Re: [Aspell-user] Using .cwl files

[Agustin Martin]

On Thu, Mar 06, 2008 at 10:36:41PM -0800, John Cage wrote:
> Ya got me... :-)   This is pretty foreign to me but I'd like to learn...
> 
> You wrote "$ preunzip foo.cwl".  Is 'preunzip' some utility I should download 
> somewhere? 

Should be built from aspell sources. However I do not know if win32 aspell
does provide a prezip/preunzip executable.

> Is this a DOS prompt command?

Was actually a unix prompt, but as long as you put the binary in the path
should work also from a win+DOS prompt.

> Which dictionary file should I work with? 
> In the 'dict' sub-directory of my Aspell directory, I see numerous 'multi' 
> files and the following files:
> "en_US-w-accents.multi",
> "en.multi", 
> "en_US.multi", 

Files containing a list of hash or multi files to be used under that name.

> "american-w-accents.alias", 

alternative name for a language having a .multi file.

> "en_US-only.rws",

The actual hash file.

> en.dat

Contains info about the dict and language, needed by aspell

None of them is to be changed, you should look at the non-processed dictionary,
like the ones available from the aspell site. there you will have either .wl
(non compressed wordlists) or .cwl (prezip compressed wordlists). If the
relevant dict uses affix compression you will not see plain words (unless
unique), but something of the form root/{affix-prefix flags} and fine
modification requires care and knowledge. IIRC the english dict does use
affix compression, so things are to be done with care unless you remove
full root+{af,pre}fix entries. Since en affix table has few flags this might
not be a problem.

> Doesn't the dictionary (Aspell) just use a single file which translates
> to my need to edit only a single file???

That is the .{wl,cwl} file, but you will need to build the hash file (the
.rws) later. Once you uncompress the .cwl and edit the resulting .wl, put
all the english .dat files (en.dat, affix and phonet) in the same dir and

aspell create master --local-data-dir=. --lang=en ./en-stripped.rws < 
en-stripped.wl

Then create an en-stripped.multi file contaning just "add en-stripped.rws"
(look at the other .multi files if in doubt) and instruct your app to use
the stripped dict. If a normal english dict is present en .dat files will
already be there and you will just need to add en-stripped.rws and
en-stripped.multi files to the place where the aspell dicts stuff is
located (do not know where that is in win).  Document somewhere what you
have done and where the dict comes from.

I am cc'ing the aspell-user list, in case Kevin Atkinson (aspell upstream)
or anybody else wants to correct me or add more info. Please keep the list
in replies, so more people can contribute and/or benefit.

-- 
Agustin