On Fri Aug 22 13:46:45 2014, coke wrote:
> On Fri Aug 22 09:52:24 2014, coke wrote:
> > On Fri Aug 20 04:39:00 2010, masak wrote:
> > > <blixtor> perl6: sub foo { my $s; for 1..3 { $s += $_ } } ; say
> > > foo()
> > > <p6eval> rakudo e45bf6: OUTPUT«666»
> > > <blixtor> what is the rationale for this returning '666'
> > > <jnthn> We embeded the devil in Rakudo.
> > > <jnthn> Also, for in Perl 6 is just a synonym for "map"
> > > <blixtor> and it's leaking from time to time, I see ;)
> > > <jnthn> So you're doing something like
> > > <jnthn> (1..3).map: { $s += $_ }
> > > <jnthn> Which returns $s since that's the last thing you touched
> > > <jnthn> So you end up with a list with $s in it 3 times
> > > <blixtor> ahh, ok, that explains the difference to perl5, which
> > > doesn't return anything
> > > <jnthn> Yes, it's a difference from Perl 5.
> > > <masak> why doesn't that return 136, though?
> > > <masak> that's what I'd expect.
> > > <sorear> masak: because Rakudo map is rw
> > > <sorear> it's not supposed to be
> > > <sorear> in theory, take should decontainerize
> > > * masak submits rakudobug
> > > <masak> it's in there, I'm sure.
> > > <masak> but it's nice to have this example on file.
> > > <sorear> like return
> > > <masak> aye.
> > > <jnthn> erm
> > > <jnthn> what?
> > > <jnthn> We never returned or taked here
> > > <masak> true.
> > > <masak> which is why it's a different ticket, I now realize :)
> > > <sorear> jnthn: map is implemented using take
> > > <jnthn> sorear: No.
> > > <jnthn> sorear: Well, it *may* be.
> > > <jnthn> sorear: But it certainly doesn't have to be.
> > > <jnthn> (That is, nothing in the spec says it needs to be.)
> > > <jnthn> morning, takadonet
> > > <jnthn> Anyway, it maybe should be decontainerizing somewhere here
> > > too
> > > <sorear> jnthn: well, in any case, the same underlying LTA is
> > > causing
> > > both
> > > <jnthn> I'm just not sure where.
> > > <jnthn> Anyway, a ticket does no harm and makes sure we don't
> > > forget
> > > the issue.
> >
> > Behavior has changed:
> >
> > 12:50 < [Coke]> perl6: sub foo { my $s; for 1..3 { $s += $_ } } ; say
> > foo()
> > 12:50 <+camelia> rakudo-{parrot,jvm,moar} 60cd9d: OUTPUT«Nil»
>
>
> 15:57 < masak> [Coke]:
> https://rt.perl.org/Ticket/Display.html?id=77334 is not
> valid any more, at least not with that kind of for
> loop.
> 15:57 < masak> m: sub foo { my $s; ($s += $_ for 1..3) } ; say foo()
> 15:57 <+camelia> rakudo-moar 5bd2d4: OUTPUT«6 6 6»
> 15:58 < masak> [Coke]: need to do it that way instead :)
>
> So, no change, really. masak++
TimToady++ has ruled that `for` should not decontainerize, on the basis that
it's easy enough to make it not do so (such as with the <> postfix) but there's
not an easy way to avoid the decontainerization if you don't want it. Also,
there's usually a better way to write these things. Discussion:
http://irclog.perlgeek.de/perl6/2015-10-27#i_11440206
Tests covering this ruling in S04-statement-modifiers/for.t and
S04-statements/for.t.
