SELECT, which will
find all rows where a columns starts with a
specific string (as in LIKE('abc%')), but taking
into account only primary character differences
(disregarding case and accents).

I'm aware of using David Schnepper's work on
nocase and nocase/noaccent collations, but after
thinking again about the problem, it seems better
to try to solve the problem without them (as I want
the output sorted according all differences).

What's missing so far is a method how to 'increment'
the search string to next string of equal length sorting
higher according to primary differences.

This 'incrementing' would return 'abd' for 'abc', and 'ac_' or
'abä' (depending on locale, err... collation) for 'abz'.

Having this incremented string the query would like
(:search holding the looked for string, :next holding
the incremented string):

> select col from t where
> (col >= upper(:search) or col >= lower(:search)) and
> col < upper(:next) and col < lower(:next);

The upper/lower thing must be done as it varies from collation
to collation which casing sorts lower. I hope it don't vary
from collation to collation that the unaccented letter sort
lower.

Any ideas for implementing the 'increment'?

Any alternatives for this requirements?

Best Regards,
Peter Jacobi,
Hamburg, Germany