The ST is a space/time optimization, trading using extra space in order to save time.

Since you DON'T have the extra space, why not just eat the extra time and do the split in your comparison?

It might be worth benchmarking; it's going to be slower than ST or GRT would be, sure, but maybe it's tolerable for the data set sizes you're talking about. Sorting 759 items isn't THAT bad...

$ perl -e'for(1..759){push @x,rand}; @x=sort {$count++; $a <=> $b} @x;
+ print "$count\n"'
6783
[download]

--
Mike