Function Split, bug or error in the documentation?

[bojinlund]

use strict;
use warnings;
use 5.010;

# from documentation:
# ... and each of these:

print join(':', split(//, 'abc', 3)), "\n";
print join(':', split(//, 'abc', 4)), "\n";

# produces the output a:b:c .

say $^V;
say $^O;

__DATA__
output:
a:b:c
a:b:c:
v5.30.1

In the second call there is an extra empty string. I am using Windows 10.

[haukex]

AFAICT with a quick check, Perl has always behaved as your code shows. It appears this documentation bug was introduced with a rewrite of the split manpage for 5.16 in bd46758. The previous version did not have this issue, it said:

Empty trailing fields, on the other hand, are produced when there is a match at the end of the string (and when LIMIT is given and is not 0), regardless of the length of the match. For example:

print join(':', split(//,   'hi there!', -1)), "\n";
print join(':', split(/\W/, 'hi there!', -1)), "\n";

produce the output 'h:i: :t:h:e:r:e:!:' and 'hi:there:', respectively, both with an empty trailing field.

You should definitely open a bug report against perlfunc.pod.

[soonix]

Same here, (v5.26.1-1 linux). I assume it is documented in this part in split:

An empty trailing field, on the other hand, is produced when there is a match at the end of EXPR, regardless of the length of the match (of course, unless a non-zero LIMIT is given explicitly, such fields are removed, as in the last example).

[davies]

Eily explained this to me in Re^3: Avoiding repeated undefs - very clearly, I thought. Unless I understood him or you. :-)

Regards,

John Davies

[bojinlund]

I am trying to understand split and its documentation. To do this I have implement split by using m{} and the variables @- and @+. By this I have found some problems

This script shows some of them:

use strict;
use warnings;
use 5.010;

use Data::Dump qw(dump dd ddx);
sub split_e { }

split_e( // );
# Warning: Use of uninitialized value $_ in pattern match (m//) at pm_
+1.pl ...

my @rv = split( //, '' );
# Warning: none

my $str = '1-10,20';
my $pat = '(-)|(,)';

@rv = split( $pat, $str );
warn dump @rv;    # (1, "-", undef, 10, undef , ",", 20)

@rv = $str =~ m{$pat}g;
warn dump @rv;    # ("-", undef, undef, ",")

while ( my $rv = $str =~ m{$pat}gc ) {
    for my $ix ( 0 .. 99 ) {
        if ( defined $-[$ix] ) {
            say sprintf '$ix= %d (%d,%d)<%s>', $ix, $-[$ix], $+[$ix],
              substr $str, $-[$ix], $+[$ix] - $-[$ix];
        }
    }
}

say "Strawberry Perl $^V";
say $^O;
__DATA__
output:
$ix= 0 (1,2)<->
$ix= 1 (1,2)<->
$ix= 0 (4,5)<,>
$ix= 2 (4,5)<,>
Strawberry Perl v5.30.1
MSWin32

Questions

Magic in split

This split_e( // ); gives the warning: Use of uninitialized value $_ in pattern match (m//) at pm_1.pl ... but this split( // ); does not.

Is there some magic in split indicate by the ‘/’s in “/PATTERN/ in the documentation?

Bug in the variables @- and @+

I find that the result from split, m{} and when using $-[$ix] $+[$ix] are inconsistent.

Split gives (1, "-", undef, 10, undef , ",", 20).

m{} gives ("-", undef, undef, ",").

Using $-[$ix] $+[$ix] does not include the undefs

Is this a bug in the last case?

[ikegami]

split_e( // ), short for split_e( $_ =~ m// ), performs a match operation. You want split_e( qr// ).

split is an operator. And like all operators, it has full control over the syntax of its operands. split /.../, ... is functionally equivalent to split qr/.../, .... Keep in mind that split predates qr//.