Assinging to $_ without localizing it is dangerous.
(local $_ = $before) =~ s/^\s+|\s+$//g;
or
(my $after = $before) =~ s/^\s+|\s+$//g;
should be used in just almost all cicumstances.
| [reply] [d/l] [select] |
I think the alternation in the s/^\s+|\s+$// version causes significant time costs in large applications. I often work with tab-delimited files containing hundreds of thousands of lines. If I'm tab-splitting these lines and then trimming each one, I'm going to pick the double-regex approach each time.
I wrote some quick code that benchmarked the double-regex vs single-regex approach against three strings.
use Benchmark;
my @words = ('trim_unneeded',' front trim only','rear trim only ','
+ both side trim ');
for my $word (@words){
print "Benchmarking $word...\n\n";
timethese(1_000_000, {double => sub{ $word =~ s/^\s+//; $word =~ s
+/\s+$//; },
single => sub{ $word =~ s/^\s+|\s+$//; }})
}
The code was run on a Celeron D 2.8 GHz machine running XP with the following results:
'trim_unneeded'
Single Regex: 0.45 seconds
Double Regex: 2.27 seconds
' front trim only'
Single Regex: 0.67 seconds
Double Regex: 2.66 seconds
'rear trim only '
Single Regex: 0.67 seconds
Double Regex: 2.45 seconds
' both side trim '
Single Regex: 0.66 seconds
Double Regex: 2.44 seconds
That's after only 1,000,000 trims. In a 800,000 line file with 50 columns per line, we're talking about 40,000,000 trims. Assuming a linear scale, that means I give up about a minute of processing time per file per run. That's far less than the time it would have taken me to type two regexes.
Admittedly, it's a small optimization, and only valid for those who are processing files on the scale that I do, but for most people who end up typing the 'trim regex' often enough to complain about it on perlmonks, it probably applies. | [reply] [d/l] |
| [reply] |
($_ = $before) =~ s/^\s+|\s+$//g;
Just as a minor nitpick I'd localise $_, but then I wouldn't bind with =~, I'd just
local $_ = $before;
s/^\s+|\s+$//g;
or else
(my $after = $before) =~ s/^\s+|\s+$//g;
| [reply] [d/l] [select] |