8 things you probably didn’t know about C#

Here’s a few unusual things about C# that few C# developers seem to know about.

1. Indexers can use params

We all know the regular indexer pattern x = something[“a”] and to implement it you write:

public string this[string key] {
  get { return internalDictionary[key]; }
}

But did you know that you can use params to allow x = something[“a”, “b”, “c”, “d”] ?

Simply write your indexer like this:

public IEnumerable<string> this[params string[] keys] {
  get { return keys.Select(key => internalDictionary[key]).AsEnumerable(); }
}

The cool thing is you can have both indexers in the same class side-by-side. If somebody passes an array or multiple args they get an IEnumerable back but call with a single arg and they get a single value.

2. Strings defined multiple times in your code are folded into one instance

Many developers believe that:

if (x == "" || x == "y")

will create a couple of strings every time. It won’t.

C#, like many languages, has string interning and every string your app compiles with gets put into an in-memory list that is referenced at runtime.

You can use String.Intern to see if it’s currently in this list but bear in mind that doing String.Intern(“what”) == “what” will always return true as you just defined another string in your source. String.IsInterned(“wh” + “at”) == “what” will also return true thanks to compiler optimizations. String.IsInterned(new string(new char[] { ‘w’,’h’,’a’,’t’ }) == new string(new char[] { ‘w’,’h’,’a’,’t’ }) will only return true if you have “what” elsewhere in your program or something else at runtime has added it to the intern pool.

If you have classes that build up or retrieve regularly used strings at runtime consider using String.Intern to add them to the pool. Bear in mind once in they’re there until your app quits so use String.Intern carefully. The syntax is simply String.Intern(someClass.ToString())

Another caveat is that doing (object)”Hi” == (object)”Hi” will return true in your app thanks to interning. Try it in your debug intermediate window and it will be false as the debugger will not be interning your strings.

3. Exposing types as a less capable type doesn’t prevent use as their real type

A great example of this is when internal lists are exposed as IEnumerable properties, e.g.

private readonly List<string> internalStrings = new List<string>();
public IEnumerable<string> AllStrings { get { return internalStrings; }

You’d likely think nobody can modify internal strings. Alas, it’s all too easy:

((List<string>)x.AllStrings).Add("Hello");

Even AsEnumerable won’t help as that’s a LINQ method that does nothing :( You can use AsReadOnly which creates a wrapper over the list that throws when you try and set anything however and provides a good pattern for doing similar things with your own classes should you need to expose a subset of internal structures if unavoidable.

4. Variables in methods can be scoped with just braces

In Pascal you had to declare all the variables your function would use at the start of the function. Thankfully today the declarations can live next to their assignment and use which prevents acidentally using the variable before you intended to.

What it doesn’t do is stop you using it after you intended. Given that for/if/while/using etc. all allow a nested scope it should come as only mild surprise that you can declare variables within braces without a keyword to achieve the same result:

private void MultipleScopes() {
  { var a = 1; Console.WriteLine(a); }
  { var b = 2; Console.WriteLine(a); }
}

It’s almost useful as now the second copy-and-pasted code block doesn’t compile but a much better solution is to split your method into smaller ones using the extract method refactoring.

5. Enums can have extension methods

Extension methods provide a way to write methods for existing classes in a way other people on your team might actually discover and use. Given that enums are classes like any other it shouldn’t be too surprising that you can extend them, like:

enum Duration { Day, Week, Month };

static class DurationExtensions {
  public static DateTime From(this Duration duration, DateTime dateTime) {
    switch duration {
      case Day:   return dateTime.AddDays(1);
      case Week:  return dateTime.AddDays(7);
      case Month: return dateTime.AddMonths(1);
      default:    throw new ArgumentOutOfRangeException("duration")
    }
  }
}

I think enums are evil but at least this lets you centralize some of the switch/if handling and abstract them away a bit until you can do something better. Remember to check the values are in range too.

6. Order of static variable declaration in your source code matters

Some people insist that variables are ordered alphabetically and there are tools around that can reorder for you… however there is one scenario where re-ording can break your app.

static class Program {
  private static int a = 5;
  private static int b = a;

  static void Main(string[] args) {
   Console.WriteLine(b);
  }
}

This will print the value 5. Reorder the a and b declarations and it will output 0.

7. Private instance variables of a class can be accessed by other instances

You might think the following code wouldn’t work:

class KeepSecret {
  private int someSecret;
  public bool Equals(KeepSecret other) {
    return other.someSecret == someSecret;
  }
}

It’s easy to think of private as meaning only this instance of a class can access them but the reality is it means only this class can access it… including other instances of this class. It’s actually quite useful for some comparison methods.

8. The C# Language specification is already on your computer

Providing you have Visual Studio installed you can find it in your Visual Studio folder in your Program Files folder (x86 if on a 64-bit machine) within the VC#\Specifications folder. VS 2010 comes with the C# 5.0 document in Word format.

It’s full of many more interesting facts such as:

  • i = 1 is atomic (thread-safe) for an int but not long
  • You can & and | nullable booleans with SQL compatibility
  • [Conditional(“DEBUG”)] is more useful than #if DEBUG

And to those of you that say “I knew all/most of these!” I say “Where are you when I’m recruiting!” Seriously, it’s hard enough trying to find C# devs with a solid understanding of the well-know parts of the language.

[)amien

37 responses  

  1. You say that the C# specification says that i++ is atomic/threadsafe for int; I had always understood that not to be the case, and I can’t find anything in the documentation that says it is atomic. Can you point me to what you are seeing? As a separate reference, see Eric Lippert’s answer to this stackoverflow question: http://stackoverflow.com/questions/4628243/is-the-operator-thread-safe

    Chris Shaffer – October 29th, 2012
  2. I knew number three for sure, number one was fascinating but I just never seem to know when to use indexers. You should blog “Neat Things You Can Do With Indexers.”

    KenOctober 29th, 2012
  3. Another option for #3 is to return `internalStrings.Select(x => x)`, as `Select()` implementations are supposed to guarantee that the source object itself is not returned.

    Incidentally, the spec guarantees that `from x in e select x`, a so-called degenerate query expression, will return the same even though trivial selects are normally optimized out.

    dahlbykOctober 29th, 2012
  4. With regard to 4, I liked the fact that in Pascal there was this declare block. In C based languages I dislike the clutter that is caused by inline declarations. Although using ‘var’ as much as possible improves it somewhat.

    Mark – October 29th, 2012
  5. I enjoyed this post. Thanks for sharing! And I only knew a few of them.

    Doug – October 29th, 2012
  6. Because we have good jobs already:)

    Oleg – October 29th, 2012
  7. Wow, the params thingy in indexers is really nice. PS.: I knew almost all of it, are you recruiting? LOL

    Bruno – October 29th, 2012
  8. Really fun list (and yup I did know all of them but I’m geeky like that).

    Some ‘off the top of my head’ feedback.

    3. This is a great gotcha, however I tell our devs not to worry about it in most cases. If you expose a property as IEnumerable (other examples exist!) you are telling people not to assume the underlying implementation, and you may in fact change it in future. There are only a few cases where you absolutely want to prevent someone gaining access to the underlying variable, particular security, and yet in those cases Reflection quickly allows a determined dev to circumvent your carefully laid plans in pretty much every case. As such the real overhead of the creating ReadOnly collections, etc., is very rarely justified – particularly if the stated aim is to prevent consumers doing something unsupported. (IMHO) Bear in mind that LINQ frequently optimistically checks for an array or list before falling back to a brute force approach.

    7. It’s shocking how few people know this one (we do 200+ interviews a year and no one gets it). Another one is how few people can tell you what protected internal is (or that it exists as an access modifier)…

    8. i++ is atomic for int on Intel, on x64 OS it’s atomic for long (as is double). The Interlocked namespace is the best way to safely perform an increment is you wish it to be atomic.

    I think you mean #if DEBUG not if #DEBUG…

    Craig DeanOctober 29th, 2012
  9. @chris – In the C# v5 spec it’s section 5.5 Atomicity of variable references.

    [)amien

    Damien GuardOctober 29th, 2012
  10. “i++ is atomic (thread-safe) for int but not for long”

    Ah, so next year it won’t be… :)

    gliljas – October 29th, 2012
  11. @glijas: Lol

    Damien GuardOctober 29th, 2012
  12. section 5.5 also explicitly states: “there is no guarantee of atomic read-modify-write, such as in the case of increment or decrement.”

    So no, i++ is not atomic. But i = 8 is, if i is an int.

    Harold – October 29th, 2012
  13. Good catch Harold – I could have sworn I saw ++ covered in the atomic section but can’t find it now. I’ll put it down to reading specs at 3am. Article fixed.

    [)amien

    Damien GuardOctober 29th, 2012
  14. 5 I did not know. Could make enums in code maintenance much nicer. Re: 6 – init of static variable with another static, outside of a static constructor? run away! ;)

    @glijas: very good!

    ER – October 30th, 2012
  15. I knew about all of these except the [Conditional(“DEBUG”)] attribute, which I certainly intend to use now. Thanks!

    Atomicity of various operations is something one learns about quickly when trying to implement lock-free structures :P

    Alex – October 30th, 2012
  16. I’m in Brazil :)

    By the way, I didn’t know about [Conditional(“DEBUG”)]

    Fabio – October 30th, 2012
  17. I like the string interning. It works regardless of visibility or membership.

    class Program
    {
        static string sw = "what";
        const string cw = "what";
        public string iw = "what";
    
        static void Main(string[] args)
        {
            Program prog = new Program();
            Console.WriteLine(String.IsInterned("foo"));
            Console.WriteLine(String.IsInterned("what"));
            Console.WriteLine(String.IsInterned(cw));
            Console.WriteLine(String.IsInterned(sw));
            Console.WriteLine(String.IsInterned(prog.iw));
            Console.WriteLine((sw == cw) && (prog.iw == cw) && (sw == prog.iw));
            Console.ReadLine();
        }
    }
    ScottOctober 30th, 2012
  18. #5 is a good idea, I hadn’t thought of that. Although with my love for extension methods, I probably would have come up with it sooner or later. The rest of them, yeah, I knew. As for “where were you when I was recruiting?”, well, where were you when I was looking for a job? :-P

    KFee – October 30th, 2012
  19. I am so disappointed that I only didn’t know number 8. I was hoping for more :)

    Andrei RineaOctober 30th, 2012
  20. My guess on the “i = 1 is atomic” fact is that in x86 assembly the ADD instruction operate on 32-bit registers. Because int’s are 32-bit this can be done in one instruction, where-as long is a 64-bit int and some extra instructions would be required.

    Shannen Saez – October 30th, 2012
  21. Nice post, thanks for the tricks collection even if I knew them, except the #8, which is interesting. ;)
    For the #3, I would say that the best way to really prevent somebody from modifiying an internal collection is to never return that collection, but only copies (e.g. using linq .ToList), or make use of a read-only collection which is made for that (but everything depends on the problem you’re trying to solve, of course).
    AsEnumerable doesn’t do “nothing”, but here it’s just not its use case ;) (it’s for example basically used in linq expressions to convert IQueryable to IEnumerable, so that you can switch from remote request processing to in-memory processing).

    Fenn – October 31st, 2012
  22. If you want to hide a list with IEnumerable

    Make the code

    get { foreach(var s in list) yield return s; }

    You will not be able to cast that to a list.

    lindholm – November 1st, 2012
  23. Great post! I definitely did not know (1) and (8). After reading your post, I tried (1) on VS2012 targeting .NET 4. To my surprise, C# also let me define *other* indexers. I knew the CLR supported this, but I think this is a new feature to C#.

    private class ABC {
    private readonly List ss;

    public ABC(IEnumerable ss) {
    this.ss=ss.ToList();
    }

    public string this[int x] {
    get { return ss[x]; }
    }

    public IEnumerable this[params int[] xs] {
    get { return xs.Select(x => ss[x]); }
    }

    public string this[string s] {
    get { return s; }
    }
    }

    ngm – November 1st, 2012
  24. Ugh, looks like the comment system ate my generics brackets.

    The third indexer (which takes string) is the most surprising. I would not have thought it’d let me create both that one and the int one.

    ngm – November 1st, 2012
  25. How about this one?

    You cannot have an enum value named “__value”.

    public enum TestEnum
    {
    __value
    }

    Won’t compile. This is because the value of an enum instance is stored in a variable called “__value” which is generated by the compiler.

    Armen Shimoon – November 1st, 2012
  26. About the 7th point, it isn’t something that happens just in C#, it occurs in all Object Oriented languages.

    One uses to forget that fact, combined to the custom of respecting the objects’ interfaces.

    JeyDotC – November 1st, 2012
  27. This was a great posting! I didn’t know several, but by knowing a little IL, most items make sense. For example, wouldn’t #7 be the same as saying that access is given based on the requesting object’s typedef token? If private means ‘per typedef token’ then the CLR would have no ability, even, to discriminate based on instance rather than class. Also, the response mentioning protected intern gives a much easier question if familiar with the CLR’s famandassem and famorassem attributes.

    Joshua C – November 3rd, 2012
  28. Nice post!

    The first one I find it really useful.

    Side note:
    When you thought you were hiring, in fact you were in a “Quiz Show” or playing “trivia” with your candidates. The people you wanted to hire perhaps went to real interviews or stayed home playing trivia with their friends.

    Javier Gonel – November 8th, 2012
  29. Just to clarify I would never use these as interview questions!

    I always use real-world domain-specific coding problems looking for coding ability,good design and being able to ask me the right questions.

    My point was if you know C# to this extreme you hopefully know C# very very well.

    [)amien

    Damien GuardNovember 8th, 2012
  30. You fail to point out that C# can run managed or un-managed code which many of its other .net counter parts can not.
    I just hate the fact your code time and investment is all for Microsoft platforms only!

    Dave Rathbone – November 10th, 2012
  31. With Mono you can use that effort on Linux, Android, iOS and MacOS.

    Which makes it a great cross-platform option – each of these platforms uses C# and a lot of core .NET classes but provides direct bindings to the native UI controls etc. and then compiles to native code.

    [)amien

    Damien GuardNovember 10th, 2012
  32. I didn’t know about number 1 – that is very cool!

    I think number 3 is fundamental and is something *every* C# programmer should know!

    Adam RalphNovember 10th, 2012
  33. Thanks for this! I could use tips like these everyday, and I even fall into the category of devs who knew some of them already. Cheers!

    Tim WelshNovember 11th, 2012
  34. Bear in mind once in they’re there until your app quits so use String.Intern carefully.

    Jerome JarvisDecember 13th, 2012
  35. All things relative, every good C# developer should already know all of this. Excellent C# developers should be chuckling at the few of these ‘lesser knowns’ which point out a much larger lack of understanding, namely 3, 4, 5, and 7. Similarly, the atomicity of a 64bit (long) assignment can’t be guaranteed on a 32bit platform because the underlying hardware requires two MOV ops (or an iterative op that is also not atomic), since .NET targets both 32bit and 64bit platforms a 64bit (long) assignment would be subject to the capability of the underlying hardware. The standard takes this into account and so C# itself does not guarantee atomicity of a 64bit assignment, but as a matter of fact a “x = 1L” (long, 64bit assignment) is an atomic operation on a 64bit Intel-based computer unless you’re running your app in “the 32bit WOW”. You can verify this by launching your process on a 64bit machine and then inspecting the assembly/machine code (not the IL) to see a 64bit, atomic assignment. This is right up there with people believing that “default(T)” “where T : class” incurs no assignment, when in fact on an Intel CPU a classic mov + xor pair are generated/executed to zero out the variable location in memory (it’s no different than assigning null or a 0 value.) A very interesting read, I think 6 is interesting, but I’m also curious if that is standard behavior (ECMA compliant/required) or if it’s an implementation-specific by-product of static class initialization in Microsoft’s CLR?

    Shaun WilsonApril 5th, 2013
  36. #6) The order of static field declaration does not matter unless they are initialized inline, in which case it could be said that static field declaration order does not matter while static field initialization order does matter. Also, it is important to note when static field initialization takes place. According to C# Language Specification Version 5.0 – 10.12, the exact order of initializing static classes is:

    “…first a new set of static fields (§10.5.1) for that particular closed type is created. Each of the static fields is initialized to its default value (§5.2). Next, the static field initializers (§10.4.5.1) are executed for those static fields. Finally, the static constructor is executed.”

    Anyways, a side-note regarding an error in the specs, it references §10.4.5.1 as the location of static field initialization, but it’s wrong. The correct location is §10.5.5.1 in the Version 5.0 specifications document. The older spec document(s) have it at §10.4.5.1.

    Kip Potter – June 20th, 2013
  37. pingback

    […] For my first ever post, I thought I’d list some neat things that I know about C# that might surprise the average reader, inspired by damieng’s similar post from 2012. […]

    10 Things You Maybe Didn’t Know About C# | Memory FencesOctober 16th, 2013

Respond to this