Mirror of forum.nim-lang.org

1929 :: possible compiler bug

• geezer9 (orginal) [2016-01-07T03:41:52+01:00] view original

  In an attempt to decide if the type(...) operator produces the static compile-time type or the run-time type, I tried the following experimental code:

  import typetraits
  
  type Thing = ref object of RootObj
      name:string
  
  type Aged = ref object of Thing
      age:int
  
  # does type.name only report STATIC declared type
  var x = new(Thing)
  
  echo("x's type=",int.name)     #this COMPILED
  #echo("x's type=",x.type.name) #gives SYNTAX ERROR got (type x.name)
  
  var y = new(Aged)
  echo("y's type=",y.type.name)
  
  var z = y                     #this COMPILED
  #var z:Thing = y              #gives SYNTAX ERROR got (type z.name)
  
  echo("z's type=",z.type.name)
  
  # hack for runtime (dynamic) type
  method typeName(th:Thing):string {.base.} = "Thing"
  method typeName(ag:Aged):string = "Aged"
  
  echo("x typeName=",x.typeName())
  echo("y typeName=",y.typeName())
  echo("z typeName=",z.typeName())
  Run

  unfortunately what I wanted to try will not compile. If either of the two indicated commands are replaced by the following alternate - I get a syntax error.

  Is this a nim compiler bug??? I am using

  Nim Compiler Version 0.12.0 (2015-10-27) [Linux: i386]

  on a 32bit Linux box.

---

• skyfex (orginal) [2016-01-07T12:14:55+01:00] view original

  This works:

  echo("x's type=",name(x.type))
  Run

  Don't ask me why it's not equivalent.

---

• skyfex (orginal) [2016-01-07T12:42:33+01:00] view original

  If you change it to

  type Thing = ref object of RootObj
      foo:string
  Run

  Then it compiles. So the compiler is confused by the fact that there's a field called "name"

---

• skyfex (orginal) [2016-01-08T09:52:49+01:00] view original

  It seems that this is because of the compiler treating type(x).name as type x.name, but only when x has a field called name, through special case handling. Very likely a bug.

---