Mirror of forum.nim-lang.org

11543 :: Passing seq or array to cpp function that takes iterators.

• pablocael (orginal) [2024-05-04T09:18:27+02:00] view original

  Hi there,

  Im trying to export the std::nth_element function from cpp to nim by creating a wrapper function like this:

  
  {.emit:"""
  
  #include <algorithm>
  
  void nth_element(float* arr, int size, int n) {
    std::nth_element(arr, arr + n, arr + size);
  }
  """.}
  
  proc nth_element(array: ref object, size: int, n: int) {.importcpp: "nth_element(@)".}
  
  
  proc main() =
      var arr = newSeq[float](10)
      for i in 0 ..< arr.len:
          arr[i] = i.float32
      nth_element(arr, arr.len, 5)
      echo arr, " ", arr[5]
  
  main()
  Run

  However I have difficulty in understanding how to pass pointers between cpp and nim. In this case I want to pass an array buffer to the nth_element function by using any available strategy (in the example I have tried using pointers but I did not succeeded).

  Is there a way of doing that?

  Thanks in advance

---

• PMunch (orginal) [2024-05-04T09:41:08+02:00] view original

  Try passing a[0].addr and changing the type of your sequence to seq[cfloat]. Arrays in C/C++ are just a pointer to consecutive elements in memory. Nim sequences are also that, but with a couple fields to track length and capacity. So by getting the pointer to the first element you have a pointer to contiguous elements. Changing to cfloat is just to make sure that the float types used my C and the one used by Nim is the same. float in Nim is set to float64 while float in C is often 32 bits. This would lead to very strange output.

---

• pablocael (orginal) [2024-05-04T11:02:07+02:00] view original

  Thanks, this works:

  {.emit:"""

  #include <algorithm>

  void nth_element(float* arr, int size, int n) {

  std::nth_element(arr, arr + n, arr + size);

  } """.}

  proc nth_element*(array: ptr float32, size: int, n: int) {.importcpp: "nth_element(@)".}

  var arr = newSeqfloat32

  for i in 0 ..< arr.len:

  arr[i] = i.float32

  nth_element(addr arr[0], arr.len, 5)

  However if I try to export the nth_element function and import from another module, I have an error:

  .cache/nim/test1_d/@mtest1.nim.cpp: In function ‘void NimMainModule()’:

  .cache/nim/test1_d/@mtest1.nim.cpp:1249:25: error: ‘nth_element’ was not declared in this scope

  1249 | nth_element((&arr__test49_u28.p->data[((NI)0)]), T23, ((NI)5));

  | ^~~~~~~~~~~

  Error: execution of an external compiler program 'g++ -c -std=gnu++17 -funsigned-char -w -fmax-errors=3 -fpermissive -pthread -I/home/....../.choosenim/toolchains/nim-2.0.4/lib -I/home/....development/vex-core/tests -o /home/...../.cache/nim/test1_d/@mtest1.nim.cpp.o /home/...../.cache/nim/test1_d/@mtest1.nim.cpp' failed with exit code: 1

  it seems it cannot find the cpp declaration from the importing module.. how can I make that declaration available? Or should I compile this module first and import it compiled?

---

• demotomohiro (orginal) [2024-05-04T15:21:33+02:00] view original

  If you don't need to wrap nth_element in C++ code, following code can be imported in other module and nth_element can accept other types.

  proc nth_element[T](first, nth, last: ptr T) {.header: "<algorithm>", importcpp: "std::nth_element(@)".}
  
  proc nthElement*[T](x: var openArray[T]; nth: int) =
    nth_element(x[0].addr, x[nth].addr, cast[ptr T](cast[uint](x[^1].addr) + sizeof(T).uint))
  
  when isMainModule:
    var x = [4, 3, 2, 1]
    
    nthElement(x, 2)
    echo x
    
    var y = [5, 4, 3, 2, 1]
    nthElement(y.toOpenArray(1, 4), 2)
    echo y
  Run

---