Bug #93335 for Method-Signatures: Wishlist for Perl 5.20 signature compatibility: document that a trailing equals sign works

Tue Feb 25 06:38:00 2014 perl [...] toby.ink - Ticket created

Subject:

Wishlist for Perl 5.20 signature compatibility: document that a trailing equals sign works

This already appears to work in Method::Signatures, but I can't seem to find where it's documented: method foo ($x=) { ... } method bar (@y=) { ... } That is, in a default, if `EXPR` is omitted, then `EXPR` the default is `()` - i.e. undef for scalars, and the empty list for non-scalars.

Wed Feb 26 23:30:36 2014 BAREFOOT [...] cpan.org - Correspondence added

Show quoted text

> This already appears to work in Method::Signatures, but I can't seem > to find where it's documented: > > method foo ($x=) { ... } > method bar (@y=) { ... } > > That is, in a default, if `EXPR` is omitted, then `EXPR` the default > is `()` - i.e. undef for scalars, and the empty list for non-scalars.

It _isn't_ documented, because, if it's working, it's working by accident. :-/ Can't say I'm that fond of it, overall ... is this official syntax from somewhere else? I don't necessarily propose that we specifically disallow it if it's working now, but I don't know that I want to bless it officially by adding POD for it either. Perhaps an outside source would sway me. :-)

Wed Feb 26 23:30:36 2014 The RT System itself - Status changed from 'new' to 'open'

Thu Feb 27 11:52:57 2014 perl [...] toby.ink - Correspondence added

I can't say I'm massively fond of it either. It has no parallel in existing Perl 5 code. You can't for example do: my $foo = ; But this is documented syntax for Perl 5.19.9. http://search.cpan.org/~tonyc/perl-5.19.9/pod/perlsub.pod#Signatures includes the following code sample: sub foo ($thing, $=) { print $thing; }

Thu Feb 27 12:50:24 2014 mschwern [...] cpan.org - Correspondence added

At first I was like "no no no, that's just turning a user mistake into syntax" and now I'm like "hey, that's a really intuitive way to declare an optional variable without needing extra syntax".

My thinking goes like this. Variables should be required by default (I hope 5.20 does this). Optional variables should have defaults, so why not make the syntax for declaring something optional to be the same as declaring a default?

PRESTO! No more need for the extra $foo? syntax. One less bit of special syntax for users to remember. And it forces the function author to think "hey, this argument should have a default".

As to the argument that there's no precedent for '$foo=' in Perl, there's a lot more precedent than '$foo?'.

My only concern is that a signature like '$foo=, $bar=' might be harder to parse... but it seems it already works so whatever.

Take this all with a grain of salt as I haven't looked at Method::Signatures, 5.20 signatures or Perl in general in months.

Mon Mar 24 00:07:09 2014 BAREFOOT [...] cpan.org - Correspondence added

Show quoted text

> I can't say I'm massively fond of it either. It has no parallel in > existing Perl 5 code. You can't for example do: > > my $foo = ; > > But this is documented syntax for Perl 5.19.9.

Yes, all properly documented and official-looking, but that doesn't keep it from being any less STUPID. :-/ Show quoted text

> At first I was like "no no no, that's just turning a user mistake into > syntax" > and now I'm like "hey, that's a really intuitive way to declare an > optional > variable without needing extra syntax". > > My thinking goes like this. Variables should be required by default (I > hope > 5.20 does this). Optional variables should have defaults, so why not > make the > syntax for declaring something optional to be the same as declaring a > default? > > PRESTO! No more need for the extra $foo? syntax. One less bit of > special syntax > for users to remember. And it forces the function author to think > "hey, this > argument should have a default".

I remain unswayed. :-/ `$foo=` just looks stupid. If you really don't like the `$foo?` syntax, you're perfectly free to write `$foo = undef`, which is a) legal, b) obvious, and c) already working in all signature implementations. The only downside is, it's (at least) 5 more characters than `?`. Show quoted text

> As to the argument that there's no precedent for '$foo=' in Perl, > there's a lot > more precedent than '$foo?'.

I disagree, since I consider Perl 6 to be Perl (even though I know that's not always a popular viewpoint). My main point is, the very first time I ever saw `$foo?` (which was in a Perl 6 something-or-other--most likely either Apocalypse or Exegesis), I knew exactly what it meant. The first time I saw `$foo=` I thought: WTF is that abomination?? I'm just sayin'. ;-> Anyhow, all we're quibbling about is whether to document it or not, and I can't realistically mount much of a defense against doing so, I suppose. But let me sulk on it for a bit longer. :-D