C#: Strings with same contents

Question

I have heard and read that a string can not be changed (immutable?). That should be correct I guess. But I have also heard that two strings with the same contents share the same memory-space (or what you call it). Is this correct?

And if so, does that mean that if I create a List with thousands of strings, it wouldn't really take up much space at all if most of those strings were equal to each other?

Answer 1

EDIT: In the answer below I've referred to the intern pool as being AppDomain-specific; I'm pretty sure that's what I've observed before, but the MSDN docs for String.Intern suggest that there's a single intern pool for the whole process, making this even more important.

Original answer

(I was going to add this as a comment, but I think it's an important enough point to need an extra answer...)

As others have explained, string interning occurs for all string literals, but not on "dynamically created" strings (e.g. those read from a database or file, or built using StringBuilder or String.Format.)

However, I wouldn't suggest calling String.Intern to get round the latter point: it will populate the intern pool for the lifetime of your AppDomain. Instead, use a pool which is local to just your usage. Here's an example of such a pool:

public class StringPool
{
    private readonly Dictionary<string,string> contents =
        new Dictionary<string,string>();

    public string Add(string item)
    {
        string ret;
        if (!contents.TryGetValue(item, out ret))
        {
            contents[item] = item;
            ret = item;
        }
        return ret;
    }
}

You'd then just use something like:

string data = pool.Add(ReadItemFromDatabase());

(Note that the pool isn't thread-safe; normal usage wouldn't need it to be.)

This way you can throw away your pool as soon as you no longer need it, rather than having a potentially large number of strings in memory forever. You could also make it smarter, implementing an LRU cache or something if you really wanted to.

EDIT: Just to clarify why this is better than using String.Intern... suppose you read a bunch of strings from a database or log file, process them, and then move onto another task. If you call String.Intern on those strings, they will never be garbage collected as long as your AppDomain is alive - and possibly not even then. If you load several different log files, you'll gradually accumulate strings in your intern pool until you either finish or run out of memory. Instead, I'm suggesting a pattern like this:

void ProcessLogFile(string file)
{
    StringPool pool = new StringPool();
    // Process the log file using strings in the pool
} // The pool can now be garbage collected

Here you get the benefit of multiple strings in the same file only existing once in memory (or at least, only getting past gen0 once) but you don't pollute a "global" resource (the intern pool).

Answer 2

This is more or less true. It is called "string interning". String literals will be present in memory only once and every variable set to the same value points to this single representation. Strings that are created in code are not automatically interned though.

http://msmvps.com/blogs/manoj/archive/2004/01/09/1549.aspx

Answer 3

If I remember correctly, string that are hard-coded in code are pooled separately. This is called "Interned" and there is a method to query whether a string is: String.IsInterned Method

On that page under "Remarks" you can read:

The common language runtime automatically maintains a table, called the "intern pool", which contains a single instance of each unique literal string constant declared in a program, as well as any unique instance of String you add programmatically.

Hope this helps you a bit, and correct me if I'm wrong.

Matthias