Different sorting results on different CLR versions

Question

While comparing strings in C#, different clr gives different results on Windows 7 sp1 x64. Here is sample code:

List<string> myList = new List<string>();
myList.AddRange(new[] { "!-", "-!", "&-l", "&l-", "-(", "(-", "-*", "*-", ".-", "-.", "/'", "-/" });
myList.Sort();
Console.WriteLine(Environment.Version);
myList.ForEach(Console.WriteLine);
Console.WriteLine();
Console.WriteLine(string.Compare("!-", "-!"));
Console.WriteLine("!-".CompareTo("-!"));

Here is the sample output:

If I set Target Framework to 4.0:

4.0.30319.18444
!-
-!
&l-
&-l
(-
-(
*-
-*
.-
-.
/'
-/

-1
-1

If I set Target Framework to 2.0:

2.0.50727.5485
-!
!-
&-l
&l-
-(
(-
-*
*-
-.
.-
-/
/'

1
1

Am I missing anything?

In .NET 4.0 the Unicode changed from 5.0 to 5.1, so perhaps that is the difference. Sadly there is very little documentation on this — xanatos, Feb 25 '15 at 11:22
`List.Sort()` uses the default comparer, and the default comparer for `string` uses the current culture. This, in turn, depends on the collation tables supplied with .NET, and those are subject to change. If you want consistent results, use an ordinal-based comparison (`List.Sort(StringComparer.Ordinal)`). — Jeroen Mostert, Feb 25 '15 at 11:29
@JeroenMostert The problem is present even with the InvariantCulture, that should be "stable" — xanatos, Feb 25 '15 at 11:30
No. The invariant culture is still a culture, and still depends on collation tables. The only thing guaranteed to be stable (as it does not depend on any collation at all) is ordinal-based comparison. — Jeroen Mostert, Feb 25 '15 at 11:31
This has been already [discussed here](http://stackoverflow.com/questions/23087995/string-comparison-and-sorting-when-strings-contain-hyphens) and solution is to use [Ordinal comparer](http://stackoverflow.com/a/19371082/2530848) — Sriram Sakthivel, Feb 25 '15 at 11:31
@SriramSakthivel Using the Ordinal comparer is good only in limited cases, because with the Ordinal comparer `è > f` — xanatos, Feb 25 '15 at 11:44

Maxime Rouiller · Answer 1 · 2015-03-17T12:32:26.780

Please ensure that you are sorting with the MyList.Sort(StringComparer.Ordinal).

Unless Unicode start changing the code of their characters, it should provide a constant sorting order. Ordinal will be based off the actual code ID that were assigned to them.

If I take your first example comparing this :

-!
!-

The hyphen is U+002D and the exclamation mark is U+0021. Those codes haven't changed since at least the ASCII tables. I would consider checking your sorting parameters to make sure you compare only on ordinal and not on actual neutral/specific cultures.

Different sorting results on different CLR versions

1 Answers1