That is no a split then, that is a pattern match. The problem remains that a line like 'a::b:: c' will match according to your rules but the result will be 'a', ':b', 'c'. You can do this easily with recursion:

my @split_bits = qw( : :: \s+ );

my @result = wierd_split( 'foo:bar::baz  boo:foo::bar::baz', \@split_b
+its );
print "Got $_\n" for @result;

sub wierd_split {
    my ( $str, $to_do, $got ) = @_;
  return (@$got, $str) unless @$to_do;
    my $next_split = shift @$to_do;
    print scalar @$to_do, " splitting '$str' on '$next_split'", "\n";
    my ( $want, $next ) = $str =~ m/(.*?)$next_split(.*)/m;
    print "'$want' '$next'\n";
    push @{$got}, $want;
    # and now for a little recursion
    wierd_split( $next, $to_do, $got ) if $want;
}
__DATA__
2 splitting 'foo:bar::baz  boo:foo::bar::baz' on ':'
'foo' 'bar::baz  boo:foo::bar::baz'
1 splitting 'bar::baz  boo:foo::bar::baz' on '::'
'bar' 'baz  boo:foo::bar::baz'
0 splitting 'baz  boo:foo::bar::baz' on '\s+'
'baz' 'boo:foo::bar::baz'
Got foo
Got bar
Got baz
Got boo:foo::bar::baz
[download]

tachyon

s&&rsenoyhcatreve&&&s&n.+t&"$'$`$\"$\&"&ee&&y&srve&&d&&print

Here's a non-recursive function:

sub wierd_split (\@@) {
    my $regex = shift;
    local $_  = @_ ? shift : $_;

    my    @res;
   ($res [@res],  $_) = split shift @$regex, $_, 2 while @$regex && le
+ngth;
    push  @res => $_ if length;

    @res
}

print "Got $_\n" for wierd_split @{[':', '::', '\s+']} => 'foo:bar::ba
+z frog';
__END__
Got foo
Got bar
Got baz
Got frog
[download]

Adding all the funny stuff of split, like using it in void context, its third parameter, or captures in the regex is left as an exercise for the reader.

Abigail

This gets away with being non recursive only by dint of doing the recurse through a while loop on a single line. I would call that a technicality! Instructional as always.

cheers

tachyon

s&&rsenoyhcatreve&&&s&n.+t&"$'$`$\"$\&"&ee&&y&srve&&d&&print