Add dtype property to arrays

[tomcaruso]

Description

I try to create a numpy array with np.zeros with the dtype given by another array. When I try to pyccelise this file pyccel return the issue Attribute dtype not found.

To Reproduce

import numpy as np
from pyccel.decorators import template
@template(name='T', types=['float[:]', 'complex[:]'])
def eval_jacobians_3d(n1: int, array: 'T'):
    arr_coeffs_x = np.zeros((n1), dtype=array.dtype)

Error details

ERROR at annotation (semantic) stage
pyccel:
 |fatal [semantic]: kernels.py [60,91]| Attribute dtype not found

Expected behavior

It should pyccelise it without crashing, I guess.

Language

Fortran

[EmilyBourne]

You are right, pyccel doesn't yet implement the dtype function or attribute.
We should create a class containing only a __new__ function which returns the dtype of an array using the dictionary DtypePrecisionToCastFunction. This new class should then be added as an attribute in the numpy array class description here:

pyccel/pyccel/ast/class_defs.py

Line 136 in 3dec5c0

NumpyArrayClass = ClassDef('numpy.ndarray',

While you are waiting for someone to pick up this issue, as a temporary workaround you can use:

import numpy as np
from pyccel.decorators import template
@template(name='T', types=['float[:]', 'complex[:]'])
def eval_jacobians_3d(n1: int, array: 'T'):
    arr_coeffs_x = np.zeros((n1), dtype=type(array[0]))