> At the risk of arguing over semantics, > I'll say again: there is no "one correct" natural order standard, > and therefore it is not "plain and simple" because there is no just > "one" such order.
I don't think there is no commonly accepted "natural sort". For example, I found another one that uses the same order as the python one that I showed above. The so-called version sort in corutils' sort is just not natural sort and it should not be called natural sort. $ printf '%s\n' G . | csvtk sort -k 1:N G . $ printf '%s\n' 1G 1. | csvtk sort -k 1:N 1G 1. $ printf '%s\n' 1G13 1.02 | csvtk sort -k 1:N 1G13 1.02 Wikipedia also explains what natural sort is and provided a few implementation links. I don't think any of them implemented the version sort as the natural sort. https://en.wikipedia.org/wiki/Natural_sort_order > and note that even the above blog writes: > "... Don't let Ned's clever Python ten-liner fool you. Implementing a > natural sort is more complex than it seems ... ". I don't understand this sentence. There is an implementation with just a few lines in python. Unless this implementation is wrong, then there is a simple implementation at least in python. -- Regards, Peng