[lltx] issue with unicode-math, spanish and LuaLaTeX

[Ulrike Fischer]

Am Tue, 12 Jul 2011 16:38:23 +0200 schrieb Philipp Stephani:

> Am 12.07.2011 16:22, schrieb Ulrike Fischer:
>> Hm. I understand \mathcode and \Umathcode actually look at the same
>> value and so are not independent. But I have the impression the
>> problem is less with \mathcode (which after all gives a value) but
>> with \mathchardef which doesn't accept it if the \mathcode is too
>> large.
> 
> This is not an issue of a \mathcode being too large, but about 
> overlapping representations. Say you have a character with class 0, 
> family 0 and code point U+1111; its \Umathcodenum is 0x1111. But the 
> \mathcode of a character with class 1, family 1, and code point U+11 is 
> 0x1111 as well---how should \mathchardef distinguish between the two?

But this ambiguity is with *small* numbers -- in the range 0-7FFF
(or 0-8000). And as far as I can see for these values \mathchardef
simply assumes that they mean values in "old" mathcode syntax (which
is imho the right decision). So why shouldn't \mathchardef be
extended so that in the range above is works like (the non-existing)
\Umathcharnumdef?  

(I agree that the syntax ambiguity of \mathcode values in the range
0-7FFF is a problem and that \mathcode/\mathchardef should better be
avoided. The suggestion is only for compability reasons)


Btw: while looking at the \Umathcodenum I saw that some arguments
are converted to curious high numbers. Mostly the letters in the
ascii range:

\Umathchardef\test="0"0"24
\showthe\Umathcodenum\test

> 36

\Umathchardef\test="0"0"41
\showthe\Umathcodenum\test

> 14799924




-- 
Ulrike Fischer