Re: Speck (Spell checker for Emacs)

[Eli Zaretskii]

> Date: Tue, 21 Apr 2015 12:34:06 +0200
> From: martin rudalics <rudalics@gmx.at>
> Cc: help-gnu-emacs@gnu.org
> 
> Initially, speck was written without Hunspell support.  I added it as
> soon as the first executables for Windows appeared (that was quite some
> time before any other Emacs Spell Checker was able to talk to Hunspell).
> The customization variables are still from that early period, there was
> no "normal" installation procedure at that time IIRC.  Moreover, I never
> upgraded my spell engines since then.  So probably support is missing or
> incorrect in this area.

FWIW, when I was reading the Hunspell-related parts of Speck, I found
myself wondering time and again why its invocation is so complicated.
You don't really need to tell it where its dictionaries are, because
it already knows where to find them: that's what a correct
installation guarantees.

Likewise with encoding: you should use UTF-8 throughout, as that's the
native encoding for Hunspell; anything else it will need to convert.

Finally, Speck should really learn from ispell.el how to set up most
or all of the variables by asking Hunspell itself, instead of asking
the user to figure out how to do that by hand.

> I faintly remember Xah Lee spending hours on setting up Hunspell for
> Windows until he gave up.  Maybe that was the reason.

Upstream Hunspell is buggy, even today, and the Windows build is even
more so.  It needs non-trivial patches to work correctly.  (E.g., I
believe it still produces byte offsets instead of character offsets,
which won't work in Emacs with UTF-8 encoding.)  I suspect that people
who "gave up" on Hunspell simply used one of those "compile and
forget" Windows ports of Hunspell, made by people who never test their
ported binaries before they make them available for download.

Hunspell which has its bugs fixed and works correctly doesn't need any
setup with ispell.el/flyspell.el, and only a bit more so with Speck.