Mirror of forum.nim-lang.org

6273 :: Understanding "Error: cannot evaluate at compile time"

• spip (orginal) [2020-04-29T02:37:54+02:00] view original

  This small bit of code barks at compile time with (4, 35) Error: cannot evaluate at compile time: size.

  type
    B[S: static int] = object
  
  template foo(size: static int): B[size] =
    B[size]()
  
  echo foo(16).repr
  Run

  The error is about the size in B[size] in the template declaration. Why does the compiler complains as not knowing size value at compile time as it is declared static int?

  When replacing static int by int, the error turns to Error: type mismatch: got <B[16]> but expected 'B[size]'.

  The only combinations to get it compiling are if I replace size by 16, or if I remove all the static in the code. But programming is not trying syntax at random and I would like to understand why it fails when I use static.

---

• juancarlospaco (orginal) [2020-04-29T07:28:16+02:00] view original

  type B[S: static[int]] = object
  template foo(size: static[int]): auto = B[size]()
  echo foo(16).repr
  Run

  Works on my machine

---

• DeletedUser (orginal) [2020-04-29T08:35:20+02:00] view original

  The issue is typedesc/static in template is slightly broken, see issue 1 (this exact issue), issue 2 (points out a case where static works and typedesc doesnt), issue 3 (this exact issue for typedesc).

---

• spip (orginal) [2020-05-01T21:05:25+02:00] view original

  I've been reading all the issues related to static and templates, macros and typedesc. I see that I'm entering the fuzzy zone where Nim typing is not totally defined and the syntax is moving. The discussions on generics vs type variables shows that there's no consensus on how types variability must be managed in Nim. And I think that it impacts finding a clean way to solve these types of bugs.

  Meanwhile, I have to find a way to change my code to stay in Nim safer areas, if possible.

---