Re: PATCH: _a2ps completion

[Peter Stephenson <pws@pwstephenson.fsnet.co.uk>]

Sven Wischnowsky wrote:
> Is this a valid reason to add a `glob-pattern' (or `file-pattern')
> style that is tested in _files and used to replace a function-supplied 
> pattern if there is any or to make _files use the pattern even if he
> calling function did not use -g? Or is this a case of `if you want
> that behaviour, copy and modify the completion function'?

I vaguely thought about this (i.e. asking you to do it :-)).  The only
reason I didn't mention it was that I vaguely thought there wouldn't be
enough context information.  But this turns out to be nonsense:

% a2ps ^Xh
tags in context :complete::a2ps::argument-rest
...

should be precise enough for anyone.

In fact, my original idea was for `glob-override', which would only modify
things if you already specified -g.  But as long as it doesn't mess up
completion of anything other than ordinary files --- i.e. it shouldn't
spoil directory completions, just alter things with the tags files,
globbed-files, or all-files --- it should be OK to use a more general
facility.

Preferably it should allow you to specify the tag you want altered,
i.e. you may want to alter just globbed-files, leaving all-files doing its
thing, or for special effects you might want all-files not to live up to
its name (e.g., I want to get rm to complete discardable files like *.o by
default (globbed-files), otherwise anything else (all-files), but I've
decide I never want it to complete vmlinuz). In which case, I suppose
`file-pattern' is the better name, since it can apply to all *files tags.
You could argue it should be `selected-files' to go with `ignored-files',
but I don't think they're really similar enough for that.

I suppose I'm going to have to include Clint's patch in 9333.

-- 
Peter Stephenson <pws@pwstephenson.fsnet.co.uk>