Mirror of forum.nim-lang.org

10817 :: The bool type

• sls1005 (orginal) [2023-12-31T10:07:46+01:00] view original

  I've seen that the current implementation uses C _Bool (or C++ bool) for a Nim bool, the C macro true (or the C++ boolean literal true) for Nim true and the C macro false (or the C++ boolean literal true) for Nim false. This can cause some problems:

  • A Nim bool is stated to be one byte in the offical documentation. However, a C _Bool is not required to be one byte in the C99 standard, and a C++ bool is also not required to be one byte. Nim now uses a static assertion to ensure that they are of the size of one byte, so I think this is not a major problem.
  • It's a well-known problem in C that the true and false macros are not defined as (_Bool)1 and (_Bool)0 but instead as 1 and 0, which means they are integral literals of int type instead of _Bool type. This only happens in C and not in C++, and could have been changed in C23.
  • Although I'm not sure about this, the C _Bool might have been deprecated and replaced with the type bool in C23.

  In addition to that, it's not stated in the documentation whether a Nim bool is the same as a C/C++ bool, so it's not clear if we should use the Nim bool to wrap a C/C++ function that takes or returns a C/C++ bool, or if we should use something like

  type
    CBoolImpl = distinct uint8
    cbool* {.importc: "bool", header: "<stdbool.h>".} = CBoolImpl
  
  let
    CTrue* {.importc: "((bool)(true))", header: "<stdbool.h>".}: cbool
    CFalse* {.importc: "((bool)(false))", header: "<stdbool.h>".}: cbool
  
  #This is incomplete. Don't actually use it.
  Run

  In the Cython language, there used to be a similar issue, which is solved by having different bool types: cpython.bool (which is the same as a bool in Python), libcpp.bool (which is the same as a bool in C++), and bint (which is the default, a bool in Python and an int in C). So I think having different bool types could be a viable way. Another way is to update the documentation.

---

• Amun (orginal) [2023-12-31T10:59:54+01:00] view original

  I only use this type internally in my C library wrappers. I export only Nim types. That's why I don't define C-style TRUE/FALSE at all. I use converters.

  type cbool = distinct byte
  
  converter from_cbool(b: cbool): bool = b != 0
  converter to_cbool(b: bool): cbyte = b.cbyte
  Run

  And you don't have to worry about C99 bool size constraints, the standard does not require an implementation to have even fixed width types. How many such implementations are still alive today? Close to none.

---

• Araq (orginal) [2023-12-31T11:50:10+01:00] view original

  The manual clearly states: "The size of the bool type is one byte."

  In practice it's always compatible with C++ and not with C. But that's not a big problem as C programmers do not understand types anyway and everything uses int (aka cint) instead.

---