Mirror of forum.nim-lang.org

3620 :: How to rewrite this C++ code to Nim properly

• maxcroud (orginal) [2018-03-08T20:53:12+01:00] view original

  Hello,

  I have some C++ code which context I want to re-create in Nim. Briefly, there is buffer of bytes. As it's just a pointer, I can easily designate an offset (current). Do not care about unpack_int, it's just a function combining some bytes from the buffer into an integer.

  The questions is how to make this buffer concept in Nim to be safe and efficient?

  I wanted to use slices, but they hold an immutuable sequence.

  const unsigned char* current, start, end;
  
  void reset(const void* data, int size) {
          start = (const unsigned char*) data;
          end = start + size;
          current = start;
  }
  
  int get_int() {
          if (current >= end) return 0;  // update
          int i;
          current = unpack_int(current, &i);
          return i;
          if (current > end) return 0;  // update
  }
  
  int unpack_int(const unsigned char *src, int *out) {
          int sign = (*src >> 6) & 1;
          *out = *src & 0x3F;
          do {
                  if(!(*src & 0x80)) break;
                  src++;
                  *out |= (*src & (0x7F)) << (6);
                  
                  if(!(*src & 0x80)) break;
                  src++;
                  *out |= (*src & (0x7F)) << (6+7);
                  
                  if(!(*src & 0x80)) break;
                  src++;
                  *out |= (*src & (0x7F)) << (6+7+7);
                  
                  if(!(*src & 0x80)) break;
                  src++;
                  *out |= (*src & (0x7F)) << (6+7+7+7);
          } while(0);
          
          src++;
          *out ^= -sign; // if (sign) *i = ~(*i)
          return src;
  }
  Run

---

• cdome (orginal) [2018-03-08T21:50:33+01:00] view original

  Code above does never check when end of buffer is hit, it is kind of important for design. Does it always ints in the buffer or something else as well?

---

• maxcroud (orginal) [2018-03-08T23:14:21+01:00] view original

  cdome, I slightly simplified the code and missed end of buffer checking, yes. Now it's updated. There are get_int, get_string and get_raw functions, there are not only ints in the buffer.

---

• mashingan (orginal) [2018-03-09T05:49:05+01:00] view original

  The use of ptr in Nim is always not safe.

  For pointer math operation, check this post: https://forum.nim-lang.org/t/1188#7366

  If it's sequence of bytes, you can use seq[Byte] or simply use cstring together with cast

---

• boia01 (orginal) [2018-03-09T16:53:05+01:00] view original

  I think you could find inspiration in the implementation of BitSets and IntSets since they both do low-level bit manipulation and operate on arrays of bytes. BitSets have a fixed memory size while IntSets are dynamically allocated/resized and use refs (GC'ed memory).

  https://github.com/nim-lang/Nim/blob/5f685bb0e6b2d68b9796e6a7d29469ce3f3b38ec/compiler/bitsets.nim https://github.com/nim-lang/Nim/blob/df4f707743879a0ea4363fcef446d89d8b421513/lib/pure/collections/intsets.nim

---