Structures and unions

User defined types are created by decorating a class that stores the corresponding member annotations. Each of the types, struct and union, has its own decorator function, @cstruct and @cunion, respectively.

@cstruct(cls=None, /, *, alias=None, wrapped=True)

Class decorator that returns a C struct class that inherits from ctypes.Structure. The decorated class must annotate the struct members with valid types.

<name> : <type>

The type field takes all types from the basic types section, plus all pointer and array types derived from them. User defined types are also valid, together with ctypes types.

import fancytypes as ft

# 128-bit complex number as defined by the C99 standard
@ft.cstruct
class complex128:
    real : ft.real64 # Real part
    imag : ft.real64 # Imaginary part

# Sample struct that can store an N-dimensional array
@ft.cstruct
class ndarray:
    data : ft.pointer(ft.real64) # Pointer to the first data array element
    dim : ft.pointer(ft.int32) # Pointer to the first dimension array element
    ndim : ft.int32 # Number of dimensions

Parameters:

• alias (str, optional) – User defined alias for the type, default is an explicit list with the fields

• wrapped (bool, optional) – Flag to wrap the type or return a ctypes class, default is True

@cunion(cls=None, /, *, alias=None, wrapped=True)

Class decorator that returns a C union class that inherits from ctypes.Union. Union members are annotated following the same rules described in @cstruct.

Parameters:

• alias (str, optional) – User defined alias for the type, default is an explicit list with the fields

• wrapped (bool, optional) – Flag to wrap the type or return a ctypes class, default is True

Structure and union types with wrapped=True have a corresponding numpy.dtype and can be used as a datatype in NumPy arrays. Pointer types are first converted to ctypes.c_void_p, which NumPy interprets as unsigned integers. This conversion does not affect the actual structure or union type.