This may work as well...
@hugearray = ( ... );
@toremove = ( ... );
# remove from @hugearray elements indexed in @toremove
@hugearray[@toremove] = ("RemoveMe") x @toremove;
push @hugearray, "SentinelFlag";
for (my $_;;) {
$_ = shift @hugearray;
last if $_ eq "SentinelFlag";
next if $_ eq "RemoveMe";
push @hugearray, $temp;
}
This works by flagging the elements you want to delete, then using
@hugearray as a circular shift register, shift-pushing each element in the array in turn, until it gets back to the beginning. As it goes, it doesn't push back in the elements to be deleted.
How does it fare efficiency-wise? This routine touches each data element once, and likely does at most a single implicit array-copy. I think it is probably an O(n+m) operation (n=@hugearray, m=@toremove). The splice solution is O(m log m + m*O(splice(n)). Since perl arrays are not linked lists, I think splice is an O(n) operation, meaning that the splice solutions are O(m log m + m*n). Which is faster? My guess would be the shift-register, but I don't know.