Debugging Numerical/Tensor API Bugs: Aliasing, Axis Handling, and Truncation

You are given a set of pseudo APIs similar to numpy/tensor utilities. The code contains multiple bugs that cause incorrect outputs or unexpected behavior for some inputs. Your tasks: Identify root causes (including aliasing/shared storage, incorrect axis handling, and truncation of remaining elements in from_ndarray). Propose concrete fixes (implementation or correct usage). Provide minimal test cases (inputs + expected outputs) that reproduce and prevent regressions. Abstract API Matrix.zeros(shape): returns a zero matrix of given shape. to_ndarray(matrix, axis=?): converts a custom Matrix to an ndarray, optionally flattening/reordering by an axis. from_ndarray(arr, shape): converts an ndarray into a Matrix; may truncate/pad if element counts mismatch. Typical issues to look for Matrix.zeros may create shared rows (e.g., [[0]*n]*m), so updating one element affects multiple rows. to_ndarray may ignore or mishandle axis=1. from_ndarray may silently truncate remainders during reshape, causing data loss. State the expected behavior, what constitutes a bug, and how unit tests should catch it. You do not need to provide a full runnable project, but the fixes must be specific enough to implement. Example Input # Test aliasing in zeros m = Matrix.zeros((2,3)) m[0,0] = 1 print(m) Output [[1,0,0],[0,0,0]] # only the first row changes