Print

Print


Vladimir,

 > This is my first post to this mailing list, so I'm sorry.

why being sorry for that? you pick up on a disscusion that happened
not too long ago on this list and with a new topic

 > LaTeX has some `built in' definitions of \lccode, \uccode, \sfcode
 > values for symbols with codes above 128 (which are used in Russian
 > language). But these settings do not conform to the common Russian
 > encodings (and it is evident that these `frozen' values can not
 > be in agreement with all possible languages ;-) ).
 >
 > Next, it is written in cfgguide.tex, that the values of above mentioned
 > registers should not be changed e.g. in a hyphenation files
 > of a particular languages.
 >
 > So, I think, that the correct way to solve this problem is to
 > set the values of \lccode, \uccode, \sfcode and \mathcode,
 > in a macros which switches to a language (\selectlanguage
 > in Babel).

well, not really. what we mean here is that they should not be changed
at all. the reason is that otherwise due to a deficiency of TeX you
can't do multi-lingual typesetting at all properly.

the reason is that TeX does support several hyphenation patterns but
unfortunately it uses the lccode table during hyphenation and it
doesn't support several lccode tables. even worse: it looks at this
table only at the end of a paragraph. as a result, if you have a
paragraph that mixes several languages then all languages might be
hyphenated according their correct hyphenation patterns, BUT before
this happens all words are translated to lowercase using the lccode
table for the LAST language in the paragraph. That makes the result
quite arbitrary.

therefore, to support fonts with different encodings within the same
paragraph (you don't even need different languages since in tex
hyphenation is tied to the font encoding so you need different
patterns for different fonts even for the same language) we have to
enforce a single lccode table.

that's an unfortunate fact of life.

now that doesn't help very much, i agree. so the questions we have to
ask is what can be done about it

 a) just support a single lccode table

 b) don't care about rubbish hyphenation when languages are mixed
    and allow for several hyphenation tables

 c) allow to change lccode tables only between paragraphs and disable
    hyphenation within language fragments in a paragraph that do not have
    the have the right (meaning: current) lccode table

 d) hope for a successor to TeX to fix it

right now we basically have situation a) which means LaTeX does not
support changing the lccode table. this does not mean that packages
can't do it but any such code is likely to break at some time in the
future and we, from the LaTeX3 project don't feel able to support
problems with a LaTeX system that does use a modified table. in
addition documents written for such a system will produce strange
results on others, ie we can't have portability there and definitely
not real multi-lingual.

b) means putting lccode changes into the framework of Babel as it is
now. that is sometimes done right now and Babel supports setting any
resetting things when entering or leaving a language environment so if
you are happy with those poor results when mixing language that sort
of works

c) is the situation where i think we can get to as long as we use TeX
as a basis and it is the scheme i intend to adopt for the new language
interface for LaTeX for which the conceptual work is mostly done and
trial implementation is done in parts.  this would give a clean
interface and only minor inconvenience, ie, if languages are mixed
within a paragraph then LaTeX will not not hyphenate part of it
(warning you about it) and you have to put in explicit hyphens there.
but it would not produce rubbish in this case, eg something you do
only notice after publication.

d) would be my wish and indeed there is a successor that does fix it
namely Omega. but unfortunately we currently have two competing
successors and neither is fit to be used as a basis for development
because both solve important issues but both solve different ones and
neither is widespread (and part of the reason why is that neither
solves enough to make it worth switching for the majority of users).
Sorry Phil, but as good and nice e-TeX is in parts, the big problems that
Omega solves aren't touched by it and as long as this is not happening
it isn't a serious candidate. --- what i would hope (but i fear in
vain) is that we take Omega and extend it with the e-tex features as
they are now and freeze that (FREEZE!) or vice versa and then we have
something that we could promote and more importantly could use to
develop serious code for, eg the LaTeX kernel.

 > Are there some ready means to preserve the `language environment'
 > before switching to another language and to restore it after
 > changing back (or to some third language)?
 > Is it a normal practice to `tune' \lccode, \uccode, \sfcode
 > and \mathcode values in a language-switching mechanism?

it is not a "normal" practice for the reasons outlined before.  as for
\mathcode this is something for which LaTeX has a high-level interface
but it is also something that is considered static and in that case i
don't really see that it could reasonably be maintained or even
should.

 > Here I see another problem. Consider some multilingual
 > phrase and let's assume that we need to do
 > \uppercase{} or \lowercase{} of this phrase. If the languages contained
 > in this phrase have a conflicting \lccode and \uccode
 > values, than the result of \uppercase{} will be incorrect.
 > However, some time ago I solved this problem by redefining
 > TeX's builtins \uppercase and \lowercase so that
 > these new macros split the multilingual phrase into
 > one-language pieces and do \uppercase of these pieces,
 > and then merge them again.

again a deficiency of TeX that really isn't solvable generally, yes
you can do some hacks by trying to write your own interpreter within
TeX but it is so easy to make it fall over, just think about the user
putting something in a macro to save typing, so how do you find those
hidden language changes? by parsing macro code? (well we all have fun
working on exotic code but it isn't worth it for serious use.

reliably this can right now only be solved by using uppercase fonts in
such places. Or, using Omega ...

 > May be, all this was already solved?

well yesno :-(

Phil replied to Vladimir:

 > One idea under consideration for e-TeX V2 is the option to save and
 > restore an entire set of register values (e.g. the current set of
 > lccodes, etc); this may prove useful in solutions to this question.

a nice feature and probably useful for a number of things BUT the
problem is a different one here and as i said above a real solution
for me would be a merger of e-TeX and Omega (and pdftex to help
promoting a switch)


good night
frank
technical director LaTeX3 project (only ;-)