next trap

[coke]

Was just working on one of the xt/ tests, where I had

for @($thing) -> $thing {
    doit()
}

and 10 lines inside of doit, I had a

next if....

instead of a

return if

which ended up skipping in the caller's for loop. I was very confused as to why it was not going through each element.

@cfa suggested this is a trap.

[patrickbkr]

Is it verified this is not a bug?

[cfa]

No, not a bug.

Note that this isn't unique to Raku; Perl[^1] has the same behaviour:

$ raku -e 'sub ponder { next }; for (1..3) { ponder; say $_ }; say "done"'
done
$ perl -E 'sub ponder { next }; for (1..3) { ponder; say $_ }; say "done"'
done

but there, warnings can help identify a potential bug:

$ perl -E 'sub ponder { next }; for (1..3) { ponder; say $_ }; say "done"' -w
Exiting subroutine via next at -e line 1.
Exiting subroutine via next at -e line 1.
Exiting subroutine via next at -e line 1.
done

Accidentally using next, last etc. to exit a sub should probably be detected by the compiler. When the above is intended behaviour, one could opt out of the error / disable the warning (again, as in Perl):

$ perl -E 'sub ponder { no warnings "exiting"; next }; for (1..3) { ponder; say $_ }; say "done"' -w
done

For now, I'd say @coke's example is worth documenting as a trap. Alternatively, we could add some examples demonstrating the use of next etc. within subs for flow control (sub my-next { … next … } and the like).

[^1]: I know Raku isn't Perl but the comparison here is useful, I think?

[cfa]

Alternatively, we could add some examples demonstrating the use of next etc. within subs for flow control (sub my-next { … next … } and the like).

P.S. Instead of "alternatively", perhaps "additionally"?

This behaviour is probably worth documenting as part of control flow anyway.

[2colours]

No, not a bug.

Is it in the specs? If not, I'd still lean towards a fix. A Block should behave like this but not a Sub.

[cfa]

@2colours, I replied there.

[2colours]

Right, thank you... Welp, seems like we cannot get away without taking notes of it, as it appears with last in the specs.

Having said that, it still seems like low-hanging fruit from design perspective, therefore I'd keep the other issue also open.