advanced patterns when using np.vstack with 1D arrays and dtype
I tried several approaches but none seem to work. I'm working with an unexpected behavior when using `np.vstack` to stack several 1D arrays. I'm working with NumPy version 1.24.3, and I expected it to maintain the dtype of my input arrays. However, the output seems to be changing the dtype to a more generic one, which is causing problems in subsequent calculations. For example, I have the following code: ```python import numpy as np arr1 = np.array([1, 2, 3], dtype=np.int32) arr2 = np.array([4.5, 5.5, 6.5], dtype=np.float64) arr3 = np.array([7, 8, 9], dtype=np.int32) result = np.vstack((arr1, arr2, arr3)) print(result) print(result.dtype) ``` I expected `result` to maintain the integer types of `arr1` and `arr3`, but the output shows: ``` [[1. 2. 3. ] [4.5 5.5 6.5] [7. 8. 9. ]] float64 ``` The dtype of the resulting array is `float64`, which I did not expect given that I have two integer arrays. This is problematic since I need the result to be consistent with integer operations. I've tried forcing the dtype by using `astype(np.int32)` on the result, but it raises a ValueError due to the presence of floats. Is there a way to preserve the dtype of the input arrays when stacking them with `np.vstack`? It seems like the behavior is due to NumPy promoting the types to ensure compatibility, but I'm not sure how to handle this without losing the integrity of my integer data. My development environment is Ubuntu. What am I doing wrong? This is part of a larger desktop app I'm building.