Mirror of forum.nim-lang.org

1915 :: Why no OverflowError?

• ManfredLotz (orginal) [2016-01-02T12:00:27+01:00] view original

  Hi there,

  Yesterday, I encountered nim and today I tried a fibonacci sample from Rosetta Code

  proc Fibonacci(n: int, current: int64, next: int64): int64 =

  if n == 0:

  result = current

  else:

  result = Fibonacci(n - 1, next, current + next)

  proc Fibonacci(n: int): int64 =

  result = Fibonacci(n, 0, 1)

  echo( Fibonacci(91))

  which works fine. If I change 91 to 92 I get:

  Error: unhandled exception: over- or underflow [OverflowError]

  which is what I would expect.

  Now I changed int64 to uint64.

  For number 93 Fibonacci(93) works fine and gives: 12200160415121876738.

  For number 94 I get 1293530146158671551 which doesn't look right. I would have expected an OverflowError.

  Is this a bug, or did I miss anything in my code?

  -- Manfred

---

• Jehan (orginal) [2016-01-02T12:09:31+01:00] view original

  Unsigned integers have modulo (UINT_MAX+1) semantics. This is intentional (and mirrors C/C++ conventions). Unsigned ints are intended for manipulating machine words at the bit level, and overflows would just get in the way (e.g. when trying to implement cryptographic algorithms).

---

• ManfredLotz (orginal) [2016-01-02T12:25:24+01:00] view original

  Aaah, understand.

  I see it is even mentioned in the manual: "Unsigned operations all wrap around; they cannot lead to over- or underflow errors."

  Next time I will try to find it in the docs before asking.

  Thanks, Manfred

---