[tex-hyphen] tex patterns as lua files

[Manuel Pégourié-Gonnard]

Le 27/04/2010 11:36, Élie Roux a écrit :
>> Once again: I would like to discuss the proper inclusion of dtx file
>> once (since I don't understand yet how exactly that works),

Well, it's rather simple: you have one source file (hyphen-luatex.dtx) and when 
you process it with tex, it gives two runtime files (hyphen.cfg and 
hyphen-luatex.lua), while processin with pdflatex creates the documentation file 
(hyphen-luatex.pdf). Of course this is just for this particular dtx.

>> You may put the dtx file under hyph-utf8/source/generic/hyph-utf8
>> (feel free to decide whether you want a subdir or not) and the rest
>> under hyph-utf8/tex/[generic or luatex - no
>> idea]/hyph-utf8/[something]
>
> I'll do that!
>
Definitely hyph-utf8/tex/luatex/hyph-utf8/hyphen.cfg (rather than 
.../generic/...) if we want to be sure our hyphen.cfg is found before the 
standard one when running luatex (and conversely, that the standard one is found 
first when not running luatex).

>> And the same question is still applicable: should I rename the files
>> to hyph-xx.xxx.txt?
>
> It would make sense for me too, It would be more clear.
>
Would make sense to me, too. While we're at it, I'd like to suggest some other 
name than "plain" for the directory part : this is rather confusing since it can 
be understood as "plain tex" in this context. Something like plaintext, or, 
clearer but longer plaintext-patterns seems more appropriate to me. (This 
doesn't really matter from a purely technical POV anyway.)

>>> What do you think?
>>
>> It can either be auto-genereted and kept inside hyph-utf8 tree or
>> auto-generated by tlmgr and kept somewhere else, but maybe the first
>> option is better since you don't depend on tlmgr that much and it
>> doesn't bring any speedup or any other advantage if auto-generated by
>> tlmgr. Just more external dependencies and the need to fix tlmgr in
>> case that you will need any changes in file format. So I suggest that
>> we move language.dat.lua somewhere inside hyph-utf8's TDS tree as
>> well.
>
> It also makes sense for me, the file can be in its complete form even
> if languages.dat is not,

Good point.

> so the first solution seems better.
>
There still are a few edge cases which I'm investigating, but currently I agree.

Manuel.