«

Also sprach Anno Siegel:

> John W. Krahn <> wrote in comp.lang.perl.misc:
>> Anno Siegel wrote:
>> > Gunnar Hjalmarsson <> wrote in comp.lang.perl.misc:
>
> [using index() instead of s///]
>
>> > my $i = length $thisPage;
>> > substr( $thisPage, $i, length $find) = $replace while
>> > ( $i = rindex $thisPage, $find, $i) >= 0;
>> >
>> > That way only the unchanged part of the string is ever searched.
>>
>> Also, using the four argument substr() should be faster.
>
> How so? I never heard of that.

John is right according to a benchmark:

#!/usr/bin/perl -w

use strict;
use Benchmark qw/cmpthese/;

my $string = "0" x 100;

cmpthese(-2, {
arg4 => sub {
substr $string, rand length $string, 1, "0";
},
arg3 => sub {
substr($string, rand length $string, 1) = "0";
},
});
__END__
Rate arg3 arg4
arg3 512250/s -- -43%
arg4 903253/s 76% --

The reason for arg4 being faster is the fact that perl needs to attach
magic to the return value of the 3-argument substr(). In the 4-argument
case this is not the case.

> I use "=" with substr() assignments because it reads better. Four argument
> substr is for when I need the old value of the substring too.

In most cases though the argument of readability supersedes speed, so
now you're right.

Tassilo
--
use bigint;
$n=71423350343770280161397026330337371139054411854 220053437565440;
$m=-8,;;$_=$n&(0xff)<<$m,,$_>>=$m,,print+chr,,while(($ m+=<=200);

Tassilo v. Parseval <> wrote in comp.lang.perl.misc:
> Also sprach Anno Siegel:
> > John W. Krahn <> wrote in comp.lang.perl.misc:

> >> Also, using the four argument substr() should be faster.
> >
> > How so? I never heard of that.
>
> John is right according to a benchmark:

Yup. I ran one too. The difference is significant.

> The reason for arg4 being faster is the fact that perl needs to attach
> magic to the return value of the 3-argument substr(). In the 4-argument
> case this is not the case.

Ah... the return value from 4-arg substr is indeed not magic:

my $x = 'aaaaabbbbbcccc';
my $ref = \ substr( $x, 5, 5, 'ZZZZZ');

$$ref = 'XXXXX'; # has no effect on $x
print "$x\n"; # aaaaaZZZZZcccc

After removing the fourth argument from substr() $x changes to
"aaaaaXXXXXcccc".

Live and learn.

Anno

Tassilo v. Parseval <> wrote in comp.lang.perl.misc:
> Also sprach Anno Siegel:
> > John W. Krahn <> wrote in comp.lang.perl.misc:

> >> Also, using the four argument substr() should be faster.
> >
> > How so? I never heard of that.
>
> John is right according to a benchmark:

Yup. I ran one too. The difference is significant.

> The reason for arg4 being faster is the fact that perl needs to attach
> magic to the return value of the 3-argument substr(). In the 4-argument
> case this is not the case.

Ah... the return value from 4-arg substr is indeed not magic:

my $x = 'aaaaabbbbbcccc';
my $ref = \ substr( $x, 5, 5, 'ZZZZZ');

$$ref = 'XXXXX'; # has no effect on $x
print "$x\n"; # aaaaaZZZZZcccc

After removing the fourth argument from substr(), $x changes to
"aaaaaXXXXXcccc".

Live and learn.

Anno

Tassilo v. Parseval <> wrote in comp.lang.perl.misc:
> Also sprach Anno Siegel:
> > John W. Krahn <> wrote in comp.lang.perl.misc:

> >> Also, using the four argument substr() should be faster.
> >
> > How so? I never heard of that.
>
> John is right according to a benchmark:

Yup. I ran one too. The difference is significant.

> The reason for arg4 being faster is the fact that perl needs to attach
> magic to the return value of the 3-argument substr(). In the 4-argument
> case this is not the case.

Ah... the return value from 4-arg substr is indeed not magic:

my $x = 'aaaaabbbbbcccc';
my $ref = \ substr( $x, 5, 5, 'ZZZZZ');

$$ref = 'XXXXX'; # has no effect on $x
print "$x\n"; # aaaaaZZZZZcccc

After removing the fourth argument from substr(), $x changes to
"aaaaaXXXXXcccc".

Live and learn.

Anno

Anno Siegel wrote:

>You're right, though I wouldn't bother with storing the length of
>anything. Working backwards runs smoother:
>
> my $i = length $thisPage;
> substr( $thisPage, $i, length $find) = $replace while
> ( $i = rindex $thisPage, $find, $i) >= 0;
>
>That way only the unchanged part of the string is ever searched.

I'm sure you're aware that the behaviour of this version is not the same
as the original one for overlapping matches. Try replacing "papa" in
"papapaya", for example.

--
Bart.

Bart Lateur <> wrote in comp.lang.perl.misc:
> Anno Siegel wrote:
>
> >You're right, though I wouldn't bother with storing the length of
> >anything. Working backwards runs smoother:
> >
> > my $i = length $thisPage;
> > substr( $thisPage, $i, length $find) = $replace while
> > ( $i = rindex $thisPage, $find, $i) >= 0;
> >
> >That way only the unchanged part of the string is ever searched.
>
> I'm sure you're aware that the behaviour of this version is not the same
> as the original one for overlapping matches. Try replacing "papa" in
> "papapaya", for example.

Frankly, no, I didn't notice. Thanks for pointing it out.

Anno