I get the following:
DB<1> x split /(a|b)+/, '12aba34ba5'
0 12
1 'a'
2 34
3 'a'
4 5
If I change that to:
DB<8> x split /(a|b|c|d|e)+/, '12abc34de5'
0 12
1 'c'
2 34
3 'e'
4 5
So, it's only capturing the last delimiter. If you do a regex the same way you get the same thing:
DB<12> x '12abc34de5' =~ /(a|b|c|d|e)+/g
0 'c'
1 'e'
Interesting. I'm not sure why. Maybe it has something to do with a list being returned in a scalar context (the last element is captured that way).
If you want to capture the whole delimiter you need to put the +
inside the parens:
DB<16> x split /([abcde]+)/, '12abc34de5'
0 12
1 'abc'
2 34
3 'de'
4 5