Conversion from Sage matrix to sympy matrix fails for matrices with rows full of zeros

[AlbertoJimenezRuiz]

Title: Conversion from Sage matrix to sympy matrix fails for matrices with rows full of zeros

Description: If a matrix created with the sage package features a row with only zeros, the constructor of ImmutableMatrix produces a matrix without those rows.

Steps to reproduce:
Run this code (needs sage):

from sage.all import Matrix as sage_matrix
from sympy.matrices import ImmutableMatrix

A=sage_matrix([[1,0],[0,0]])
# print(A)
rows_A=A.rows()
# print(rows_A)
matrix1 = ImmutableMatrix(rows_A)
print(matrix1)
matrix2 = ImmutableMatrix([(1, 0), (0, 0)])
print(matrix2)

• Output:

Matrix([[1, 0]])
Matrix([[1, 0], [0, 0]])

• Expected output:

Matrix([[1, 0], [0, 0]])
Matrix([[1, 0], [0, 0]])

The problem:
After investigating, the rows method of sage returns a list of sage.modules.vector_integer_dense.Vector_integer_dense. The vector_integer_dense class evaluates to false if all its elements are zero, while the python list type does so only if it is empty.

Possible fix:

The function _handle_creation_inputs of the class MatrixBase in the file sympy/sympy/matrices/matrixbase.py includes logic which discard rows that evaluate to false. Commenting out lines 3854-3855 fixes the issue.

class MatrixBase(Printable):
...
def _handle_creation_inputs(cls, *args, **kwargs):
...
    if not is_sequence(row) and \
            not getattr(row, 'is_Matrix', False):
        raise ValueError('expecting list of lists')

    if hasattr(row, '__array__'):
        if 0 in row.shape:
            continue
        #elif not row:
        #    continue

Workaround:

Converting the Vector_integer_dense as shown: rows_A = [list(x) for x in rows_A].

Tested Versions:

• OS: Ubuntu 24.04 x86_64
• python: Python 3.11.9 | packaged by conda-forge | (main, Apr 19 2024, 18:36:13) [GCC 12.3.0] on linux
• sympy: 1.12
• sage: 10.2

[oscarbenjamin]

The _handle_creation_inputs function is an abomination. It would have been better not to overload the constructor so much.

I don't know why there is a check to skip a row. It seems to come from gh-16462.

Removing the two indicated lines does not seem to cause any test failures. I think it seems like a reasonable change.