Mirror of forum.nim-lang.org

2657 :: Thin C wrappers.

• tgregory (orginal) [2016-12-09T18:51:49+01:00] view original

  Hello, I'm currently exploring Nim's FFI capabilities. I went through the both Manual and Tutorial but did not manage to find a detailed explanation on how various Nim types are layed out in memory.

  Here is what i do understand so far:

  • Adding {.pure.} pragma creates C struct directly mirroring Nim's object layout (without any hidden VT fields)
  • It is possible to have a C union via {.union.} pragma

  What I basically want to achieve is to create C and Nim type definitions which will have the same memory layout. Is it safe to assume the following type equality:

  
  C            Nim
  double       float64
  int          int
  size_t       int (with operations though unsigned module)
  int          bool
  Run

  Another problem is the Nim's sequence memory layout. The typical C code for a sequence like type (assuming straightforward memory allocation strategy) would be:

  struct arr {
    size_t len;
    int*   elements;
  };
  Run

  Is there any way to map such a struct to a sequence, so it could be passed between Nim and C without copying data?

---

• Arrrrrrrrr (orginal) [2016-12-09T20:29:57+01:00] view original

  You might want to checkout compatible types (those that start with 'c', for example, cint, cdouble, etc) http://nim-lang.org/docs/system.html.

---

• jlp765 (orginal) [2016-12-10T00:12:41+01:00] view original

  Look at manual.html#foreign-function-interface

  as it describes the .union pragma

---