On Aug 3, 7:10 am, [email protected] ("Chas. Owens") wrote:
> On Tue, Aug 3, 2010 at 09:47, Chas. Owens <[email protected]> wrote:
> > On Tue, Aug 3, 2010 at 08:44, Shawn H Corey <[email protected]> wrote:
> >> On 10-08-03 06:43 AM, Rob Coops wrote:
>
> >>> Third you could of course when you are printing the values from the array
> >>> add the linefeeds:
> >>> print join("\n", @myarray);
> >>> or
> >>> foreach my $value ( @myarray ) {
> >>> print $value . "\n";
>
> >> When printing, use a list; it's faster.
>
> >> print $value, "\n";
> > snip
>
> > I hate it when some makes a blanket statement of "it's faster" without
> > providing a benchmark or a reason. In the simple case, the comma is
> > slower than the period and interpolation is slower still.
>
> Whoops, accidentally sent too early. My point was that even though
> concatenation is faster than pass a list or an interpolated string
> (contrary to the claim made earlier), it isn't really that much
> faster. Especially once you factor in the cost of writing to disk
> (which the benchmark avoids to prevent noise). Now, you might make
> the argument that the comma is clearer than the concatenation, but
> that point is debatable. For one thing, these two statements are not
> identical:
>
> print "foo", "\n";
> print "foo" . "\n";
>
> The first will be affected by the value of $, and the second won't.
> The second will also benefit from compile time constant folding:
>
> perl -MO=Deparse -e 'print "foo" . "\n"'
> print "foo\n";
>
> perl -MO=Deparse -e 'print "foo", "\n"'
> print 'foo', "\n";
>
> Personally, I use string interpolation even though it is the slowest
> of the three methods. I find string interpolation to be superior to
> passing a list because it is not affected by $, and more readable than
> having tiny periods floating around. It also requires fewer
> characters on average (you usually have at least one set of double
> quotes in the expression already).
I agree totally.
>
> To sum up: if you are looking for something like this to speed up your
> program than you are well and truly hosed. Choose a method that looks
> right to your team and has the right behavior. Profile your code to
> find spots were you are slow and optimize those spots by changing the
> algorithm or moving code down a level to C (via XS or Inline::C), not
> by relying on folk-wisdom micro-optimizations.
>
> One item:
> Rate string comma period
> string 3588672/s -- -3% -11%
> comma 3714590/s 4% -- -7%
> period 4014817/s 12% 8% --
>
> #!/usr/bin/perl
>
> use strict;
> use warnings;
>
> use Benchmark;
>
> open my $bit_bucket, ">", "/dev/null"
> or die "$!";
>
> my $s = "foo";
>
> my %subs = (
> string => sub { print $bit_bucket "$s\n" },
> comma => sub { print $bit_bucket $s, "\n" },
> period => sub { print $bit_bucket $s . "\n" },
> );
>
> for my $sub (keys %subs) {
> print "$sub: ", $subs{$sub}(), "\n";
>
> }
>
> Benchmark::cmpthese -1, \%subs;
>
> Ten items:
> Rate string comma period
> string 1052183/s -- -1% -2%
> comma 1061926/s 1% -- -1%
> period 1071850/s 2% 1% --
> ...
There was a recent thread about the slowness of
concatenation on Win32 (can't remember, maybe
a perl issue though).
Adding a couple of small tweaks for Win32 and
re-running seems to confirm:
use File::Spec qw(devnull);
open my $bit_bucket, ">", File::Spec->devnull()
or die "$!";
...
Result:
Rate period string comma
period 453142/s -- -4% -6%
string 469903/s 4% -- -2%
comma 481771/s 6% 3% --
--
Charles DeRykus
--
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
http://learn.perl.org/
