[Dict-common-dev] Re: Debian transition to Aspell 0.60

Agustin Martin agustin.martin@hispalinux.es
Thu, 27 Jan 2005 13:13:48 +0100


On Fri, Jan 21, 2005 at 01:40:47AM -0800, Brian Nelson wrote:
> You may find the new aspell packages available at:
> 
>   http://people.debian.org/~pyro/pending/
> 
> Please test building dictionaries with the new packages and let me know
> if you have any problems.  Also, if you find any flaws in my proposal,
> please let me know.
> 

(Note, I am bcc'ing aspell dicts maintainers rather than cc'ing them, alioth
do not like long cc lines and your first mail did not get archived. Sorry
those that receive this mail twice. Not {b}cc'ing the aspell depending
pakages)

Please find updated aspell-gl-minimos packages in

  http://dict-common.alioth.debian.org/aspell6/

(with parts in the sources subdir). It is built using affix compression
(7,3M -> 363k). Since this might be of help to others I am giving some info
about the required steps, xx stands for the two letter code of your
language (e.g. en), or for the two letters code + variant (e.g. gl-minimos
in my package)

I find this very interesting when all dicts are built from the same source,
or can be done this way.

** Building the hash:

You need to adapt the xx.dat file with the affix info, e.g.

# --------------- .dat file --------------------
# xx data file
name           xx
charset        iso8859-1
soundslike     xx
affix          xx
affix-compress true
# ----------------------------------------------

The last two lines are the important ones to add.

You need to have the myspell affix file (e.g. xx_XX.aff), although aspell
will expect it as xx_affix.dat (a symlink should suffice for hash
building).

You also need the munched wordlist in the ispell format (myspell one simply
has an extra leading line with the number of roots for parsing speedup). If
you only have the myspell dict, something like 

cat xx_XX.mydict | sed '1d' | aspell create .. options .. 

should help, but you can use directly the ispell munchlist if you have it
without stripping the first line,

cat languagexx.ispelmunchlist | aspell create .. options ..

** Installing everything

Besides the usual files, remember to install xx_affix.dat (with that name)
in /usr/lib/aspell-0.60 (Remember that /usr/share/aspell seems no longer used
by dicts, and that you must install now in /usr/lib/aspell-0.60 rather than
in /usr/lib/aspell)

> I'm going on a pseudo-vacation for the next two weeks and would really
> like to upload the new packages when I return.  I'd appreciate it if
> someone would setup a staging area to collect all of the newly built
> dictionaries so that they can all be uploaded together.  Please
> coordinate on the dict-common-dev@lists.alioth.debian.org mailing list,
> unless of course Agustin objects.  ;)
> 

No objection at all, of course.

Some other remarks for aspell dicts.

Please register your aspell dict for use in emacs, look at

  http://dict-common.alioth.debian.org/dsdt-policy.html#aspell-registration

You do not need to add all the entries the equivalent ispell dict has, just
the ones that are relevant to aspell (e.g., no need to add most of the tex
variants).

Cheers,

-- 
Agustin