Working with Python arrays

Python has a builtin array module supporting dynamic 1-dimensional arrays of primitive types. It is possible to access the underlying C array of a Python array from within Cython. At the same time they are ordinary Python objects which can be stored in lists and serialized between processes when using multiprocessing.

Compared to the manual approach with malloc() and free(), this gives the safe and automatic memory management of Python, and compared to a Numpy array there is no need to install a dependency, as the array module is built into both Python and Cython.

Safe usage with memory views

from cpython cimport array
from array import array
cdef array.array a = array('i', [1, 2, 3])
cdef int[:] ca = a

print ca[0]

A Python array is constructed with a type signature and sequence of initial values. For the possible type signatures, refer to the Python documentation for the array module.

Notice that when a Python array is assigned to a variable typed as memory view, there will be a slight overhead to construct the memory view. However, from that point on the variable can be passed to other functions without overhead, so long as it is typed:

from cpython cimport array
from array import array
cdef array.array a = array('i', [1, 2, 3])
cdef int[:] ca = a

cdef int overhead(object a):
    cdef int[:] ca = a
    return ca[0]

cdef int no_overhead(int[:] ca):
    return ca[0]

print overhead(a)  # new memory view will be constructed, overhead
print no_overhead(ca)  # ca is already a memory view, so no overhead

Zero-overhead, unsafe access to raw C pointer

To avoid any overhead and to be able to pass a C pointer to other functions, it is possible to access the underlying contiguous array as a pointer. There is no type or bounds checking, so be careful to use the right type and signedness.

from cpython cimport array
from array import array

cdef array.array a = array('i', [1, 2, 3])

# access underlying pointer:
print a.data.as_ints[0]

from libc.string cimport memset
memset(a.data.as_voidptr, 0, len(a) * sizeof(int))

Cloning, extending arrays

To avoid having to use the array constructor from the Python module, it is possible to create a new array with the same type as a template, and preallocate a given number of elements. The array is initialized to zero when requested.

from cpython cimport array
from array import array

cdef array.array int_array_template = array('i', [])
cdef array.array newarray

# create an array with 3 elements with same type as template
newarray = array.clone(int_array_template, 3, zero=False)

An array can also be extended and resized; this avoids repeated memory reallocation which would occur if elements would be appended or removed one by one.

from cpython cimport array
from array import array

cdef array.array a = array('i', [1, 2, 3])
cdef array.array b = array('i', [4, 5, 6])

# extend a with b, resize as needed
array.extend(a, b)
# resize a, leaving just original three elements
array.resize(a, len(a) - len(b))