NumPy structured / regular array from ctypes pointer to array of structs

Question

Say I have the following C function:

void getArrOfStructs(SomeStruct** ptr, int* numElements)

And the following C struct:

typedef struct SomeStruct
{
    int x;
    int y;
};

I am able to successfully get a Python list:

class SomeStruct(Structure):
    _fields_ = [('x', c_int),
                ('y', c_int)]
ptr, numElements = pointer(SomeStruct()), c_int()
myDLL.getArrOfStructs(byref(ptr), byref(numElements)))

I want to get a NumPy structured / regular array.

1. Structured vs Regular array: which one is preferable (in terms of terminology)?
2. How can I do it? I'm looking for an efficient way (without copy each cell). I tried NumPy's frombuffer() functions, but was only able to use it with regular C arrays.

Answer 1

Views of numpy arrays share a data buffer

In [267]: x=np.arange(6).reshape(3,2)
In [268]: x.tostring()
Out[268]: b'\x00\x00\x00\x00\x01\x00\x00\x00\x02\x00\x00\x00\x03\x00\x00\x00\x04\x00\x00\x00\x05\x00\x00\x00'
In [269]: x.view('i,i')
Out[269]: 
array([[(0, 1)],
       [(2, 3)],
       [(4, 5)]], 
      dtype=[('f0', '<i4'), ('f1', '<i4')])

In this example the databuffer is a C array of 24 bytes, which can viewed in various ways - a flat array, 2 columns, or structured array with 2 fields.

I haven't worked with ctypes but I'm sure there's something equivalent to np.frombuffer to construct an array from a byte buffer.

In [273]: np.frombuffer(x.tostring(),int)
Out[273]: array([0, 1, 2, 3, 4, 5])
In [274]: np.frombuffer(x.tostring(),'i,i')
Out[274]: 
array([(0, 1), (2, 3), (4, 5)], 
      dtype=[('f0', '<i4'), ('f1', '<i4')])