Re: compadd option passing (Re: Completion function for bitkeeper?)

[Oliver Kiddle <okiddle@xxxxxxxxxxx>]

I sent this two weeks ago and just realised that it never turned up on
the mailing list.

Bart wrote:

> } I agree that it isn't ideal. The best alternative I can think of would
> } be to pass compadd options to the various tag handling functions
> } instead of the completion functions.
> 
> That makes it harder for people writing simple completion functions to
> make use of the compadd options, though.  Maybe that never comes up, but
> it seems an unnecessary limitation.

True. But it could help other issues by doing more of the hard work
with picking out suffixes and descriptions and passing on only the
right stuff. Auto-removable suffixes especially stand to benefit if we
can store them at each stage. It needs a lot more thought though. 

> I suggest instead that we choose a single option (or other syntactic
> marker) that is documented as being passed to completion functions to
> delimit the compadd options from the any options of the function itself.
> E.g. soemthing like "everything between -o and a bare '-' is a compadd
> option":

> Alternately always pass the compadd options in an array parameter, and

Either do avoid name clashes but, to be honest, with all of this I think
either of those solutions are uglier than the situation we have at the
moment.

If we really want to clear it up in such a way, How about a
single `-' argument be defined as the start of (or end of?) compadd
options.

So you have to use for example.
  _sccs_files -e - -X 'sccs file'

If there is no `-' argument, it can assume that all options are compadd
options (which will be most function calls anyway).

_all_labels already looks for `-' in deciding where to put compadd
options so it'd also be the easiest to implement.

A possibility included in this is that we only do anything special for
`-' in _arguments actions (to clear up our initial space vs. { ... }
confusion).

Oliver