naturalsort 1.0.3

Usage

Here’s an example of regular sorting (based on the ASCII order of individual characters) compared to natural order sorting:

> # Import the sorted() alternative.
> from natsort import natsort
>
> # This is plain old sorting (what we DON'T want).
> sorted(['1', '5', '10', '50'])
['1', '10', '5', '50']
>
> # This is natural order sorting (what we DO want).
> natsort(['1', '5', '10', '50'])
['1', '5', '10', '50']

Why another natsort module?!

There was already a natsort package available on PyPI before I uploaded the first release of my naturalsort package, so why did I upload another package with a very similar name? Because the two packages implement different forms of natural order sorting!

My main use case for natural order sorting has always been for sorting filenames and pathnames, specifically those containing software version numbers. I wrote my naturalsort module years ago because I couldn’t find any for Python, but never published it.

At some point I got sick of manually copying versions of my natural order sorting module back and forth between projects so I decided to either find an alternative available on PyPI or publish my own module. That’s when I found the natsort package and started using it in several projects.

At some point I got bitten in the ass because I didn’t properly test the natsort package for my use case. Here’s a simple scenario which works as I expect it to:

> from natsort import natsorted
> natsorted(['1.8.1-r26', '1.8.1-r30', '2.0-r2', '2.0-r7', '2.0-r11'])
['1.8.1-r26', '1.8.1-r30', '2.0-r2', '2.0-r7', '2.0-r11']

However as I said my actual use case was for sorting filenames with version numbers embedded in them, for example:

> from natsort import natsorted
> versions = ['1.8.1-r26', '1.8.1-r30', '2.0-r2', '2.0-r7', '2.0-r11']
> natsorted(['my-package-%s' % v for v in versions])
['my-package-2.0-r2',
 'my-package-2.0-r7',
 'my-package-2.0-r11',
 'my-package-1.8.1-r26',
 'my-package-1.8.1-r30']

This result really surprised me when I saw it for the first time, although it is the intended result of the natsort package: The hyphen before the version number is interpreted as a negative sign, which explains why 2.0 now comes before 1.8.1.

So there’s a long answer to a simple question: the two packages do different things. Use the naturalsort package if you need to reliably sort version numbers regardless of separators and use the natsort package if you need to sort strings containing more complex numbers like floating point numbers with negative signs and exponentials.

Contact

The latest version of naturalsort is available on PyPI and GitHub. For bug reports please create an issue on GitHub. If you have questions, suggestions, etc. feel free to send me an e-mail at peter@peterodding.com.

License

This software is licensed under the MIT license.

© 2013 Peter Odding.