What is this declaration in Cython? cdef PyObject **workers. Is it a pointer to a pointer?

user_2770149 0
Viewed 172

I'm trying to utilize the concepts in this sample code to run some Cython code in parallel, but I can't seem to find any information in the Cython documentation about what this notation actually means.

    cdef FLOAT_t[:] numbers
    cdef unsigned int i
    cdef INDEX_t n_workers
    cdef PyObject **workers
    cdef list ref_workers #Here to maintain references on Python side
    
    def __init__(Parent self, INDEX_t n_workers, list numbers):
        cdef INDEX_t i
        self.n_workers = n_workers
        self.numbers = np.array(numbers,dtype=float)
        self.workers = <PyObject **>malloc(self.n_workers*cython.sizeof(cython.pointer(PyObject)))
        
        #Populate worker pool
        self.ref_workers = []
        for i in range(self.n_workers):
            self.ref_workers.append(Worker())
            self.workers[i] = <PyObject*>self.ref_workers[i]
    
    def __dealloc__(Parent self):
        free(self.workers)

Does the ** notation mean that it is a pointer to a pointer of a PyObject? I understand that the <> notation is meant to dereference the pointer, so is this line:

self.workers = <PyObject **>malloc(self.n_workers*cython.sizeof(cython.pointer(PyObject)))

allocating an unknown amount of memory, since the size of the PyObject is unknown until self.workers is filled with dereferenced PyObjects?

cython python c++
1 Answers

Not only is it a pointer to a PyObject* pointer it's also the pointer to the first element of an array of PyObject* pointers.

You can see it is allocating memory to accommodate self.n_workers, presumably the workers are implemented using a PyObject derivative, so in memory you will have:

self.workers -> self.workers[0] (PyObject* for 1st worker)
                self.workers[1] (PyObject* for 2nd worker)
                ....
                self.workers[N-1]  (PyObject* for last worker)
user_16079370
Related