Unexpected shape when for lambdifyed matrix

[StefanKaeser]

Hey,

I stumbled over something using lambdify on a matrix expression with version 1.5.1 and wanted to ask if this behavior is intended.

I wanted to lambdify a matrix and give numpy arrays as input for some symbols and constant ints for others.
As a minimal example I construct following matrix

import sympy
import numpy as np

a, b, c, d = sympy.symbols("a b c d")
matrix = sympy.Matrix([ [a, b], [c, d]])
matrix_fct = sympy.lambdify([a, b, c, d], matrix)

If I would now like to evaluate the matrix for multiple values of a but fixed b, c and d, I assumed that this expression

a_values = np.array([0, 1, 2, 3])
matrix_values = matrix_fct(a_values, 1, 2, 3)

would produce matrix_values as a (2, 2, 4) shapped numpy array. But instead I get

print(matrix_values)
print(matrix_values.shape)
>>>[[array([0, 1, 2, 3]) 1]
[2 3]]
>>> (2, 2)

To get the desired I need to use constant arrays

a_values = np.array([0, 1, 2, 3])
a_shape = a_values.shape
matrix_values = matrix_fct(a_values, 1 * np.ones(a_shape), 2 * np.ones(a_shape), 3 * np.ones(a_shape))

print(matrix_values)
print(matrix_values.shape)
>>>[[[0. 1. 2. 3.]
  [1. 1. 1. 1.]]

 [[2. 2. 2. 2.]
  [3. 3. 3. 3.]]]
>>>(2, 2, 4)

Is this behavior intended?
I find the array that I get for the first version highly unusable and don't see in what situation I would want it.

[sylee957]

Sympy lambdify is basically dispatching to `numpy.array([[a, b], [c, d]]), and if numpy is not broadcasting the scalars into ragged array by default, we shouldn't do that because it breaks other consistency.
I'm alternatively thinking of providing a fully symbolic API for matrix shape, because I don't think that there is any way to bind matrix shape into any code generation purposes unless it is accessed in implicit way.