Subject: Re: ispell blues (was: localization formats proposal)
From: Martin Sevior (msevior@mccubbin.ph.unimelb.edu.au)
Date: Sun Jul 22 2001 - 00:40:01 CDT
Hi Frodo,
> > Sorry I don't know. Frodo does though. Care to post a link to a HOWTO
> > Frodo?
>
> I was not really sure myself. One option is simply to use pspell, which
> will do the hard work for you - but I gather that is not an option on
> all platforms.
>
I think the question was: How to I create an ispell hash file. I believe
Karl would like native spell checking in his own version(s)? of Norwegian
for which no hash currently exists. Pspell is just a portable interface to
ispell here. What Karl wants, and I don't know how to do, is to actually
create an ispell hash file.
> I still find it a bit wasteful to have AbiWord use its own ispell
> libraries, when the system-wide ones are perfectly fine. I would like
> to suggest the following solution:
>
Except that many systems no longer have ispell hashes. RH moved to aspell.
In addition supporting the myriad system dependent location(s) of the
ispell hashes is a major pain.
One of the immediate problems we face for version 0.90 is how to package
AbiWord and how many binaries to we produce? We've lost Sam TH who created
most of the non-RH binaries of AbiWord. I can produce redhat 7.1
rpms. Thomas can make some QNX binaries. Who will make the Windows
binaries and what will be done with our 20 ispell hashes for Windows?
> * When we distribute binary packages, we include all AbiWord
ispell > libraries
I'm totally in favour of this, provided we also provide a minimal
English-only binary as well. I can do this for RH rpm. For other distros,
we'll need other volenteers.
> * When somebody compiles from source and selects ispell, not pspell,
> we try to find any system-wide installed ispell dictionaries. If
> none are found, we again install the AbiWord ispell libraries. If
> we do find a system-wide ispell, we determine what options are used
> in it (see below) and compile our own ispell support with the same
> options. Perhaps we still offer the option to use AbiWord's own
> dictionaries, though.
>
We can always point developers to how to compile with pspell or ispell.
> I have attached a small C program that will detect for you what options
> an ispell hash is compiled with (rewriting it in C++ is left as an exercize
> for the reader). When run on the (old) AbiWord ispell hashes you get this:
>
> (bash-2.05.0) frodo@arda:~/projects/abiword/ispell-detect$ ./ispell-detect ../abidistfiles/dictionary/*.hash
> ../abidistfiles/dictionary/BigEndian32.american.hash: big endian (options: 0002) #undef NO8BIT
> #define NO_CAPITALIZATION_SUPPORT
> #define MASKBITS 32
> ../abidistfiles/dictionary/LittleEndian32.american.hash: little endian (options: 0006)
> #undef NO8BIT
> #define NO_CAPITALIZATION_SUPPORT
> #define MASKBITS 64
>
> Comments?
Thanks for writing this program Frodo. It will be useful as is already.
However it doesn't appear to tell us the number of keys in the hash.
Debian makes hash files withh 100, while redhat makes (made) them
with 128.
Cheers
Martin
This archive was generated by hypermail 2b25 : Sun Jul 22 2001 - 00:40:15 CDT