Ah, I think we both have different definitions of fields.

Your definition (please correct if I'm wrong) is whatever comes out of the split that is currently implemented in Sort::Fields. In perl terms it would be the output from:

perl -MData::Dumper -e'$_=" a b c"; print Dumper(split /\s+/, $_)'
$VAR1 = '';
$VAR2 = 'a';
$VAR3 = 'b';
$VAR4 = 'c';
[download]

Using your definition, there are 4 fields here (list indices 0..3)

My definition is what a user of Sort::Fields would naturally count as a field (ignoring initial whitespace), before its musty innards get hold of it. In perl terms, this would be along the lines of:

perl -MData::Dumper -e' $_=" a b c"; print Dumper(split)'             
$VAR1 = 'a';
$VAR2 = 'b';
$VAR3 = 'c';
[download]

Using my definition, there are 3 fields here (list indices 0..2)

I claim that Sort::Fields already will ignore leading whitespace in any field (my definition) except field 1. When there exists leading whitespace in what I call field 1, the currently implemented split in Sort::Fields will return a null for field 1 (your definition).

Is this a good description of the situation?