Arg decoding with a template?

[Jim Dennis]

In article <rdeemts4mrarb8b5ghqieio68n1jv32tl1 at 4ax.com>, 
  Dale Strickland-Clark wrote:

>David Bolen <db3l at fitlinxx.com> wrote:

>>-- David

> I guess what I had in mind was a simple but effective equivelent that
> I used to use with BCPL. A function called rdargs, which I think also
> became part of AmigaDOS

 For those of us who are unfamiliar with these; do you care to 
 show an example of the syntax and explain why you think it's superior?

> I've just had a quick hunt with Google and I can't find a definition
> for it anywhere.

> I'll probably end up using getopt just because I don't have time to 
> do a proper job right now.

 It might be possible to create a subclass of getopt which 
 provides a similar API (list of methods) but accepts your 
 preferred syntax.

 BTW: It would be *REALLY* nice if you'd wrap you lines at <72 chars
 and provide a leading space (margin) to your posts.  80 character
 widths are still a dominant convention for USENET/netnews and 
 public mailing lists, and leaving a few characters for quoting
 prefixes is always appreciated. 
 
 Someone in this thread mentioned the convenience of '?-completion'
 on *switches* as well as filenames.  Of course in the latest versions
 of the popular UNIX/Linux shells there is support for "programmable 
 completion."  However, it would be difficult to make that work with
 arbitrary command arguments.  I suppose one could emulate it with 
 a couple of features:

 	For each (known) command: 
		Build a list of known switches

	Compile that into a database/dbm hash

	On (tab completion):
		if known:
			match command (arg[0]) and
			match/glob switch "head" on known switches
			present a list of completions
		else:
			call command with --help:
			grep for - lines that match current switch "head"

 ... That would be cool.