[+] does not behave like the other [binop]s on single Iterables

[markjreed]

The Problem

Given a single scalarized Iterable, [+] reduces its elements.

Expected Behavior

[+] should behave like the other reduced arithmetic operators.

Actual Behavior

[+] appears to decontainerize a single Iterable, reducing across its elements. This seems to be because it uses Array.sum when available, instead of just doing an actual reduce with +.

Steps to Reproduce

say [+] $(1,2,3)
# result: 6
# expected result: 3

Note that replacing + with any of the other three arithmetic ops here does produce 3.

Environment

• Operating system: macOS 13.0
• Compiler version (perl6 -v or raku -v):
  Welcome to Rakudo™ v2022.12.
  Implementing the Raku® Programming Language v6.d.
  Built on MoarVM version 2022.12.

[lizmat]

Thanks for the report!

[thundergnat]

Actually, due to single argument, I would expect this to produce 6. A stronger case could be made that say [+] $(1,2,3), (note trailing comma) should return 3.

Arguably [-] $(1,2,3), [*] $(1,2,3) and [/] $(1,2,3) are misbehaving and acting like [-] $(1,2,3),, [*] $(1,2,3), and [/] $(1,2,3),

[lizmat]

Yeah, but they have been exposing this behaviour for a long time already. So fixing would probably need to be done at a language level.

[markjreed]

@thundergnat I agree that it would at least be more useful for the other operators to behave like [+] in this regard. It should be consistent either way, though.

@lizmat So for those of us keeping score at home, you tried to fix this, it broke other stuff, so the fix was reverted. Right?

[lizmat]

Yes, the fix was reverted, in the sense that [+] no longer uses sum internally, so it now works just like the other meta infix ops, and so should have the same semantics. So if you're using [+], the fix is still there.

[thundergnat]

Well, I am mildly of the opinion that the meta reduce binops should follow the single argument rule, for consistency if nothing else, but also think that it is a relatively minor issue in general and that core devs time may be better spent on more critical things.

And, like lizmat said, they have been that way for some time; changing them should at least be discussed in problem-solving, with maybe some input from jnthn or even (wistfully) Timtoady.