Mirror of forum.nim-lang.org

5451 :: Integers in Nim

• manfred_h (orginal) [2019-11-04T15:37:47+01:00] view original

  I wonder whether integers work in Nim as most people would expect.

  Is anybody surprised by the output of the program below?

  const c_1bn = 1_000_000_000
  
  let i_4bn = 4 * c_1bn
  
  proc test_1[T](n : T) = echo n, "A"
  
  proc test_1[T : int|uint](n : T) = echo n, "B"
  
  proc test_1(n : int64) = echo n, "C"
  
  proc test_1(n : int16) = echo n, "D"
  
  test_1(c_1bn)
  test_1(2'i8)
  test_1(3'u16)
  test_1(i_4bn)
  
  
  proc test_2[T](n : T) = echo n, "A"
  
  proc test_2[T : int](n : T) = echo n, "B"
  
  proc test_2(n : uint) = echo n, "C"
  
  proc test_2(n : int16) = echo n, "D"
  
  test_2(5_000_000_000)
  test_2(6'i8)
  test_2(7'u16)
  
  
  proc test_3[T](n : T) = echo n, "A"
  
  proc test_3[T : uint](n : T) = echo n, "B"
  
  proc test_3(n : int) = echo n, "C"
  
  proc test_3(n : int64) = echo n, "D"
  
  test_3(8_000_000_000)
  test_3(9'i8)
  test_3(10'u16)
  Run

  (Wrong answers: 1abcd2bcd3bcd4abcd5bcd6bcd7bcd8abc9bcd10bcd)

---

• Araq (orginal) [2019-11-04T15:47:46+01:00] view original

  The program doesn't compile for me:

  temp2.nim(13, 7) Error: ambiguous call; both temp2.test_1(n: T: int or uint) [declared in C:Usersrumpfprojectsnimtemp2.nim(7, 6)] and temp2.test_1(n: T) [declared in C:Usersrumpfprojectsnimtemp2.nim(5, 6)] match for: (int literal(1000000000))

---

• manfred_h (orginal) [2019-11-04T15:53:09+01:00] view original

  Given that the template in line 7 is more specific than the template in line 5, is the compile time error the expected result?

---

• Araq (orginal) [2019-11-04T15:58:28+01:00] view original

  "more specific" refers to ref T vs T, it doesn't refer to type generic type constraints in the form [T: int] (which is super weird anyway, int is not a typeclass).

---

• manfred_h (orginal) [2019-11-04T16:19:21+01:00] view original

  
  proc xyz[T : int](x : T) = echo x
  
  xyz(1'i8)
  xyz(1'i32)
  Run

  generates two distinct procedures for int8 and int32 in the c code. How does that happen if int is not a generic type, which would be treated as a type class according to the spec?

---

• lucian (orginal) [2019-11-04T16:48:43+01:00] view original

  this is interesting,I was expecting that T was lifted to integer. Instead is just validated it can be used as integer but the type stays as passed from parameter.

  proc xyz[T : int](x : T) =
      echo "type of param is ", type(x)
      echo "type of generic is ", T
      echo x
  
  xyz(1'i8)
  # type of param is int8
  # type of generic is int8
  Run

---

• manfred_h (orginal) [2019-11-04T17:20:08+01:00] view original

  see for https://github.com/nim-lang/Nim/issues/12552 for how

  proc xyz[T : int](x : T)
  Run

  behaves

---

• manfred_h (orginal) [2019-11-05T12:53:15+01:00] view original

  A little puzzler:

  proc add_size[T : int](n : T) : T = n + n.sizeof.T
  proc add_size(n : int16) : int16 = n + 20'i16 # Bonus !
  
  let n_1 : int8  = 1
  let n_2 : int16 = add_size(n_1)
  let n_22 : int32 = add_size(n_2)
  let n_26 : int = add_size(n_22)
  let n_34 : int64 = add_size(n_26)
  
  echo n_34 # output : 34  (on 64bit system)
  Run

---