UPDATE: Please don't trust the benchmark, turns out, that when operating on the same hash they are sabotaging each other*

> Is (keys %hash)[$idx++] optimized to be performant, or will it create a long list of keys to be thrown away?

it's way faster than each !?!

UPDATE Sorry I misread the results, it's late. :)

The approach with (keys %hash)[$idx++] is far slower.

    
Compilation started at Fri Jun 19 02:17:48

C:/Perl_524/bin\perl.exe  d:/exp/hash_iter.pl 

*** Testing gen_idx_keys
ok 1 - 'gen_idx_keys' iterates over all entries
ok 2 - 'gen_idx_keys' cycles and iterates again

*** Testing gen_copy_hash_and_each
ok 3 - 'gen_copy_hash_and_each' iterates over all entries
ok 4 - 'gen_copy_hash_and_each' cycles and iterates again

*** Testing gen_copy_keys
ok 5 - 'gen_copy_keys' iterates over all entries
ok 6 - 'gen_copy_keys' cycles and iterates again

*** Testing gen_each
ok 7 - 'gen_each' iterates over all entries
ok 8 - 'gen_each' cycles and iterates again
1..8
                        Rate   idx_keys  copy_keys       each copy_has
+h_and_each
idx_keys              3.05/s         --      -100%      -100%         
+     -100%
copy_keys          2361328/s  77510489%         --        -6%         
+      -23%
each               2522717/s  82808095%         7%         --         
+      -17%
copy_hash_and_each 3047966/s 100049391%        29%        21%         
+        --

Compilation finished at Fri Jun 19 02:18:05
[download]