How can I get a "smart" sort of strings using Python?

Question

I have users (mostly Germans) who need to select items from a long list. I will implement autocomplete, but I also want to present the elements to them in the expected order. I asked a couple of users for typical strings to sort them and found that it was (basically) consistent. However, implementing this order is difficult:

user_expectation(l)                               "        <        @        1        2        10       10abc    A        e        é        E        Z
sorted(l)                                         "        1        10       10abc    2        <        @        A        E        Z        e        é
sorted(l, key=lambda w: w.lower())                "        1        10       10abc    2        <        @        A        e        E        Z        é
ns.natsorted(l)                                   1        2        10       10abc    "        <        @        A        E        Z        e        é
ns.natsorted(l, alg=ns.I)                         1        2        10       10abc    "        <        @        A        E        Z        e        é
ns.natsorted(l, alg=ns.LOCALE | ns.LF | ns.G)     1        2        10       10abc    "        <        @        A        E        e        Z        é
ns.natsorted(l, alg=ns.LOCALE | ns.LF | ns.G), en 1        2        10       10abc    <        "        @        A        E        e        é        Z
ns.natsorted(l, alg=ns.LOCALE | ns.LF | ns.G), de 1        2        10       10abc    <        "        @        A        E        e        é        Z
ns.natsorted(l, alg=ns.LF | ns.G), de             1        2        10       10abc    "        <        @        A        e        E        Z        é

Hence:

• Special characters first - the order does not matter if it is consistent.
• Further. Sort numbers numerically by matching prefix (from here ['1', '10', '2'])
• Characters (Latin1?)
  • Without first accent
  • Lower case before capital case (although this is probably not so important
  • Accented / special later

the code

# -*- coding: utf-8 -*-

from __future__ import unicode_literals
import natsort as ns
import locale

def custom_print(name, l):
    s = u"{:<50}".format(name)
    for el in l:
        s += u"{:<5}\t".format(el)
    print(u"\t" + s.strip())

l = ['"', "<", "@", "1", "2", "10", "10abc", "A", "e", "é", "E", "Z"]
custom_print("user_expectation(l)", l)
custom_print("sorted(l)", sorted(l))
custom_print("sorted(l, key=lambda w: w.lower())",
             sorted(l, key=lambda w: w.lower()))
custom_print("ns.natsorted(l)", ns.natsorted(l))
custom_print("ns.natsorted(l, alg=ns.I)", ns.natsorted(l, alg=ns.I))
custom_print("ns.natsorted(l, alg=ns.LOCALE | ns.LF | ns.G)",
             ns.natsorted(l, alg=ns.LOCALE | ns.LF | ns.G))
locale.setlocale(locale.LC_ALL, 'en_US.UTF-8')
custom_print("ns.natsorted(l, alg=ns.LOCALE | ns.LF | ns.G), en",
             ns.natsorted(l, alg=ns.LOCALE | ns.LF | ns.G))
locale.setlocale(locale.LC_ALL, 'de_DE.UTF-8')
custom_print("ns.natsorted(l, alg=ns.LOCALE | ns.LF | ns.G), de",
             ns.natsorted(l, alg=ns.LOCALE | ns.LF | ns.G))
custom_print("ns.natsorted(l, alg=ns.LF | ns.G), de",
             ns.natsorted(l, alg=ns.LF | ns.G))

natsort IGNORECASE, LOWERCASEFIRST, LOCALE (de en), GROUP . , . ? ( LF, , )

Answer 1

natsort version >= 5.1.0, .

.

import re
import natsort as ns

def special_chars_first(x):
    '''Ensure special characters are sorted first.'''
    # You can add error handling here if needed.
    # If you need '_' to be considered a special character,
    # use [0-9A-Za-z] instead of \W.
    return re.sub(r'^(\W)', r'0\1', x)
    # An alternate, less-hacky solution.
    #if re.match(r'\W', x):
    #    return float('-inf'), x
    #else:
    #    return float('inf'), x

l = ['"', "<", "@", "1", "2", "10", "10abc", "A", "e", "é", "E", "Z"]
print(ns.natsorted(l, key=special_chars_first, alg=ns.G | ns.LF))

['"', '<', '@', '1', '2', '10', '10abc', 'A', 'e', 'é', 'E', 'Z']

, - ( - , '_') '0', , ( natsort ).

Answer 2

key -, (, ).

sorted ( natsorted) , .

, .

import re
def replace_special(s):
  # add more characters to regex, as required
  return re.sub('[<@]', ' ', s)

l = ['"', "<", "@", "1", "10", "2", "10abc", "A", "e", "é", "E", "Z"]

sorted(l, key=replace_special)
['<', '@', '"', '1', '10', '10abc', '2', 'A', 'E', 'Z', 'e', 'é']