Mirror of forum.nim-lang.org

7441 :: conversion int -> float : what's the rule?

• HJarausch (orginal) [2021-01-27T18:25:02+01:00] view original

  I don't understand the implicit conversion of an int to float

  
  let
    x=3.14
    two= 2
  
  const zwei=2
  
  echo 2*x        # OK
  echo zwei*x   # OK
  
  # but
  echo two*x  # type mismatch: got <int, float64>
  Run

  Why doesn't Nim convert an int to the larger type float as do C and C++ ?

---

• haxscramper_ (orginal) [2021-01-27T18:27:59+01:00] view original

  https://nim-lang.org/docs/lenientops.html#%2A%2CI%2CF

---

• lqdev (orginal) [2021-01-27T18:32:19+01:00] view original

  In Nim converting an int to a to float64 is a narrowing conversion, because an int can be an int64 in case you're running on a 64-bit architecture. Converting from an int64 to a float64 is a narrowing conversion, because a float64 can represent integers precisely only up to 2^53, whilst an int64 can obviously represent integers up to 2^63 (or 2^64 in case of uint64).

---

• HJarausch (orginal) [2021-01-27T18:44:24+01:00] view original

  Thank you both!

  I wished I could tell Nim to use int32 as default int (on a 64bit machine)

---

• Stefan_Salewski (orginal) [2021-01-27T18:49:23+01:00] view original

  It is more like this: In Nim literal numbers have no fix type, so 2*x is ok as 2 is automatically considered as 2.0 of float type. The same seems to work for const numbers, so zwei*x works as const zwei has not the fixed int type, but seems to be a general number. But two*x does not work, because let two= 2 gives the variable two the static type int. If you like less strict math better you may import lenientops.

---

• mratsim (orginal) [2021-01-28T09:41:07+01:00] view original

  C integer promotion rules are one of the biggest source of bugs in C codebase. Requiring opt-in for them is a feature.

---