fix #37393: interpolation in for outer

[simeonschaub]

This also simplifies the logic a bit. This does now throw an error during lowering for for outer +i = 1:3; end, which was technically valid before, but I don't think that should be a problem?
fixes #37393

[JeffBezanson]

The simplification is nice but a bit too good to be true --- we allowed for outer in x so that outer still generally works as a variable (not stealing it as a keyword).

[simeonschaub]

Oh, you are right! I misunderstood what parse-pipe< does. I think we should be able to rewrite most of parse-iteration-spec using parse-eq* though. Do you agree with that approach?

[simeonschaub]

I also can't quite figure out what this line is for:

julia/src/julia-parser.scm

Line 1668 in 36473b1

((and (eq? lhs ':) (closing-token? t))

Is this even needed anymore?

[JeffBezanson]

I think the : line is for the discontinued comprehension syntax:

julia> [i for i = 1:10, :]
ERROR: syntax: comprehension syntax with `:` ranges has been removed around REPL[4]:1

parse-eq* might work, hard to say without trying it :)

[simeonschaub]

parse-eq* might work, hard to say without trying it :)

It almost works. Unfortunately, the current parse-iteration-spec allows

[i for i
in 1:3]

as valid syntax, which I guess was just an oversight, but since it's actually used in

julia/base/compiler/ssair/inlining.jl

Line 1326 in f902584

for argt

, I guess we still need to allow that?

[simeonschaub]

Ok, this should now be ready to review. It keeps the (somewhat questionable) current behavior regarding newlines, while still fixing #37393.

[simeonschaub]

@JeffBezanson Would you mind taking another look at this?

[simeonschaub]

I am going to rebase once #37247 is merged