> Tatsuo Ishii <ishii@postgresql.org> writes:
> > similarity -> generate_trgm -> find_word -> iswordchr -> t_isalpha -> isalpha
>
> > if locale is C and USE_WIDE_UPPER_LOWER defined which is the case in
> > most modern OSs.
>
> Quite. And *if locale is C then only standard ASCII letters are letters*.
> You may not like that but it's not wrong; in fact any other behavior
> would be wrong.
> *if locale is C then only standard ASCII letters are letters*.
That's just the definition/implementaion of isalpha. My point is,
using isalpha is quite correct for pg_trgm or not. Text search, oracle
compat functions and any other string handling functions in PostgreSQL
behave quite different from what pg_trgm does.
The essential question is, are there any reason for pg_trgm to limit
to only ASCII range characters?
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese: http://www.sraoss.co.jp