[EXPERIMENT] smartmatch and given/when

[p5pRT]

Migrated from rt.perl.org#119317 (status was 'open')

Searchable as RT119317$

[p5pRT]

From @rjbs

Okay, the ~~/given/when feature, introduced in perl 5.10.0, has been discussed
plenty already. Seriously, plenty. For one of the many threads, consider​:

  http​://www.nntp.perl.org/group/perl.perl5.porters/2012/08/msg191210.html

Sprout did a bit of work on implementing things as discussed and this work then
languished, mostly because I didn't realize he was waiting on a rely, and then
because he was on hiatus. At present, it's still sort of languishing, mostly I
think, because I need to reply.

Steps to resolve this​:

  * review the work currently done (sprout/smartpatch)
  * respond to it
  * figure out what remains undone
  * test that
  * decide whether it makes sense
  * establish final acceptance criteria
  * finish up

(See also http​://markmail.org/message/eplhjfyep6yigwtl )

I will try to get this moving again as much as I can so we can bring it to a
conclusion.

--
rjbs

[p5pRT]

From @rjbs

* Ricardo SIGNES <perlbug-followup@​perl.org> [2013-08-15T21​:53​:06]

Sprout did a bit of work on implementing things as discussed and this work

In chosing the phrase "a bit of work," I did not mean to make any claim about
the effort involved. I apologize if I seemed to trivialize hard work!

--
rjbs

[p5pRT]

The RT System itself - Status changed from 'new' to 'open'

[p5pRT]

From @cpansprout

On Thu Aug 15 18​:53​:06 2013, rjbs wrote​:

Okay, the ~~/given/when feature, introduced in perl 5.10.0, has been
discussed
plenty already. Seriously, plenty. For one of the many threads,
consider​:

http​://www.nntp.perl.org/group/perl.perl5.porters/2012/08/msg191210.html

Sprout did a bit of work on implementing things as discussed and this
work then
languished, mostly because I didn't realize he was waiting on a rely,
and then
because he was on hiatus. At present, it's still sort of languishing,
mostly I
think, because I need to reply.

Steps to resolve this​:

* review the work currently done (sprout/smartpatch)
* respond to it
* figure out what remains undone
* test that
* decide whether it makes sense
* establish final acceptance criteria
* finish up

(See also http​://markmail.org/message/eplhjfyep6yigwtl )

I will try to get this moving again as much as I can so we can bring
it to a
conclusion.

IIRC, the last two issues that had not been ironed out were
• What the fallback should be​: eq or fatal
• How ‘break’ should behave​: like ‘next’ (my preference) or buggy (your
preference)

OK, that last bit was tongue-in-cheek. However, you never gave a full
explanation of how you want ‘break’ to behave. I was poking you with
questions (does it apply to while(<>)?), and getting short answers that
led me to believe you had not really thought it through.

This is all from memory. I haven’t gone to look back at the archives.
(I don’t have access to markmail.org, so I don’t know which message you
refer to above.)

Another note​: Amidst the whirling threads about how smartmatch should
behave, I pointed out that we already have a generic ‘matcher’ object in
perl​: the coderef. You thought that having something that accepts
coderefs or regexps was a good idea. More recently I have thought about
allowing qr//->($match_against_this). Regardless of smartmatch, that
might be a good idea. What do you think?

--

Father Chrysostomos

[p5pRT]

From @Hugmeir

On Sun, Aug 18, 2013 at 3​:41 AM, Father Chrysostomos via RT <
perlbug-followup@​perl.org> wrote​:

On Thu Aug 15 18​:53​:06 2013, rjbs wrote​:

Okay, the ~~/given/when feature, introduced in perl 5.10.0, has been
discussed
plenty already. Seriously, plenty. For one of the many threads,
consider​:

http​://www.nntp.perl.org/group/perl.perl5.porters/2012/08/msg191210.html

Sprout did a bit of work on implementing things as discussed and this
work then
languished, mostly because I didn't realize he was waiting on a rely,
and then
because he was on hiatus. At present, it's still sort of languishing,
mostly I
think, because I need to reply.

Steps to resolve this​:

* review the work currently done (sprout/smartpatch)
* respond to it
* figure out what remains undone
* test that
* decide whether it makes sense
* establish final acceptance criteria
* finish up

(See also http​://markmail.org/message/eplhjfyep6yigwtl )

I will try to get this moving again as much as I can so we can bring
it to a
conclusion.

IIRC, the last two issues that had not been ironed out were
• What the fallback should be​: eq or fatal
• How ‘break’ should behave​: like ‘next’ (my preference) or buggy (your
preference)

OK, that last bit was tongue-in-cheek. However, you never gave a full
explanation of how you want ‘break’ to behave. I was poking you with
questions (does it apply to while(<>)?), and getting short answers that
led me to believe you had not really thought it through.

This is all from memory. I haven’t gone to look back at the archives.
(I don’t have access to markmail.org, so I don’t know which message you
refer to above.)

Another note​: Amidst the whirling threads about how smartmatch should
behave, I pointed out that we already have a generic ‘matcher’ object in
perl​: the coderef. You thought that having something that accepts
coderefs or regexps was a good idea. More recently I have thought about
allowing qr//->($match_against_this). Regardless of smartmatch, that
might be a good idea. What do you think?

You can already implement this without having to modify the core​:

  package Regexp { use overload q[&{}] => sub { my ($re) = $_[0]; sub {
$_[0] =~ $re } }, fallback => 1 }
  say qr/f(oo)/->("foo")

So perhaps it should be tested on CPAN first. I don't see much use for it,
but YMMV*; personally, I'd prefer something like this​:

  sub Regexp​::match { return $_[1] =~ $_[0] }
  say qr/f(oo)/->match("foo")

But I guess that's what autobox is for.

[*] For the curious, yes, I've spent an inordinate amount of time reading
TV Tropes today.