Re^3: Merge hash into another hash

For simple hashes, what's wrong with just:

my %target = (
  #initialize
); 

%target = (%target, %source)
[download]

? (with the issues of complex cases, source checking, and dereferencing aside (but perhaps fixable))

jason.

Comment on Re^3: Merge hash into another hash Download Code

Replies are listed 'Best First'.
Re^4: Merge hash into another hash by linuxer (Curate) on Jul 17, 2009 at 09:13 UTC
Hi, let me try to explain with my limited knowledge of perl's internal behaviour ;o) Your solution converts `%target` and `%source` into flat lists, combines it to one list and assigns that list to `%target`... Grandfather's solution only inserts the key-value pairs from `%source` into `%target`, overwriting already existing pairs. This is less expensive than converting into lists and reassigning these lists. I tried a benchmark which shows that Grandfather's solution is faster than yours. I have to admit, that - until this post - I preferred and used the way you presented (mainly because of readability), but now I have to reconsider ;o) #!/usr/bin/perl # vim: set ts=4 sw=4 et sta: use strict; use warnings; use Benchmark qw( cmpthese ); my %target; my %source; ## prefill @target{'aa'..'zz' } = 'aa' .. 'zz'; @source{'ww'..'yy', 1..100} = ( 'ww'..'yy', 1..100 ); cmpthese( -1, { grandfather => sub { my %t = %target; @t{keys %source} = values %source; }, sleepyjay => sub { my %t = %target; %t = ( %t, %source ); }, }); __END__ Rate sleepyjay grandfather sleepyjay 598/s -- -62% grandfather 1585/s 165% -- Rate sleepyjay grandfather sleepyjay 649/s -- -62% grandfather 1706/s 163% -- [download]	[reply] [d/l] [select]

Replies are listed 'Best First'.

Re^4: Merge hash into another hash
by linuxer (Curate) on Jul 17, 2009 at 09:13 UTC

Hi, let me try to explain with my limited knowledge of perl's internal behaviour ;o)

Your solution converts %target and %source into flat lists, combines it to one list and assigns that list to %target...

Grandfather's solution only inserts the key-value pairs from %source into %target, overwriting already existing pairs.

This is less expensive than converting into lists and reassigning these lists.

I tried a benchmark which shows that Grandfather's solution is faster than yours.

I have to admit, that - until this post - I preferred and used the way you presented (mainly because of readability), but now I have to reconsider ;o)

#!/usr/bin/perl
# vim: set ts=4 sw=4 et sta:
use strict;
use warnings;
use Benchmark qw( cmpthese );

my %target;
my %source;

## prefill
@target{'aa'..'zz' } = 'aa' .. 'zz';
@source{'ww'..'yy', 1..100}     = ( 'ww'..'yy', 1..100 );

cmpthese( -1, {
   grandfather => sub {
        my %t = %target;
        @t{keys %source} = values %source;
    },
   sleepyjay   => sub {
       my %t = %target;
       %t = ( %t, %source );
    },
});

__END__
              Rate   sleepyjay grandfather
sleepyjay    598/s          --        -62%
grandfather 1585/s        165%          --

              Rate   sleepyjay grandfather
sleepyjay    649/s          --        -62%
grandfather 1706/s        163%          --
[download]

[reply]
[d/l]
[select]