The only way I see is to assign the data to a tied hash, which automatically reacts on collisions...of course all nested hashes need to be tied too.

something like

my $datastring =slurp $file;
tie %h, "CollisionHash";
eval "\%h = $datastring"
[download]

As I feared (though it makes sense)

tied_hashes do it with list assignments (they are sequential) but not with ano-hash assignment (they are precompiled, hence collisions are optimized)

FWIW

use strict;
use warnings;

use Data::Dump qw/pp/;

package HoA;
require Tie::Hash;
our @ISA = qw(Tie::StdHash);

sub STORE {
  my ($this,$key,$value) =@_;
  push @{$this->{$key}}, $value;
}


package main;

my $data_str= do { local $/;<DATA>};


tie my %h, "HoA";

my $h=\%h;


%h = eval "$data_str";
pp $h;

%h = eval "%{ { $data_str } }";
pp $h;


__DATA__
(
   alpha => 1,
   alpha => 2
)
[download]

shows

{
  # tied HoA
  alpha => [1, 2],
}
{
  # tied HoA
  alpha => [2],
}
[download]

Cheers Rolf

( addicted to the Perl Programming Language)

I mean, when I do `do $file`, Perl would read the entire data structure at once and overwrite hash keys with the last value it reads. I wish there was a way for me to traverse through the data structure (which Perl would be doing internally to be able to find key collisions) manually.

It's correct that do file means slurp + eval but your data only defines an anonymous hash on the top level.

{
  alpha => {
    beta => {
      gamma => theta,
      delta => lambda,
    },
    beta => {
      gamma => zeta,
    },
  },
},
[download]

you need to add a variable name!

Cheers Rolf

( addicted to the Perl Programming Language)