Re: Bug 130397 (Was: Emacs - Ispell problem with i[no]german dictionary)

[Kenichi Handa]

In article <address@hidden>, Agustin Martin <address@hidden> writes:
>>  Should I install my
>>  patch for CVS Emacs.  Or, is it better to wait for you or
>>  some other maintainer work on it?

> I have retested your patch and my workaround with a good ifrench dict
> (it was indeed buggy), and both give reasonable results for
> iso-8859-{1,15} dict/buffer combinations, with some of the expected
> misalignments due to iso-8859-15 chars, but your patch does the good
> work for an utf-8 buffer with iso-8859-15 only chars and an
> iso-8859-15 dict emacs entry, while my workaround does nothing there.

> I suggest you to install your patch for CVS Emacs, so it can have a wider
> testing. 

Ok, I've just installed it.

> I will try to adapt both systems for use in dictionaries-common ispell.el,
> with your patch as primary choice and the workaround as a fallback.

I see.  Thank you.

The remaining problem is that a character sequence
recognized as a single word by Emacs won't be recognized as
a word by ispell-get-word because of an existence of
character that can't be encoded by a coding system specified
for the current ispell dictionary.  For instance, if we
check a buffer containing "español" by american dictionary,
"espa" and "ol" are cheched separately.

This problem is not that serious as the misalignment error,
but is better to be fixed somehow.  Stephen suggested that
such a word should be skipped or marked as "uncheckable".  I
think the latter is better because 'ñ' in the above case may
be a typo.

This kind of situation can be detected by checking if the
the first and last characters of a word detected by
ispell-get-word is surely at word-boundary.  But, I don't
know where to check it nor how to mark something
"uncheckable" or "unknown word".  It requires a work by
someone who knows ispell.el well.

---
Ken'ichi HANDA
address@hidden