{"id":21273,"date":"2008-08-12T10:00:00","date_gmt":"2008-08-12T10:00:00","guid":{"rendered":"https:\/\/blogs.msdn.microsoft.com\/oldnewthing\/2008\/08\/12\/the-implementation-of-iterators-in-c-and-its-consequences-part-1\/"},"modified":"2008-08-12T10:00:00","modified_gmt":"2008-08-12T10:00:00","slug":"the-implementation-of-iterators-in-c-and-its-consequences-part-1","status":"publish","type":"post","link":"https:\/\/devblogs.microsoft.com\/oldnewthing\/20080812-00\/?p=21273","title":{"rendered":"The implementation of iterators in C# and its consequences (part 1)"},"content":{"rendered":"

\nLike\n\nanonymous methods<\/A>,\niterators in C# are very complex syntactic sugar.\nYou could do it all yourself (after all, you did<\/I> have to do\nit all yourself in earlier versions of C#),\nbut the compiler transformation makes for much greater convenience.\n<\/P>\n

\nThe idea behind iterators is that they take a function with\n\nyield return<\/CODE><\/A>\nstatements\n(and possible some yield break<\/CODE> statements)\nand convert it into a state machine.\nWhen you yield return<\/CODE>, the state of the function is\nrecorded, and execution resumes from that state the next time the\niterator is called upon to produce another object.\n<\/P>\n

\nHere’s the basic idea:\nAll the local variables of the iterator (treating iterator parameters\nas pre-initialized local variables, including the hidden this<\/CODE>\nparameter)\nbecome member variables of a helper class.\nThe helper class also has an internal state<\/I> member that keeps\ntrack of where execution left off and an internal current<\/I>\nmember that holds the object most recently enumerated.\n<\/P>\n

\nclass MyClass {\n int limit = 0;\n public MyClass(int limit) { this.limit = limit; }<\/p>\n
 public IEnumerable<int> CountFrom(int start)\n {\n  for (int i = start; i <= limit; i++) {\n   yield return i;\n  }\n }\n}\n<\/PRE>\n
\nThe CountFrom<\/CODE> method produces an integer\nenumerator that spits out the integers starting at start<\/CODE>\nand continuing up to and including limit<\/CODE>.\nThe compiler internally converts this enumerator into\nsomething like this:\n<\/P>\n
\n class MyClass_Enumerator : IEnumerable<int> {\n  int state$0 = 0;\/\/ internal member\n  int current$0;  \/\/ internal member\n  MyClass this$0; \/\/ implicit parameter to CountFrom\n  int start;      \/\/ explicit parameter to CountFrom\n  int i;          \/\/ local variable of CountFrom<\/p>\n
  public int Current {\n   get { return current$0; }\n  }<\/p>\n
  public bool MoveNext()\n  {\n   switch (state$0) {\n   case 0: goto resume$0;\n   case 1: goto resume$1;\n   case 2: return false;\n   }<\/p>\n
 resume$0:;<\/FONT>\n   for (i = start; i <= this$0.<\/FONT>limit; i++) {\n    current$0 =<\/FONT> i;\n    state$0 = 1;\n    return true;\n resume$1:;<\/FONT>\n   }<\/p>\n
   state$0 = 2;\n   return false;\n  }\n  … other bookkeeping, not important here …\n }<\/FONT><\/p>\n
 public IEnumerable<int> CountFrom(int start)\n {\n  MyClass_Enumerator e = new MyClass_Enumerator();\n  e.this$0 = this;\n  e.start = start;\n  return e;<\/FONT>\n }\n<\/PRE>\n
\nThe enumerator class is auto-generated by the compiler\nand, as promised, it contains two internal members for the\nstate and current object,\nplus a member for each parameter\n(including the hidden this<\/CODE> parameter),\nplus a member for each local variable.\nThe Current<\/CODE> property merely returns the current object.\nAll the real work happens in MoveNext<\/CODE>.\n<\/P>\n
\nTo generate the MoveNext<\/CODE> method, the compiler\ntakes the code you write and performs a few transformations.\nFirst, all the references to variables and parameters need to\nbe adjusted since the code moved to a helper class.\n<\/P>\n