Let s/// return according to spec. WAS: Give the result of s/// a Bool mixin.

[peschwa]

This refers to RT #82108. The performance and semantic concers
raised by jnthn++ obviously still exist.

For reference: https://rt.perl.org/Public/Bug/Display.html?id=82108

[moritz]

I'm less than thrilled by mixing Bool into basic types in the setting.

It might "do what I mean" in this particular context, but the result is normally a string that can be passed to anywhere else, and that anywhere else might be very surprised by strings not boolifying in the normal way.

The problem must really be solved at the spec level first, not monkey-patched in rakudo.

(IMHO the same goes for tr///, but I kept quiet there, hoping it would be just a very localized thing; but I don't want the mixing-in theme to spread).

[peschwa]

@moritz: Right, mixins in the setting aren't great. That's not (anymore) what this PR implements, though, even though I clearly forgot to mention that it isn't. Sorry about the confusion.
The spec has been update here, which this PR implements now. I force-pushed the (as of the spec-commit correct and clarified) update into the PR branch, since the ticket is still the correct one.

[peschwa]

There has been a bit more discussion here, which coupled with my own re-reading of S05 has resulted in the latest changes.
I'm rather sure this is not yet completely to spec, and thus am still working on it.
Here are notable spectest-results of the current state of the branch merged against nom, running current spectest.

[peschwa]

I would consider this branch "to spec", as it is now.
The most glaring change from the current implementation in nom is, that m:g// does not automatically return a List, and thus further processing of the return value has to explicitely force list context, e.g.:

    # previously:
    for m:g/./ { .say } # prints every single character
    # as of this branch:
    for m:g/./ { .say } # prints the whole match once, as per S05 'Capturing from repeated Matches'
    for @(m:g/./) { .say } # prints every single character

Because this might break a few things in a few places (outside of Rakudo itself, I did take care of that, to the best of my knowledge) there should maybe be some kind of deprecation cycle before this becomes mainline. Obviously some tests broke as well, I have created a branch in perl6/roast locally, which I can push. It mostly contains fixes regarding the difference mentioned above.

Edit: Here is the branch mentioned above.

[peschwa]

There's been a bit more discussion on this around here, which suggests merging of this PR should be delayed until after the list refactor pmichaud++ has taken upon himself, mainly to conserve the for m:g/$pat/ { ... } idiom.

[peschwa]

This PR is likely obsolete, as infix:<~~> has been changed to match Lists as truthy on its RHS.