You don't normally need to initialize (autovivification will take care of this) but you can use the ||= construction:

for (@somelist) {
      my($foo,$bar) = split;
      my $x = somecalc($foo,$bar);
      $somehash{ $foo }->{ $bar } ||= [];
      push @{ $somehash{ $foo }->{ $bar } }, $x;
}
[download]

||= basically says 'if the left-hand value isn't defined, assign the right-hand side to it'.

Chris
M-x auto-bs-mode

    # this line won't work!!
    push @$somehash{$foo.$bar}, $x;
[download]

use perlfunc:ref to see if the value of $somehash{"$foo$bar"} is an arrayref. If it isn't, then create it.

unless ( ref ($somehash{"$foo$bar"} eq 'ARRAY' ) {
    $somehash{"$foo$bar"} = [];
}
...
[download]

HTH

Philosophy can be made out of anything. Or less -- Jerry A. Fodor

As noted later, your use of ref is not necessary because you can autovivify the hash entry and push onto it in one step:

push @{$somehash{"$foo$bar"}}, $x;

However, another point is to reconsider ever using ref. As was made clear to me this node, most uses of ref indicate that the code design should be revisited. ref is typically used to handle "special cases" and make things magically "work", which typically suggests a design flaw. Go to that link and read it. Good stuff! Once I grokked it, I rewrote the main loop in the code that I was making my point with and discovered that my new loop, without using ref, was smaller and easier to maintain.

Cheers,
Ovid

Join the Perlmonks Setiathome Group or just click on the the link and check out our stats.

push @$somehash{$foo.$bar}, $x;

# is the same as
push @{$somehash}{$foo.$bar}, $x;
[download]