RFC: Allow subcommands in MAIN

[Juerd]

The combination of these two patches will allow subcommands like in "git" or "apt".

Note that f8227bb deliberately goes against spec (S06), which says:

As usual, switches are assumed to be first, and everything after the first non-switch, or any switches after a --, are treated as positionals or go into the slurpy array (even if they look like switches)

S06 is demonstrably wrong in that this would be the usual thing, since all these standard tools apparently take switches after positionals:

1;0 juerd@cxie:~$ wc $(ls $(which perl6 -a) -1) --chars
26876 /home/juerd/.rakudobrew/bin/perl6

Requiring switches at the beginning is common with BSD-style command line parsing, but that style doesn't have --long-names or --, both of which are already supported by Perl 6.

This patch set has not yet been thoroughly tested. Please discuss and test :-)

[Juerd]

Demonstration:

1;0 juerd@cxie:~$ cat gitlike.pl 
multi sub MAIN ("add", *@pathspec, Bool :v($verbose), Bool :n($dry-run), Bool :i($interactive)) { die 1 }
multi sub MAIN ("commit", *@file, Bool :$e, Bool :$dry-run, Str :$author) { die 2 }
multi sub MAIN ("diff", Str $path1, Str $path2, Bool :$no-index) { die 3 }

1;0 juerd@cxie:~$ perl6 gitlike.pl 
Usage:
  gitlike.pl add [-v] [-n] [-i] [<pathspec> ...] 
  gitlike.pl commit [-e] [--dry-run] [--author=<Str>] [<file> ...] 
  gitlike.pl diff [--no-index] <path1> <path2> 
1;2 juerd@cxie:~$ perl6 gitlike.pl add
1
  in sub MAIN at gitlike.pl line 1
  in block <unit> at gitlike.pl line 3

1;1 juerd@cxie:~$ perl6 gitlike.pl diff foo bar
3
  in sub MAIN at gitlike.pl line 3
  in block <unit> at gitlike.pl line 3

[vendethiel]

Yes yes yes! That's a very good idea 👍 .

[lizmat]

I'm afraid merging this will have to wait until we have a decision on how to technically handle versioning

[Juerd]

I understand that merging might have to wait on that decision. In the meantime, it remains important that we discuss whether we WANT to merge it, since it would require a spec change.

[b2gills]

As far as I can tell the only person who has been against this sort of change at all is Tim Toady.
(+1 from me)

[MadcapJake]

Can this be considered for inclusion in 2016.03 under v6.d.a?

[jnthn]

TimToady would need to rule on the language design issue. I suspect it
breaks back-compat, so would need to go in the 6.d setting.

On Mon, Feb 22, 2016 at 6:22 PM, Jake Russo notifications@github.com
wrote:

Can this be considered for inclusion in 2016.03 under v6.d.a?

—
Reply to this email directly or view it on GitHub
#688 (comment).

[AlexDaniel]

It may be a bit too late but: YES!!

[niner]

What if there should be a distinction between global and subcommand specific options as in:
zypper -v search --match-exact Mesa

[LLFourn]

👍 Would be cool if we could also

sub MAIN($ where 'foo'|'bar', :$baz) { ... } # PS why can't we just write sub MAIN('foo'|'bar',...)
Usage:
  gitlike.pl [ foo | bar ] [-baz]

@niner I cannot think of a way to describe main options and subcommand options with signatures

[LLFourn]

Idea: what about for global switches we introspect the dispatcher's signature:

# to use nine's example
proto MAIN(:$v #={verbose output},|) { $*VERBOSE = True if $v; {*} }
multi MAIN("search",Bool :$match-exact,Str $target) { ... }

Then the help could look like:

Options: 
  -v     verbose output
Usage:
  zypper [OPTIONS] search <target> [--match-exact]

What I'm suggesting is all cosmetic so I think @niner's concern above can be addressed after this is merged.

[Juerd]

Can someone explain what the next steps are in the process of reaching a decision?

[zoffixznet]

Can someone explain what the next steps are in the process of reaching a decision?

Does this PR break any of the current tests?

If yes,

Write your proposal into https://github.com/perl6/specs/blob/master/v6d.pod , listing yourself as stakeholder. The change will be considered for inclusion in the next language release. Note that that won't happen in a near future (I saw dates named anywhere for summer 2017 to Diwalli 2017, which is sometime in October).

If no tests break,

Then hunt down TimToady and ask him to make a decision on whether this language design change is OK. If he agrees, fix merge conflicts, and ping one of the people who can merge to merge.

Hope that helps.

[Juerd]

Basically fixed in af14584 and 0d1be77 as discussed in #2246.