Natural Order String Comparison
Martin Pool http://sourcefrog.net
Computer string sorting algorithms generally don't order strings containing numbers in the same way that a human would do. Consider:
rfc1.txt
rfc2086.txt
rfc822.txt
It would be more friendly if the program listed the files as
rfc1.txt
rfc822.txt
rfc2086.txt
Filenames sort properly if people insert leading zeros, but they don't always do that.
I've written a subroutine that compares strings according to this natural ordering. You can use this routine in your own software, or download a patch to add it to your favourite Unix program.
Sorting
Strings are sorted as usual, except that decimal integer substrings are compared on their numeric value. For example,
a < a0 < a1 < a1a < a1b < a2 < a10 < a20
Strings can contain several number parts:
x2-g8 < x2-y08 < x2-y7 < x8-y8
in which case numeric fields are separated by nonnumeric characters. Leading spaces are ignored. This works very well for IP addresses from log files, for example.
Leading zeros are not ignored, which tends to give more reasonable results on decimal fractions.
1.001 < 1.002 < 1.010 < 1.02 < 1.1 < 1.3
Some applications may wish to change this by modifying the test that calls isspace.
Performance is linear: each character of the string is scanned at most once, and only as many characters as necessary to decide are considered.
Licensing
This software is copyright by Martin Pool, and made available under the same licence as zlib:
This software is provided 'as-is', without any express or implied warranty. In no event will the authors be held liable for any damages arising from the use of this software.
Permission is granted to anyone to use this software for any purpose, including commercial applications, and to alter it and redistribute it freely, subject to the following restrictions:
The origin of this software must not be misrepresented; you must not claim that you wrote the original software. If you use this software in a product, an acknowledgment in the product documentation would be appreciated but is not required.
Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
This notice may not be removed or altered from any source distribution.
This licence applies only to the C implementation. You are free to reimplement the idea fom scratch in any language.
Get It!
strnatcmp.c, strnatcmp.h - the algorithm itself
natsort.c - example driver program.
natcompare.js - Kristof Coomans wrote a natural sort comparison in Javascript.
natcmp.rb -- An implementation by Alan Davies in Ruby.
Related Work
POSIX sort(1) has the -n option to sort numbers, but this doesn't work if there is a non-numeric prefix.
GNU ls(1) has the --sort=version option, which works
the same way.
The PHP scripting language now has a strnatcmp function based on this code. The PHP wrapper was done by Andrei Zimievsky.
Stuart Cheshire has a Macintosh system extension to do natural ordering. I indepdendently reinvented the algorithm, but Stuart had it first. I borrowed the term natural sort from him.
Sort::Versions
in Perl. "The code has some special magic to deal with common conventions in program version numbers, like the difference between 'decimal' versions (eg perl 5.005) and the Unix kind (eg perl 5.6.1)."
Sort::Naturally is also in Perl, by Sean M. Burke. It uses locale-sensitive character classes to sort words and numeric substrings in a way similar to natsort.
Ed Avis wrote something similar in Haskell.
Pierre-Luc Paour wrote a
NaturalOrderComparator
in Java.
Numacomp - similar thing in Python.
as3natcompare implementation in Flash ActionScript 3.
To Do
Comparison of characters is purely numeric, without taking character set or locale into account. So it is only correct for ASCII. This should probably be a separate function because doing the comparisons will probably introduce a dependency on the OS mechanism for finding the locale and comparing characters.
It might be good to support multibyte character sets too.
If you fix either of these, please mail me. They should not be very hard.