Re: pg_trgm

Greg Stark <gsstark@mit.edu>

From: Greg Stark <gsstark@mit.edu>

To: Tatsuo Ishii <ishii@postgresql.org>

Cc: tgl@sss.pgh.pa.us, peter_e@gmx.net, ishii@sraoss.co.jp, andres@anarazel.de, pgsql-hackers@postgresql.org, teodor@sigaev.ru

Date: 2010-05-29T14:09:12Z

Lists: pgsql-hackers

On Sat, May 29, 2010 at 9:13 AM, Tatsuo Ishii <ishii@postgresql.org> wrote:
> ! #define iswordchr(c)  (lc_ctype_is_c()? \
> !                                                               ((*(c) & 0x80)? !t_isspace(c) : (t_isalpha(c) || t_isdigit(c))) : \
>

Surely isspace(c) will always be false for non-ascii characters in C locale?

Now it might be sensible to just treat any non-ascii character as a
word-character in addition to alpha and digits, so what might make
sense is

   t_isalpha(c) || t_isdigit(c)) || (lc_ctype_is_c() && *(c)&0x80)

Though I wonder whether it wouldn't be generally more useful to users
to provide the non-space version as an option. I could see that being
useful for people in other circumstances aside from working around
this locale problem.

-- 
greg