Case insensitive pattern matching gives strange results for non-ascii character (such as UTF-8 encoded cyrillic letters): test=# select 'б' ~* 'Б' ; ?column? ---------- f (1 row) ( 'б' and 'Б' are lower and upper case variants of cyrillic 'B')
at the same time: test=# select 'б' ilike 'Б' ; ?column? ---------- t (1 row)
(PG 8.3 on Linux, UTF-8 locale)
Also, what could be the reason for that cyrillic letters are not treated by regexp engine as the part of [:alpha:], [:alnum:], \w etc. classes? Or they never meant to be?