OT: clearing a hash (was: TIMTOWTDI doesn't mean invent an outlandish approach (usually))

[Aristotle]

Of the first set of answers, one doesn't end up with an empty hash, and one is inefficient. Of the second set of answers, only one is correct, the other three are not guaranteed to work.

I must be missing something obvious, because I can't quite see why they wouldn't work. But since you say it applies for three of them, it could also be true for the inefficient answer from the first set.(Which I posted without actually looking at the question, in reply to someone posting one of the ones in the second set.) Is this so?

I'm sure I'll bang my head on the wall once you spill the beans, but what is it that I can't see?

[Abigail-II]

The '%hash = undef' is incorrect:

$ perl -wle '%hash = undef; print scalar %hash'
Odd number of elements in hash assignment at -e line 1.
Use of uninitialized value in list assignment at -e line 1.
1/8

It'll create a one element hash, whose key is the empty string, with an undefined value.

The last three solutions modify the hash while iterating over the keys, which should be a no-no. However, I just read that in the documentation of 'each' that deleting the item most recently returned by 'each()' is safe; this feature seems to have been added (or documented) in 5.6.1.

Abigail

[Aristotle]

Ah, so that's why I was puzzled - the hint on each was there the first I wanted to know about deleting keys while iterating. But you said three of the following will not necessarily work:
1. map { delete $hash{$_} } keys %hash;
2. while (my ($k,$v) = each %hash) { delete $hash{$k} };
3. for (keys %hash) { delete $hash{$_} };
4. delete $hash{$_} for keys %hash;
With the each solution being valid that's one down, two to go, and now I'm more confused than before.. of the remaining three solutions, the for loops are identical except for syntactical details. So they must either both be valid or both invalid; then the map version must be valid and the for ones not. But I can't see any reason to conclude that any of them (all four) are not guaranteed to work.

[Abigail-II]

Number 1) is safe because in this case, a list of keys will be produced before the deletes are called. For 3) and 4), I though that 'for (keys %hash)' was special cased to iterate over the hash instead of producing a list (just like 'for (@array)' is special cased), but I can't find it in the documentation right now. Anyway, regardless whether it's special cased or not, 3) and 4) are fine as well, due to the exception of modifying a hash while iterating over it by deleting the last element returned by 'each' (since 'keys' and 'values' uses the same iterator).

Abigail