Re: [PATCH] completion: add _uids, _gids

[Oliver Kiddle <opk@xxxxxxx>]

dana wrote:
> you're suggesting that with -Nu we should only offer the uids when
> $PREFIX == [0-9-]* basically? i could do that

More specifically, when the compadd for the names failed to add matches.
Which will have much the same effect in most cases.

It could also be worth checking how often -Nu is applicable. If nearly
always, maybe that should be the default.

> i forgot about that, but fwiw i think it's easily solved by just not
> passing the -g to compdescribe -I

The vague documentation for compdescribe doesn't cover the combination.
Is that just turning off the common description collecting or is it a wild
 stab in the dark - compdescribe internals aren't trivial.

> On Tue 14 Jul 2026, at 18:24, Oliver Kiddle wrote:
> > Next _describe does not do a good job of accepting more unusual compadd
> > options like -A, -D. It even conflicts for -O.
>
> they seem to work for me generally i think? how do you mean?

I've not been at a system where I can test - am typing into a phone. I
know I've hit blocks with its peculiarities which may have been since
fixed or could be.

> like it'd build the display + insert arrays and then just give them back
> to you (with expl)? when wouldn't you want to pass those straight on to
> compadd?

Maybe just the array for use with -d, conventionally named disp. You
also get a full tag/label loop with _describe. Nesting those is not always ideal. I'll
need to prototype and test a few variations which I won't be able
to do for a few weeks.

Oliver