Mirror of forum.nim-lang.org

720 :: Countup with floats

• thomas (orginal) [2015-01-05T21:31:44+01:00] view original

  The documentation for the countup generator states that the step may only be positive, but it doesn't specify the type. When I try to compile the following code I get a compiler error:

  for t in countup(0.0, 1.0, 0.1):
  Run

  Is there an alternative function in the standard library I should be looking for?

---

• Araq (orginal) [2015-01-05T22:23:55+01:00] view original

  No, but it's trivial to implement. However most people are already concerned with system.nim's size. Maybe start a 'floatutils' module? Or add it to 'math'?

---

• brianrogoff (orginal) [2015-01-05T23:32:00+01:00] view original

  The docs aren't wrong. They state 'Counts from ordinal value a up to b with the given step count.' OK, 'step count' is confusing, it should really just be 'step'. But that's not your issue, yours is that floats are not ordinals.

  Be careful and test whatever you write, because floating point ranges like this are vulnerable to being off on account of floating point inaccuracies. For example, if your generator code is roughly

  iterator countup*(a: float, b: float, step = 1.0): float {.inline.} =
    var res:float = a
    while res <= b:
      yield res
      res += step
  Run

  then

  for t in countup(0.0, 0.6, 0.2):
      echo(t)
  Run

  will not print the 0.6 (on my machine) because of innacuracies.

---

• thomas (orginal) [2015-01-06T11:01:42+01:00] view original

  Excellent, thanks for that summary.

---