如何基于向量实现n维数组左上角子数组的自动切片？

Great question! When working with n-dimensional arrays (like the NumPy-style array in your example), you don’t need to manually write out slice ranges for every dimension—Python’s unpacking and generator expressions make this trivial and scalable.

Core Idea #

The key is to convert your vector v_n into a tuple of slice objects, where each slice corresponds to 0:v_n[i] for the i-th dimension. NumPy arrays (and most array-like libraries) accept tuples of slices as indexers, which lets you dynamically generate the slice for any number of dimensions.

Step-by-Step Implementation #

Let’s use NumPy (since your example aligns with its syntax) to demonstrate:

1. First, import NumPy and define your n-dimensional array and dimension vector:

import numpy as np

# Example 3D array (adjust dimensions as needed)
A_3 = np.arange(5*6*7).reshape(5, 6, 7)
# Vector specifying the upper bounds for each dimension
v_3 = [2, 3, 4]

2. Generate the dynamic slice tuple:

# Create a slice(0, v) for each value in v_n, then pack into a tuple
slice_indices = tuple(slice(0, dim_size) for dim_size in v_3)

3. Apply the slice to your array:

sublist = A_3[slice_indices]

4. Verify the result—its shape will match v_n:

print(sublist.shape)  # Output: (2, 3, 4)

Testing with Your Original 2D Example #

For your original 2D array A and vector v=[10, 3], this method works exactly as expected:

# Your original 2D array
A = np.array([
    [1.,2.,3.,0.,0.,0.,0.],
    [4.,5.,6.,0.,0.,0.,0.],
    [7.,8.,9.,0.,0.,0.,0.],
    [10.,11.,12.,0.,0.,0.,0.],
    [13.,14.,15.,0.,0.,0.,0.],
    [16.,17.,18.,0.,0.,0.,0.],
    [19.,20.,21.,0.,0.,0.,0.],
    [22.,23.,24.,0.,0.,0.,0.],
    [25.,26.,27.,0.,0.,0.,0.],
    [28.,29.,30.,0.,0.,0.,0.],
    [0.,0.,0.,0.,0.,0.,0.],
    [0.,0.,0.,0.,0.,0.,0.],
    [0.,0.,0.,0.,0.,0.,0.],
    [0.,0.,0.,0.,0.,0.,0.]
])
v = [10, 3]

slice_indices = tuple(slice(0, dim) for dim in v)
result = A[slice_indices]
# Result is exactly the 10x3 upper-left subarray you wanted

Optional: Add Dimension Validation #

To avoid errors if v_n doesn’t match the number of dimensions in A_n, add a quick check:

if len(v_n) != A_n.ndim:
    raise ValueError("Vector v_n must have the same length as the number of dimensions in A_n")

Why This Works #

NumPy’s array indexing system natively supports tuple-based indexers. By generating a slice for each dimension dynamically, you avoid writing repetitive code like A_n[0:v1, 0:v2, 0:v3, ...]—this scales seamlessly to 1D, 2D, 10D, or any number of dimensions.

内容的提问来源于stack exchange，提问作者Jack Rolph