bug#13639: [emacs] ispell.el: hunspell dicts autodetection under Emacs.

[Eli Zaretskii]

> Date: Wed, 20 Feb 2013 18:50:45 +0100
> From: Agustin Martin <agustin.martin@hispalinux.es>
> 
> > > > > Sorry, I should have written WORDCHARS.
> > > > 
> > > > Why do we need that?
> > > 
> > > This is what ispell.el calls otherchars. Parsing WORDCHARS ensures that
> > > both
> > > hunspell and ispell.el think about the same characters in that category.
> > 
> > I think you are mistaken, that's not my reading of hunspell(4).
> 
> Sorry for the late reply,
> 
> (Opening a new thread specifically about hunspell dicts autodetection and
> using new cloned bugreport #13639 specific about this)
> 
> Although WORDCHARS description in hunspell(4)
> 
> WORDCHARS characters
>    WORDCHARS extends tokenizer of Hunspell command line interface
>    with additional word character. For example, dot, dash, n-dash, numbers,
>    percent sign are word character in Hungarian.
> 
> is too hungarian biassed and does not mention usual apostrophe AFAIK it
> mostly refers to the same as 'otherchars', although hunspell may accept
> that in locations not in the middle of a word.

I didn't just read the man page, I also looked into several *.aff
files that install with Hunspell dictionaries.  It is clear to me that
WORDCHARS is at least unreliable, even if your interpretation is
correct (of which I'm still unconvinced): some *.aff files don't have
that entry at all (e.g., en_GB.aff, whose OTHERCHARS should include
the ' character, and also ru_RU.aff); others, like he_IL.aff, have
that entry mention all the CASECHARS, in addition to OTHERCHARS.  I
wouldn't bet my money on what that entry gives us.

> The good news are that I started working on hunspell dicts autodetection.

Good news, indeed!  Thanks!