highest - finding the highest numbers in a list

This software is owned by The university of California, Irvine - not any version of the GPL. GPL is a fine series of licenses, but the owners of the software need it to be distributed under these terms.
highest is a program that finds the highest numbers in a sequence of lines containing numbers.

Usage:

seki-strombrg> highest -h
Usage: /Dcs/seki/strombrg/bin/highest [-n numtokeep] [-f fieldno] [-a] [-r] [-h]
Usage: -a says to sort in ASCII order, rather than numerically
Usage: -r says to sort in reverse
Usage: -h says to give usage information (help)

A comparison of some different methods of finding the largest file in a directory hierarchy.

GNU find and xargs (to work on files with newlines in them)?	Traditional *ix find and xargs (newlines are a problem)	Small to medium number of files	Large number of files	Huge number of files	Commentary	O notation: n is the number of files, m is the number of files to keep, n assumed always >> m (n much larger than m), and c is a (usually) large constant	Concurrency	Command
Yes	No	Yes	No	No	Very simple, but poor accuracy (filenames with newlines are troublesome) and scalability	O(n*log(n)+m)	Poor - sort is most likely going to read the whole list before starting to sort	find /mtpt -type f -print \| xargs du -s \| sort -nr \| head -10
No	Yes	Yes	No	No	Pretty simple, good accuracy, but poor scalability	O(n*log(n)+m)	Poor - sort is most likely going to read the whole list before starting to sort	find /mtpt -type f -print0 \| xargs -0 du -s \| sort -nr \| head -10
No	Yes	Yes	No	No	Kind of complex, good accuracy, but a little better scalability if you are able to assume that all your numbers will be over 10000	O((n/c)*log(n/c)+m)	Poor - sort is most likely going to read the whole list before starting to sort	find /mtpt -type f -size 10000c -print0 \| xargs -0 du -s \| sort -nr \| head -10
No	Yes	No	Yes	No	More complex, but pretty good accuracy and scalability	O(n+n*log(m))	Good - highest should do its processing interleaved well with the processing done by the find and du	find /mtpt -type f -print0 \| xargs -0 du -s \| highest
No	Yes	No	No	Yes	Pretty complex, but pretty good accuracy and excellent scalability, assuming you now all the largest numbers will be above 10,000	O(n/c+n*log(m))	Good - highest should do its processing interleaved well with the processing done by the find and du	find /mtpt -type f -size 10000c -print0 \| xargs -0 du -s \| highest

Here's a graph comparing the performance of highest and GNU sort. As you can see, highest never beats GNU sort in the ranges compared, but:

GNU sort stops working. This is because highest doesn't require disk space to do its work; GNU sort does.
Looking at the graph, you can see a slight tendency for the GNU sort line to cross the highest line. I believe that before long, highest would be outperforming GNU sort significantly - but only for pretty large datasets and only if GNU sort has enough disk space to work.

Back to Dan's tech tidbits