On Sat, Jun 13, 2020 at 1:27 PM Sean McAfee <eef...@gmail.com> wrote:
> On Sat, Jun 13, 2020 at 10:21 AM Brad Gilbert <b2gi...@gmail.com> wrote:
>
>> That was just a dumb example.
>> An incredibly dumb example.
>>
>> So what happens is that `Bool.pick` chooses The Bool values of either
>> `True` or `False`.
>> It does this at every position in the string.
>>
>> 'TrueFalse' ~~ / <{ Bool.pick }>
>>
>
> I saw this and figured that maybe a state variable could help:
>
> 'TrueFalse' ~~ /<{ $ //= Bool.pick }>/
>
> ...but it doesn't. I suppose that means the bracketed code is not
> compiled just once, like an anonymous subroutine, but is fully re-evaluated
> on every match attempt. Is that the intended behavior?
>
`state` variables act like they were declared in the next outer lexical
scope.
{ { state $v; $v++ } }
{ my $v; { $v++ } }
The next outer lexical scope that we see for `<{ }>` is the regex itself;
/ <{ $ //= Bool.pick }> /
/ <{ (state $v) //= Bool.pick }> /
That last one is almost like this:
/ :my $v; <{ $v //= Bool.pick }> /
Almost.
It really acts like there is a scope defined by the <> and a separate one
defined by the {}.
(This may be just an implementation detail that is leaking out.)
---
At any rate the right way to fix it is to actually run `Bool.pick` exactly
once at the beginning.
/ :my $to-match = ~Bool.pick; <{ $to-match }> /
/ :my $to-match = ~Bool.pick; "$to-match" /
