Parseable repr for undef

[nickrobinson251]

• attempt to resolve undef does not have parseable repr #33204
• WIP becuase not sure which tests will need updating
• Allows us to keep the verbose printing in REPL (i've not changed this text, although we could if others see a benefit to that too) [EDIT: updated below]

julia> undef
array initializer with undefined values

• But gives us a parseable representation if requested (with show/repr) -- which is the motivation for the change

julia> repr(undef)
"UndefInitializer()"

• To me (others may disagree), a bonus is we get more familiar Julia syntax in the first case here, which helps to better distinguishes the two cases:

julia> fill(undef, 2)  # thing you probably don't want - Vector of Initializers
2-element Array{UndefInitializer,1}:
 UndefInitializer()
 UndefInitializer()

julia> Vector(undef, 2)  # thing you probably wanted - Vector of undefined values
2-element Array{Any,1}:
 #undef
 #undef

[nickrobinson251]

I'd appreciate being advised on:

• do the tests here seem reasonable? can/should I test differently?
• does this need news? (fwiw, no tests needed updating, but it's definitely public facing)

[JeffBezanson]

Oops, just commented on exactly this on your other PR. Sorry, a bit behind on PRs and issues :)

[JeffBezanson]

Looks good to me. I think the only question is whether to output UndefInitializer() (which is perfectly clear, and sufficiently scary 😄 ) or undef, which is how it's usually entered.

[nickrobinson251]

I think the only question is whether to output UndefInitializer() (which is perfectly clear, and sufficiently scary 😄 ) or undef, which is how it's usually entered.

I slightly prefer UndefInitializer() over undef mostly because it is visually more different to #undef

and secondly because it is clearer (in the fill example) that these are "Initializers" (having the canonical "create an instance of a type" syntax) not "undefined values".

Also this seemed to be prefered by folks at #33204 (comment).

...but i do not feel strongly either way.

[JeffBezanson]

Ok, I'm convinced :)

[nickrobinson251]

This PR now results in:

julia> undef
UndefInitializer(): array initializer with undefined values

julia> repr(undef)
"UndefInitializer()"

julia> fill(undef, 2)
2-element Array{UndefInitializer,1}:
 UndefInitializer()
 UndefInitializer()

julia> Vector(undef, 2)
2-element Array{Any,1}:
 #undef
 #undef

[StefanKarpinski]

Should probably have a NEWS entry.

[nickrobinson251]

Should probably have a NEWS entry.

Ah, sorry! Since this was already merged, there's a separate PR to add news: #33254