Different behaviour on combining diacritics between String.Split() and String.IndexOf()

Question

If I have a string that contains combining diacritics, there seems to be some confusion between behaviour between different string functions. If I use String.IndexOf(), it will combine the diacritic and find the correct character. If I use String.Split(), for some reason it will not combine them and will not find the combined character.

Example code:

string test = "abce\u0308fgh";
Console.WriteLine(test.IndexOf("e"));
Console.WriteLine(test.IndexOf("ë"));

This will work as expected, meaning the e is not found, but the ë is. But for some reason this doesn't behave similarly:

string test = "abcde\u0308fgh";
Console.WriteLine(test.Split('e').Length.ToString());
Console.WriteLine(test.Split('ë').Length.ToString());

For some reason Split() will not combine the diacritic and will split by e, but not by ë.

Is there some reason for this functionality and is there a way to either have an IndexOf() function that doesn't combine the diacritic, or preferably a Split() function that does?

Edit: Noticed I had earlier written wrong code, it had 'e' and not "e"

string test = "abce\u0308fgh";
Console.WriteLine(test.IndexOf('e'));
Console.WriteLine(test.IndexOf('ë'));

This behaves as the Split() also, so it is not between the methods, it's between taking a character or a string.

Answer 1

Actually, when I copy and paste your example code into a blank program, I get exactly the behavior I might expect: both IndexOf() and Split() do not treat the combined character as the passed in ë search character. I.e. the call to IndexOf('ë') returns -1 for me, consistent with how you describe the behavior of Split().

That said, if you want Split() to treat such two-character representations of single-character versions as if they were in fact originally the single-character version, you can just call string.Normalize() before Split(). For example:

Console.WriteLine(test.Normalize().Split('ë').Length);

The Normalize() method has an overload to let you control the exact type of normalization, should that be required (it's not in the example you've provided).