Re: Per command _redirect completion

[Oliver Kiddle <okiddle@xxxxxxxxxxx>]

Sven wrote:

> - Does anyone know about other places than redirections and values
>   where this might be interesting to use? Even now it is already easy
>   to define more sets.

Contexts. They've always been mixed in with the commands but I think
it'd be logical to now separate them off (with the possible exception
of -default-). There wouldn't be much use for service and patterns with
them though so the value of separating them is questionable.

The only other places I can think of are arguments to math functions
and condition codes but they are probably best left until we need them
(I thought about adding a file size math function to the stat module so
it might be useful someday with that sort of thing).

> > Have I also understood correctly, that the compdef changes allow
> > multiple uses of -T so that _gzip might start with:
> >     #compdef gzip gunzip gzcat=gunzip -T values GZIP
> 
> Right. Good? ;-)

Yes. Good.

Only I just wonder now: if we have to triplicate the list of commands
just to handle redirections (normal, < and >) it might end up being a
bit tedious.

> Felix Rosencrantz wrote:
> 
> > Looks good to me.  Though am seeing an error with the value
completions.  While
> > the redirs and normal completion don't seem to have a problem. 
> > 
> > host@ TZ=<TAB>
> > _normal:74: division by zero
> > host@ DISPLAY=<TAB>
> > _normal:74: ':' without '?'
> > 
> >
_normal:74:service="${${(e):-\$${__services}[$_comp_command1]}:-$_comp_command1}"
> 
> Yup, had this too, somtimes. Try the new and improved patch below...

Doesn't seem to be solved - `_do_type:31: ':' without '?'' after
TZ=<tab>

And, f < <tab> is a bad math expression: operand expected at `<'

> +++ ./Completion/Base/Core/_normal	Wed Feb 27 22:29:17 2002

> +_set_command
>  
> -return ret
> +_do_type "$skip[@]" comps "_$comp_command1" "_$comp_command2"

This last line needs the dollars and underscores swapped otherwise all
you get is service=_ and default completion.

> Only two small value-functions left.

And those aren't so small and could arguably live inside _gcc.

> > > diff -ur -r ../oz/Completion/Base/Utility/_contexts
> > > +# You may also select the assocs to use with the `-T assoc'
option.
> > 
> > I know we only renamed it a year ago but _contexts perhaps could be
> > better named.
> 
> Hm. I don't think it's too bad, but then I'm not a native English
> speaker.

It isn't particularly easy thinking of good names even when it is my
first language. My problem with the name _contexts is that it isn't
just a context (i.e. -parameter-, -equal- etc) that can be given to it
as an argument. It can be commands and now also values.

> The stuff in _normal has been put into separate function
> `_do_type', which needs a better name (for which I need help).
> 
> Between the things to do:
> 
> - Decide on a name for the concept so that we can give _do_type its
>   final name and update the docs.

How about _dispatch?

_contexts might then be _redispatch. Could they be merged as one
function? I can't see a reason why _contexts oughtn't to be honouring
pattern compdefs and they do basically the same job.

Thanks.

Oliver

__________________________________________________
Do You Yahoo!?
Everything you'll ever need on one web page
from News and Sport to Email and Music Charts
http://uk.my.yahoo.com