Re: case-table functions clobbering extra slots

[Dave Love]

Richard Stallman <address@hidden> writes:

> Of course, to really solve these problems, after the release, we need
> a more thorough change.  I think Greek uses two forms of lower-case
> sigma, one at the end of the word, one elsewhere.

Yes.

> To downcase correctly requires determining where the sigma is.

Yes, but I'm not sure it's really worth it.  Greek experts should
probably comment.

> Then there's the problem of ß in German, which I believe should
> upcase to SS.  And SS in some (but not all?)  cases should downcase
> into ß in German.

I'm not sure the latter is actually done.

> I think the mechanism of case tables is thus inadequate.  The question
> is whether to replace it or extend it.

I thought I said something about this before, but maybe I decided not
to confuse the issue.  I think you have more important things to deal
with.

I think you could make an extra extra slot for the table to specify
special cases by arbitrary (pairs of?) functions.  Emacs 22 allows
caseification to change byte length in a buffer, but I don't know
whether it could deal with changing character length of what the
situation is for strings.

> Do you have a complete list of problem cases for case conversion in
> Emacs?  Would you like to start designing a full solution?

The relevant Unicode tables are
http://www.unicode.org/Public/UNIDATA/SpecialCasing.txt and
http://www.unicode.org/Public/UNIDATA/CaseFolding.txt; I think they
reference the Unicode standard and associated technical reports.  I
assume the data are programmed into libc locale functions.

I did Turkish `i's for Emacs 22.  As far as I remember, there's a
similar issue for some other language(s) for which you don't currently
have an environment.  You can't do it in 21 because of the restriction
to the same byte length.