Re: Unicode 4.01 bidi type changes

On Tue, 27 Apr 2004, Beni Cherniavsky wrote:
> It's a tough world :-(.  You have a point.  Perhaps it's not a good idea
> after all.

I go with the U+2212 MINUS SIGN.  Glibc has already limited
support for localized numerals, so printf("%Id\n", 123) will
print Persian digits under the fa_IR locate.  Recently Hamed
(CCed) extended that, to use localized decimal-point and
thousands-separator too.  Moreover, we worked with Bruno Haible
to add the support in gettext, such that it removes the "%Id"
non-portable "I" flag on systems that does not support it.  So
when translating, you can translate "%d" to "%Id" to get
localized digits...

TODO 1:  Glibc should use U+2212 MINUS SIGN in negative numbers
printed with i18n ("I") flag.  Or if not acceptable (which is 99%
the case), the locale should be able to localize the minus sign.
Should be pretty straightforward to do after Hamed's patch.

Glibc also supports parsing localized digits by scanf("%Id"...)
too, but due to some problems in the ISO C standard, it's broken
now.  The same problems prevent Glibc to accept localized digits
by isdigit, iswdigit, strtod and other friend functions, which is
really sad :(.  So the only hope is to fix scanf("%Id") which is
already a GNU C extension.  Perhaps Glib can later hide this
problem, by using scanf instead of strtod, but it's probably
going to be a performance hit.  So:

TODO 2:  Glibc should recognize localized digits in scanf family
of functions with i18n ("I") flag.  Moreover, it should accept
U+2212 MINUS SIGN, and localized decimal-point and
thousands-separator too.

Any idea?

Hamed, we are waiting for your patch to be applied :).  It's much
needed now due to the change in Unicode data files.

Cheers,
--behdad

PS. Oh, But wait:  What about plus sign??  It all smells really
bad.  Right now there are only three characters as type ES, and
the three are:

  U+002B PLUS SIGN
  U+002D HYPHEN-MINUS
  U+FF0F FULLWIDTH SOLIDUS

Note that how:

  * PLUS SIGN is ES, while PLUS-MINUS SIGN is still in ET!

  * HYPHEN-MINUS is ES, while HYPHEN is ON and MINUS SIGN is ET!

  * FULLWIDTH SOLIDUS is ES, while SOLIDUS is moved to CS!


We're gonna experience crazy things soon...

Cheers
poor behdad.