Mirror of forum.nim-lang.org

1788 :: convert float64 binary to int64

• JamesWirth (orginal) [2015-11-10T18:24:34+01:00] view original

  I was looking for the nim equivalent to the go function math.Float64Bits and none of my searches turned up anything (bad choices of search, no doubt)

  Anyway the following worked - but I am wondering if there is a better NIM way:

  proc float64Bits(x:float64):int64 =
      var x = x
      var xi: int64
      moveMem(addr(xi),addr(x),8)
      return xi
  Run

  since using moveMem makes me nervous

---

• Varriount (orginal) [2015-11-10T21:12:12+01:00] view original

  If you want to transform one value to another while still keeping the same bit pattern, use a cast.

  
  let
    i = 5.4
    g = cast[float](i)
  
  
  Run

---

• jlp765 (orginal) [2015-11-11T03:04:55+01:00] view original

  Also try toBiggestInt()

---

• JamesWirth (orginal) [2015-11-11T21:43:06+01:00] view original

  The

  castint64

  is just what I wanted. Many thanx

  Does

  castsometype

  always just copy bits - even when sometype and othertype have different bit sizes?

---

• Varriount (orginal) [2015-11-11T21:57:44+01:00] view original

  It doesn't necessarily copy, just reinterprets the type. It will work with different bit sizes, but beware casting larger than the real type, you'll get odd behavior.

---

• JamesWirth (orginal) [2015-11-11T22:27:13+01:00] view original

  To be sure I understand, if there was:

  then x.addr and (cast[float64](x)).addr would be the same???

  Would

  result in cast[int32](w) having the value 515 - assuming little endian byte order? I realize that would be highly non-portable.

---