in reply to PDL: efficient self-referencing math?

The posted code actually has a sort of cascade effect; it starts at the highest-numbered (last) value, subtracts that from the second-last, then moves down one, subtracting that from the one below, etc:
pdl> p $x = floor random(5) * 10 [9 4 3 2 8] pdl> p Dumper($xperl = $x->unpdl) $VAR1 = ['9','4','3','2','8']; pdl> $xperl->[ $_ -1 ] -= $xperl->[ $_ ] for reverse 1 .. $#$xperl; pdl> p Dumper($xperl) $VAR1 = [14,-5,9,-6,8];
The way to achieve that would be to do the below, but multiplying $to_sub by sequence($x->dim(0)) before subtracting. On the assumption this was not intended and instead the idea was to rotate the vector one over, then subtract that from the original:
pdl> p $x = pdl '[9 4 3 2 8]' [9 4 3 2 8] pdl> $to_sub = $x->rotate(-1)->sever # MUST sever, or next mutates $x pdl> $to_sub(-1) .= 0 pdl> p $to_sub [4 3 2 8 0] pdl> p $x - $to_sub [5 1 1 -6 8]