My first pass assessment would be that maps block doesn't qualify as a sub in the normal sense:

print map{ print wantarray(); (1,2,3) } 1;;
Use of uninitialized value in print at (eval 19) line 1, <STDIN> line 
+11.

1 2 3
[download]

It was obviously called in a list context, else the return would be different:

print scalar map{ print wantarray(); (1,2,3) } 1;;
Use of uninitialized value in print at (eval 20) line 1, <STDIN> line 
+12.

3
[download]

This implies that the block is not a true sub as far as wantarray is concerned. It is just a block as with:

print do{ print wantarray(); (1,2,3) };;
Use of uninitialized value in print at (eval 23) line 1, <STDIN> line 
+15.

1 2 3
[download]

Curiously, if you define your own sub which takes a block, wantarray treats it as a normal routine:

$ perl -wE 'sub f(&) { shift->() }; say f { wantarray }'
1
[download]

I wonder if this inconsistency is worth a bug report.

Perl 6 - second systems done right

I suspect that as a built-in, map avails itself of more optimisations than it would be safe to apply to a user-written function.

I know from experience that if you implement a callback interface in XS using the full belts'n'braces approach described in perlcall, then the result is far less efficient than the carefully optimised callback interface used by List::Util::reduce().

The downside of trying to match that level of optimisation with your own XS code is that it is very easy to create the situation whereby the callback leaks memory if it contains closures. (The early versions of reduce() suffered from the same problems.)

I think it would be very difficult, if not impossible to compile user-written, runtime-compiled subs taking callbacks, to be as efficient as map. And to 'fix the inconsistency' would essentially require pessimising the implementation of map

So no, I don't think it is worth a bug report. At most, a documentation change noting the difference might be useful.

---

With the rise and rise of 'Social' network sites: 'Computers are making people easier to use everyday'

Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error.

"Science is about questioning the status quo. Questioning authority".

In the absence of evidence, opinion is indistinguishable from prejudice.

Aside from being documented ("Evaluates BLOCK or EXPR in list context"), it's very obviously evaluated in list context.

$ perl -E'say map { scalar( $_, $_ ) } 4;'
4

$ perl -E'say map { $_, $_ } 4;'
44
[download]

Even when map isn't.

$ perl -E'say scalar map { scalar( $_, $_ ) } 4;'
1

$ perl -E'say scalar map { $_, $_ } 4;'
2
[download]

wantarray "returns true if the context of the currently executing subroutine or eval is looking for a list value", emphasis mine.

$ perl -E'@a = sub { map { say wantarray // "undef" || 0 } 1; 1 }->()'
1

$ perl -E'$a = sub { map { say wantarray // "undef" || 0 } 1; 1 }->()'
0

$ perl -E'     sub { map { say wantarray // "undef" || 0 } 1; 1 }->()'
undef
[download]

map itself is in void context in each of those. The block is evaluated in list context in each of those. Only the context in which the subroutine is evaluated is changing. This is what wantarray reports.

wantarray works as documented in map.

wantarray's documentation says it all really. wantarray is sensitive to the "subroutine or eval" context and is "unspecified in the top level of a file". map isn't a sub or eval so doesn't provide any new context for wantarray.