Re: .cmap and .cset files for languages that share scripts

Gora Mohanty wrote:

>   For languages that share scripts, e.g., Hindi and Marathi that share the
> Devanagari script, packagers of the standard aspell dictionary
> distributions run into a problem because both dictionary distributions
> would contain identical u-deva.{cmap,cset} files. However, if one of the
> dictionaries is already installed, the second package reports problems
> with conflicting files, and refuses to install unless things are forced. A
> way around this would be to have separate u-deva-hi.txt and
> u-deva-mr.txt files (from which the .cmap/.cset files are generated), in
> the aspell-lang distributions. Of course, these files would be identical,
> but that might not be much of an issue as they are a single line each.
> Is that possible?

I do not see any fundamental problem with distributing identical .cmap/.cset
files in different packages.  This files will NEVER conflict with each other
since they should always be identical.  The fact that they are causing you
problems is a limitation of the Debian packaging system.

In order to implement your solution EVERY u-*.txt would need to be renamed to
u-*-<LANG>.txt as any one of them may cause a problem for you in the future.
Doing this will involve repacking all the dictionaries that use them in a
slightly incompatible way.  This is not something I am not willing to do.

Rather than renaming the file may I suggest you create a separate Debian
package which includes all the u-* character sets and than have other
dictionaries depend on this package.

You can also rename the files your self.  You don't even have to regenerate
them.  Just rename them and change the '= ' line on the top of the .cset file,
and perhapses add a comment such as "generated from u-*.txt but renamed to
avoid conflicts ...".