Re: case-table functions clobbering extra slots

[Dave Love]

Richard Stallman <address@hidden> writes:

>     set_case_table (in casetab.c) automatically setups up,
>     canon, and eqv tables from downcase table if they are not
>     set.  Existing of these tables suggests that one to many or
>     many to one mapping in upcase or downcase is supported, but
>     I don't know the detail of how those tables are computed.
>
> Could someone try setting up the upcase table for sigma,
> and see if it works right?

If you ignore the other extra slots and do

  (aset (char-table-extra-slot (standard-case-table) 0) ?ς ?Σ)

then, correctly:

  (upcase "σς")
    => "ΣΣ"
  (downcase "Σ")
    => "σ"

> What does the downcase table end up saying?

I don't understand.  It shouldn't change.

> Meanwhile, how does multibyte Emacs normally treat the sigma variants?
> Does it handle case conversion for them reasonably?

Normally

  (upcase "ς")
    => "ς"

which I was trying to fix.  I don't know what the criterion for
`reasonably' is.  It's only incorrect in that final sigma doesn't
upcase.

The above is for 8859-7.  Obviously likewise for the equivalent
mule-unicode characters.

I still don't think this is worth spending much time on in comparison
to other things which aren't getting fixed.