Re: "switch$ia" (Russ: Useless Benchmarking again)

[Russ]

I have to assume someone has done this before, but I didn't see it, so...

Benchmarking interpolated strings vs. appended strings:

use Benchmark;

timethese( 1000000, {
  interp => sub {my $R = 1; my $S = "Test${R}a"},
  append => sub {my $R = 1; my $S = 'Test' . $R . 'a'}
});

results in:

Benchmark: timing 1000000 iterations of interp, manual...
  interp:  4 wallclock secs ( 3.68 usr +  0.05 sys =  3.73 CPU) @ 2680
+96.51/s
  manual:  3 wallclock secs ( 3.79 usr + -0.13 sys =  3.66 CPU) @ 2732
+24.04/s

Slightly faster to append strings together than to interpolate them, but statistically, it's a wash...

Russ
Brainbench 'Most Valuable Professional' for Perl

[echo]

Someone please correct me if I'm wrong, ISTR interpolated strings are converted to concatenation at compile time?

[Fletch]

You are correct, at least according to the B::Deparse documentation. Interpolation is just syntactic sugar for concatenation. You can see it for yourself:

$ perl -MO=Deparse,-p,-q -e '$foo = "BAR"; $bar = "a${foo}c"; $baz="a"
+.$foo."c"'                        
($foo = 'BAR');
($bar = (('a' . $foo) . 'c'));
($baz = (('a' . $foo) . 'c'));

Update: Added concatenation to the example for comparison. Also, the output using B::Terse to dump the opcode tree may be more convincing.

[Russ]

<sheepish grin>
hehehe You're probably right.

So, in the spirit of further useless benchmarking:

use Benchmark;

timethese( 10000000, {
  5 => sub {my $R = 5;},
  6 => sub {my $R = 6;}
});

results in:

    5:  6 wallclock secs ( 6.61 usr + -0.03 sys =  6.58 CPU) @ 1519756
+.84/s
    6:  6 wallclock secs ( 5.89 usr +  0.08 sys =  5.97 CPU) @ 1675041
+.88/s

As you can see, it is slightly faster to assign a 6 than a 5, so I will use 6's more often from now on... :-)
</sheepish grin>

Russ
Brainbench 'Most Valuable Professional' for Perl