String interning?

Question

The second ReferenceEquals call returns false. Why isn't the string in s4 interned?

string s1 = "tom";
string s2 = "tom";


Console.Write(object.ReferenceEquals(s2, s1)); //true

string s3 = "tom";
string s4 = "to";
s4 += "m";

Console.Write(object.ReferenceEquals(s3, s4)); //false

--edit--

To add to the question, this discussion is about string interning and references. Not about the advantages of StringBuilder over string concatenation. Thanks.

--edit2-- Just noticed that when I do String.Intern(s4);, I still get false.

--edit3-- Getting weird now. Both s3 and s4 are interned but their references are not equal? Help!

string s3 = "tom";
string s4 = "to";
s4 += "m";
String.Intern(s4);

Console.WriteLine(s3 == s4); //true
Console.WriteLine(object.ReferenceEquals(s3, s4)); //false
Console.WriteLine(string.IsInterned(s3) != null);  //true (s3 is interned)
Console.WriteLine(string.IsInterned(s4) != null);  //true (s4 is interned)

Answer 1

Strings are immutable. This means their contents can't be changed.

When you do s4 += "m"; internally, the CLR copies the string to another location in memory which contains the original string and the appended part.

Edit: see MSDN string reference

Answer 2

The string in s4 is interred. However, when you execute s4 += "m";, you have created a new string that will not be interred as it's value is not a string literal but the result of a string concatenation operation. As a result, s3 and s4 are two different string instances in two different memory locations.

For more information on string interning, look here, specifically at the last example. When you do String.Intern(s4), you are indeed interring the string, but you are still not performing a reference equality test between those two interred strings. The String.Intern method returns the interred string, so you would need to do this:

string s1 = "tom";
string s2 = "tom";

Console.Write(object.ReferenceEquals(s2, s1)); //true 

string s3 = "tom";
string s4 = "to";
s4 += "m";

Console.Write(object.ReferenceEquals(s3, s4)); //false

string s5 = String.Intern(s4);

Console.Write(object.ReferenceEquals(s3, s5)); //true

Answer 3

In C#, each string is a distinct object, and cannot be edited. You are creating references to them, but each string is distinct. The behaviour is consistent and easy to understand.

Might I suggest examining the StringBuilder class for manipulating strings without creating new instances? It should be sufficient for anything you want to do with strings.

Answer 4

First of all, all words about immutable strings which are already written are correct. But there are some impotent things which are not written. The code

string s1 = "tom";
string s2 = "tom";
Console.Write(object.ReferenceEquals(s2, s1)); //true

display really "True", but only because of some small compiler optimization or like here because CLR ignore C# compiler attributes (see "CLR via C#" book) and place only one string "tom" in the heap.

Second you can fix the situation with following lines:

s3 = String.Intern(s3);
s4 = String.Intern(s4);
Console.Write (object.ReferenceEquals (s3, s4)); //true

Function String.Intern calculate a hash code of the string and search for the same hash in the internal hash table. Because it find this, it returns back the reference to already existing String object. If the string doesn't exist in the internal hash table, a copy of string is made and hash computed. Garbage collector don't free memory for the string, because it is referenced by hash table.

Answer 5

I answer your question here:

http://blogs.msdn.com/ericlippert/archive/2009/09/28/string-interning-and-string-empty.aspx