是否可以读取实习生池中的所有字符串？

Question

It's well known that in some certain cases when using strings in C#, the CLR does string interning as an optimization.

So my questions are:

• It possible to read all the strings that are currently in the intern pool?
• Is there a way to get a reference count to each interned string?
• Would it be possible to read the intern pool from a separate process space?
• If none of these are possible, what's the reasoning for not allowing these use cases?

I could see this being somewhat useful when monitoring memory usage in certain cases. It may also be useful when working with sensitive information (although I would think SecureString would be more preferable in many scenarios).

As far as I can tell, the only public methods related to string interning are String.Intern(string) and String.IsInterned(string)

I'm asking out of curiosity, not trying to solve a real problem. I realize that doing any logic based off of the string intern pool would be a bad idea.

Answer 1

Looking up the interned strings via code has no use case so it's feature was not added in to the language.

However looking up the strings in memory while debugging a program is a very common use case, and there are tools to do that.

You will need to use the tool WinDbg.exe that comes with the Windows SDK. After launching it and attaching it to your program you do the command

.loadby sos clr

and that will load in the extensions for debugging .NET apps. Once you have done that you can do the command

!DumpHeap -strings

and you can see all string objects in the heap.

As for telling if the object in that list that you are looking at is interned or not, I am not entirely sure how. Hopefully if you ask a new question about WinDbg and how to tell if a string is interned or not someone may be able to answer.

Answer 2

You can analyze the strings and duplicates which make sense to intern with MemAnalyzer which is based on ClrMD .

https://github.com/Alois-xx/MemAnalyzer

C>MemAnalyzer.exe -dstrings -f 50KStringsx64.dmp

    Strings(Count)  Waste(Bytes)    String
    500             20,958          String 0
    500             20,958          String 1
    500             20,958          String 2
    500             20,958          String 3
    500             20,958          String 4
    500             20,958          String 5

Summary
==========================================
Strings                       61,330 count
Allocated Size             2,529,742 bytes
Waste Duplicate Strings    2,515,898 bytes

This will give you a metric how many duplicate strings you have and which of them might make sense to intern. To find out which object references a specific string you can add

-showAddress

to show the first address of each string which might be worth interning. Then you can use Windbg and !GCRoot address to find out which object hold this string which should give you an idea in which class you need to add String.Intern calls.

Please note that the .NET String.Intern pool will never release references. If you are dealing with large datasets with different content you should use your own Dictionary pool to be able to release all interned strings when you unload the current dataset and load the next one.