Network Interface Name Sorting in Python

Evergreen problem in sorting strings: How do you implement human-expected ordering when the strings have varying-length numbers in them? Prime example in the networking world is network interface naming in various devices. Example:

>>> interfaces = ["1/1/1", "1/1/2", "1/1/10", "1/1/11", "1/1/10.32", "1/1/10.311", "1/1/10.312", "eth-1/1/1.1", "eth-1/1/1.2", "eth-1/1/1.10"]
>>> print("\n".join(sorted(interfaces)))
1/1/1
1/1/10
1/1/10.311
1/1/10.312
1/1/10.32
1/1/11
1/1/2
eth-1/1/1.1
eth-1/1/1.10
eth-1/1/1.2
>>>

That’s not ideal as the default string sorting just compares the strings character-by-character, and thus “10” comes before “2”, for example.

Let’s create a function that can be used as the sort key:

import string

def interface_sort_func(name: str) -> str:
    in_num = False
    text = ""
    num = ""
    for c in name:
        if c in string.digits:
            in_num = True
            num += c
        elif in_num:
            in_num = False
            text += num.rjust(5, "0")
            num = ""
            text += c
        else:
            text += c
    if in_num:
        text += num.rjust(5, "0")
    return text

Now let’s test the interface sorting again:

>>> print("\n".join(sorted(interfaces, key=interface_sort_func)))
1/1/1
1/1/2
1/1/10
1/1/10.32
1/1/10.311
1/1/10.312
1/1/11
eth-1/1/1.1
eth-1/1/1.2
eth-1/1/1.10
>>>

Looks good!

The idea in this example is that when sorted() compares the list values it will call the function specified in the key argument for each compared value, and the function returns a modified (or normalized) value. The original contents of the input list are not changed at any point. The function expands all embedded numbers of any length to 5 digits, with zero-padding. For example, the value “1/1/2” is compared as “00001/00001/00002” against “00001/00001/00010” (“1/1/10”), and that results in expected ordering.