Re: is* character classification

On Wed, Nov 09, 2011 at 05:14:23AM +0000, corvid wrote:
> 
> I was just looking at how the is* functions treat characters with the
> high bit set. If I set LC_CTYPE to en_US, 0xC0 is alpha, but it
> understandably isn't in en_US.UTF-8.
> 
> So I found my way to
> /usr/share/i18n/locales/  en_US -> en_GB -> i18n
> 
> And found that:
> upper 41-5A, C0-D6, D8-DE
> lower 61-7A, B5, DF-F6, F8-FF
> alpha 41-5A, 61-7A, AA, B5, BA, C0-D6, D8-F6, F8-FF
> digit 30-39
> space 09-0D, 20
> cntrl 00-1F, 7F-9F
> punct 21-2F, 3A-40, 5B-60, 7B-7E, A0-A9, AB-B4, B6-B9, BB-BF, D7, F7
> graph 21-7E, A0-FF
> print 20-7E, A0-FF
> xdigit 30-39, 41-46, 61-66
> blank 9, 20
> 
> In dillo:
> isupper not used
> islower not used
> isalpha cssparser, html
> isdigit doesn't matter
> isalnum cssparser, html, datauri, downloads, dlib.h
> dIsalnum url
> isspace cssparser, html, keys, dlib.h, findtext.hh
> dIsspace colors, src/cookies, misc, dpi/cookies, file, ftp, dlib
> iscntrl keys, auth, misc
> ispunct textblock
> isgraph not used
> isprint dlib
> isxdigit doesn't matter
> isblank not used
> isascii not used
> 
> I'm not sure that any real harm is done in these cases, but it might not
> be a bad idea to check that it's ascii when we only want ascii.

Sure.

I could imagine you already regret that you started looking into this stuff :)