Re: Problem to inspect scalars in STASH

[educated_foo]

Yes, this is annoying, and even more so when defined(@array) and defined(%hash) become deprecated. The solution I came up with is to use defined($$name) for scalars, and defined(*{$name}{THING}) for the others. I think it works on all current Perls without polluting the stash, e.g.

main @> $Test::x = 1; @Test::y = qw(2 3); %Test::z = (a => 'b');
main @> keys %Test::
y x z
main @> join ' ', "SCALARS:", grep defined(${"Test::$_"}), keys %Test:
+:
'SCALARS: x'
main @> join ' ', "ARRAYS:", grep defined(*{"Test::$_"}{ARRAY}), keys 
+%Test::
'ARRAYS: y'
main @> join ' ', "HASHES:", grep defined(*{"Test::$_"}{HASH}), keys %
+Test::
'HASHES: z'
main @> join ' ', "SCALARS:", grep defined(${"Test::$_"}), keys %Test:
+:
'SCALARS: x'

[LanX]

Thanks for your help... :)

> The solution I came up with is to use defined($$name) for scalars,

Sorry, that's not a solution because it excludes all declared package variables which became undef in the meantime.

Think of something like  $flag = () in the middle of the code. I'm trying to fix the tab-expansion in the perldebugger, actually $f#TAB# lists all symbols starting with f, no matter which slot is used.

  DB<223> sub flo {}
 => 0
 
  DB<224> $f                      # type TAB
$f     $find  $flag  $flo

As a workaround I could include all symbols where only the scalar slot is defined, but if someone decides to have equally named sub &flo AND scalar $flo where the latter is undef, I can't tell if the scalar belongs to the code or not.

While this cases seems rare enough - most people tend to use different names for different sigils - it's not a 100% solution.

If there is no better suggestion, I will need to parse the code to reliably find all scalars.

>Yes, this is annoying,

FWIW I consider this behavior a bug.

[educated_foo]

You're right, of course, and I agree that the SCALAR auto-vivification thing is a bug. I remember discussing this with some P5P folks awhile back, and IIRC this auto-creation of the SCALAR slot was fairly well-baked into the core. Still, you think that wandering through the symbol table should be easy -- it's just a hash of GLOBs, right? -- and then you run into all this stuff. *sigh*