[tex-hyphen] German patterns in TL2010

Stephan Hennig mailing_list at arcor.de
Wed Jun 2 14:42:36 CEST 2010


[Resending, to the list this time.]

Am 02.06.2010 12:34, schrieb Manuel Pégourié-Gonnard:
> Le 02/06/2010 12:07, Stephan Hennig a écrit :
>>> Those field’s values must be the same as in language.dat.
>>
>> How is this ensured?  Purely manually?
>>
> Well, in normal operation, the two files are generated from the same
> information. If someone wants to modify one of the files manually (that is,
> using the language-local.dat(.lua) files), then it's up to him to ensure that.

What about generating language.dat from language.dat.lua, if the latter 
file exists (by fmtutil)?  I guess, I'm not a big fan of the top-level 
configuration files being distribution specific (.tlp*).


>> But why introduce 'loader' in the first place, when there is no use for it?
>
> I don't know, but too much (useless) information doesn't hurt (or so I thought).
> We can remove this field if you find it too confusing.

Well, TeX has never followed the KISS approach.  But I'd prefer not to 
add to the complexity if there is no gain.  (Note, my proposal to 
generate language.dat from language.dat.lua obviously requires a 
'loader' field.)


>> Wasn't the idea of converting patterns to text-only to abandon pattern
>> loaders?  Technically, converting patterns to text-only and making LuaTeX
>> load patterns dynamically are two different things.
>
> Not completely. The point is, if patterns can be loaded at any time, we can't
> say anything about TeX's internal state, especially catcodes, lccodes, etc. So
> it seems much safer to load patterns using only the Lua interface, which doesn't
> have these problems.

I haven't thought about that.  Thanks for pointing this out!


>> But as they happen at the same time, why keep old stuff around?  In case a
>> demand for loaders arose in the future, would it be difficult to re-introduce
>> them only then?
>>
> The loaders are still used by other engines now.

But those all gather the information from language.dat, no?


>> runpattern f texmf/tex/generic/hyphen/dehypht.tex
>>
> It's used by our packaging system to know that the file
> texmf/tex/generic/hyphen/dehypht.tex belongs to hyphen-german, even though it's
> not apparent in its path or name (usually, files are expected to contain the
> name of the package as a path component). Here, pattern is to be understood as
> "shell-like pattern for matching file names"

I was wondering exactly about that "pattern" thing.  Thanks for the 
explanation!

Best regards,
Stephan Hennig


More information about the tex-hyphen mailing list