遍历 Perl 数组的最佳方法

哪种实现(在速度和内存使用方面)最适合迭代 Perl 数组？还有更好的办法吗？(不需要保留 @Array)。

实施1

foreach (@Array)
{
SubRoutine($_);
}

实施2

while($Element=shift(@Array))
{
SubRoutine($Element);
}

实施3

while(scalar(@Array) !=0)
{
$Element=shift(@Array);
SubRoutine($Element);
}

实施4

for my $i (0 .. $#Array)
{
SubRoutine($Array[$i]);
}

实施5

map { SubRoutine($_) } @Array ;
262215 次浏览
小开

1 is substantially different from 2 and 3, since it leaves the array in tact, whereas the other two leave it empty.

I'd say #3 is pretty wacky and probably less efficient, so forget that.

Which leaves you with #1 and #2, and they do not do the same thing, so one cannot be "better" than the other. If the array is large and you don't need to keep it, generally scope will deal with it (but see NOTE), so generally, #1 is still the clearest and simplest method. Shifting each element off will not speed anything up. Even if there is a need to free the array from the reference, I'd just go:

undef @Array;

when done.

  • NOTE: The subroutine containing the scope of the array actually keeps the array and re-uses the space next time. Generally, that should be fine (see comments).
小开

IMO, implementation #1 is typical and being short and idiomatic for Perl trumps the others for that alone. A benchmark of the three choices might offer you insight into speed, at least.

小开
最佳答案

  • In terms of speed: #1 and #4, but not by much in most instances.

    You could write a benchmark to confirm, but I suspect you'll find #1 and #4 to be slightly faster because the iteration work is done in C instead of Perl, and no needless copying of the array elements occurs. ($_ is aliased to the element in #1, but #2 and #3 actually copy the scalars from the array.)

    #5 might be similar.

  • In terms memory usage: They're all the same except for #5.

    for (@a) is special-cased to avoid flattening the array. The loop iterates over the indexes of the array.

  • In terms of readability: #1.

  • In terms of flexibility: #1/#4 and #5.

    #2 does not support elements that are false. #2 and #3 are destructive.

小开

If you only care about the elements of @Array, use:

for my $el (@Array) {
# ...
}

or

If the indices matter, use:

for my $i (0 .. $#Array) {
# ...
}

Or, as of perl 5.12.1, you can use:

while (my ($i, $el) = each @Array) {
# ...
}

If you need both the element and its index in the body of the loop, I would expect using each to be the fastest, but then you'll be giving up compatibility with pre-5.12.1 perls.

Some other pattern than these might be appropriate under certain circumstances.

小开

In single line to print the element or array.

print $_ for (@array);

NOTE: remember that $_ is internally referring to the element of @array in loop. Any changes made in $_ will reflect in @array; ex. 

my @array = qw( 1 2 3 );
for (@array) {
$_ = $_ *2 ;
}
print "@array";

output: 2 4 6

小开

The best way to decide questions like this to benchmark them:

use strict;
use warnings;
use Benchmark qw(:all);


our @input_array = (0..1000);


my $a = sub {
my @array = @{[ @input_array ]};
my $index = 0;
foreach my $element (@array) {
die unless $index == $element;
$index++;
}
};


my $b = sub {
my @array = @{[ @input_array ]};
my $index = 0;
while (defined(my $element = shift @array)) {
die unless $index == $element;
$index++;
}
};


my $c = sub {
my @array = @{[ @input_array ]};
my $index = 0;
while (scalar(@array) !=0) {
my $element = shift(@array);
die unless $index == $element;
$index++;
}
};


my $d = sub {
my @array = @{[ @input_array ]};
foreach my $index (0.. $#array) {
my $element = $array[$index];
die unless $index == $element;
}
};


my $e = sub {
my @array = @{[ @input_array ]};
for (my $index = 0; $index <= $#array; $index++) {
my $element = $array[$index];
die unless $index == $element;
}
};


my $f = sub {
my @array = @{[ @input_array ]};
while (my ($index, $element) = each @array) {
die unless $index == $element;
}
};


my $count;
timethese($count, {
'1' => $a,
'2' => $b,
'3' => $c,
'4' => $d,
'5' => $e,
'6' => $f,
});

And running this on perl 5, version 24, subversion 1 (v5.24.1) built for x86_64-linux-gnu-thread-multi

I get:

Benchmark: running 1, 2, 3, 4, 5, 6 for at least 3 CPU seconds...
1:  3 wallclock secs ( 3.16 usr +  0.00 sys =  3.16 CPU) @ 12560.13/s (n=39690)
2:  3 wallclock secs ( 3.18 usr +  0.00 sys =  3.18 CPU) @ 7828.30/s (n=24894)
3:  3 wallclock secs ( 3.23 usr +  0.00 sys =  3.23 CPU) @ 6763.47/s (n=21846)
4:  4 wallclock secs ( 3.15 usr +  0.00 sys =  3.15 CPU) @ 9596.83/s (n=30230)
5:  4 wallclock secs ( 3.20 usr +  0.00 sys =  3.20 CPU) @ 6826.88/s (n=21846)
6:  3 wallclock secs ( 3.12 usr +  0.00 sys =  3.12 CPU) @ 5653.53/s (n=17639)

So the 'foreach (@Array)' is about twice as fast as the others. All the others are very similar.

@ikegami also points out that there are quite a few differences in these implimentations other than speed.


提交答案