how to do a numeric sort

[Skaperen]

i think all that is needed is a sort key function for this. i want to have a list of files sorted in a numeric way so that:

Output:
python-3.4.0-foo.gz
python-3.4.1-foo.gz
python-3.4.10-foo.gz
python-3.4.2-foo.gz
python-3.4.3-foo.gz
python-3.4.4-foo.gz
python-3.4.5-foo.gz
python-3.4.6-foo.gz
python-3.4.7-foo.gz
python-3.4.8-foo.gz
python-3.4.9-foo.gz
python-3.5.0-foo.gz
python-3.5.1-foo.gz
python-3.5.2-foo.gz
python-3.5.3-foo.gz
python-3.5.4-foo.gz
python-3.5.5-foo.gz
python-3.5.6-foo.gz
python-3.5.7-foo.gz

gets sorted like:

Output:
python-3.4.0-foo.gz
python-3.4.1-foo.gz
python-3.4.2-foo.gz
python-3.4.3-foo.gz
python-3.4.4-foo.gz
python-3.4.5-foo.gz
python-3.4.6-foo.gz
python-3.4.7-foo.gz
python-3.4.8-foo.gz
python-3.4.9-foo.gz
python-3.4.10-foo.gz
python-3.5.0-foo.gz
python-3.5.1-foo.gz
python-3.5.2-foo.gz
python-3.5.3-foo.gz
python-3.5.4-foo.gz
python-3.5.5-foo.gz
python-3.5.6-foo.gz
python-3.5.7-foo.gz

anyone know of code that can do this?

[perfringo]

Something like this:

lst = ['python-3.4.0-foo.gz',
       'python-3.4.1-foo.gz',
       'python-3.4.10-foo.gz',
       'python-3.4.2-foo.gz',
       'python-3.4.3-foo.gz',
       'python-3.4.4-foo.gz',
       'python-3.4.5-foo.gz',
       'python-3.4.6-foo.gz',
       'python-3.4.7-foo.gz',
       'python-3.4.8-foo.gz',
       'python-3.4.9-foo.gz',
       'python-3.5.0-foo.gz',
       'python-3.5.1-foo.gz',
       'python-3.5.2-foo.gz',
       'python-3.5.3-foo.gz',
       'python-3.5.4-foo.gz',
       'python-3.5.5-foo.gz',
       'python-3.5.6-foo.gz',
       'python-3.5.7-foo.gz']

def by_version(row):
    return [int(num) for num in row.split('-')[1].split('.')]

sorted(lst, key=by_version)

Output:
['python-3.4.0-foo.gz', 'python-3.4.1-foo.gz', 'python-3.4.2-foo.gz', 'python-3.4.3-foo.gz', 'python-3.4.4-foo.gz', 'python-3.4.5-foo.gz', 'python-3.4.6-foo.gz', 'python-3.4.7-foo.gz', 'python-3.4.8-foo.gz', 'python-3.4.9-foo.gz', 'python-3.4.10-foo.gz', 'python-3.5.0-foo.gz', 'python-3.5.1-foo.gz', 'python-3.5.2-foo.gz', 'python-3.5.3-foo.gz', 'python-3.5.4-foo.gz', 'python-3.5.5-foo.gz', 'python-3.5.6-foo.gz', 'python-3.5.7-foo.gz']

[scidam]

Below is not an universal approach, but it might be useful:

def key(x):
    vals = map(int, x.split('.'))
    vv = [1000, 50, 1]
    return sum(x * y for x, y in zip(vv, vals))

sorted(['3.1.3', '3.1.1', '3.1.9', '3.2.1', '3.1.10'], key=key)

Output:
['3.1.1', '3.1.3', '3.1.9', '3.1.10', '3.2.1']

[Skaperen]

i'm trying these and others but running into problems because it some cases i am trying to sort there are other character values bordering the digits. so to use perfringo's code, i'd need a more magical split().

[perfringo]

i'm trying these and others but running into problems because it some cases i am trying to sort there are other character values bordering the digits. so to use perfringo's code, i'd need a more magical split().

From which side? Are these all 3.x.x versions? If so and other characters are before version number one can split on number 3 as it's redundant.

[Skaperen]

the list of Python docs files in the 3.4.x and 3.5.x ranges was just an example case. any non-decimal character value can be adjacent to the decimal digits. any run of a class (decimal digits vs. not) can be of any length.

[perfringo]

Then one can use groupby:

from itertools import groupby

lst = ['python3.4.0-foo.gz',
       'python2.4.1-foo.gz',
       'python-3.4.10foo.gz',
       'python-3.4.2-foo.gz',
       'python-3.4.3-foo.gz',
       'python-3.4.4-foo.gz',
       'python-3.4.5-foo.gz',
       'python-3.4.6-foo.gz',
       'python-3.4.7-foo.gz',
       'python-3.4.8-foo.gz',
       'python-3.4.9-foo.gz',
       'python-3.5.0-foo.gz',
       'python-3.5.1-foo.gz',
       'python-3.5.2-foo.gz',
       'python-3.5.3-foo.gz',
       'python-3.5.4-foo.gz',
       'python-3.5.5-foo.gz',
       'python-3.5.6-foo.gz',
       'python-3.5.7-foo.gz']

def versions(row):
    return [int(''.join(x for x in group)) for nums, group in groupby(row, lambda x: x.isdigit()) if nums]

sorted(lst, key=versions)

Output:
['python2.4.1-foo.gz', 'python3.4.0-foo.gz', 'python-3.4.2-foo.gz', 'python-3.4.3-foo.gz', 'python-3.4.4-foo.gz', 'python-3.4.5-foo.gz', 'python-3.4.6-foo.gz', 'python-3.4.7-foo.gz', 'python-3.4.8-foo.gz', 'python-3.4.9-foo.gz', 'python-3.4.10foo.gz', 'python-3.5.0-foo.gz', 'python-3.5.1-foo.gz', 'python-3.5.2-foo.gz', 'python-3.5.3-foo.gz', 'python-3.5.4-foo.gz', 'python-3.5.5-foo.gz', 'python-3.5.6-foo.gz', 'python-3.5.7-foo.gz']

EDIT:

Actually no need for comprehension in .join, one can use ''.join(group)

[Gribouillis]

There are also modules for this on pypi, such as this one. It could be worth trying.

[Skaperen]

that looks like a good module to have around.

i ended up implementing the numeric sort, for sorting files, this way:

my script acts as a filter, reading stdin then printing the result to stdout.

this filter script defines a special code character as Unicode U+FFF6, very unlikely to be in any text data. if the input line has this code in it, a single split around it is done and the part after the code is printed to stdout. else (if the line does not have the code in it) it does re.split(r'(\d+)',line). then it scans that list for any elements that are .isdecimal(). strings that are .isdecimal() are padded with enough leading zeros to make it a wide fixed width number. the list is then joined back to a single string. the modified string + the code + the original string are then printed to stdout. this filter will undo what it does when it processes its own result. so a Unix shell pipeline like filter|sort|filter performs the numeric sort i want. and it avoids storing the whole file in memory because sometimes i might have millions of lines to sort.

#!/usr/bin/env python3
from re import split
from sys import stdin
code = chr(0xfff6)
width = 25
regexp = r'(\d+)'
pad = '0'*width
for line in stdin:
    line = line.split('\n',1)[0]
    if not line:
        print()
        continue
    if code in line:
        line = line.split(code,1)[1]
        print(line)
        continue
    seq = split(regexp,line)
    for x in range(len(seq)):
        if seq[x].isdecimal():
            seq[x] = (pad+seq[x])[-width:]
    seq.append(code)
    seq.append(line)
    print(''.join(seq))
exit(0)

no, i didn't want to use line.rstrip(). i didn't want to lose any trailing spaces.

[Gribouillis]

Wouldn't sort --version-sort suit your needs?