如何优雅地交错合并NumPy数组？现有实现可优化吗？

Great question! You're right that manually listing each row isn't scalable or elegant—here are three cleaner, more efficient approaches to achieve your desired interleaved array: #

Method 1: List Comprehension + Concatenate (Readable & Scalable)

This approach builds small pairs of [x, row] for each row in y, then concatenates them all together. It’s easy to read and works for any number of rows in y:

import numpy as np
x = np.array([1,2,3,4,5])
y = np.array([[4,6,2,6,9], [5,9,8,7,4], [3,2,5,4,9]])

result = np.concatenate([[x], row] for row in y)

Method 2: Slice Assignment (Most Efficient)

Preallocate an empty array of the correct shape, then use slice indexing to fill in x and y rows in one go. This is ideal for large datasets since it avoids extra intermediate arrays:

result = np.empty((2 * len(y), x.size), dtype=x.dtype)
result[::2] = x  # Fill even-indexed rows (0,2,4) with x
result[1::2] = y # Fill odd-indexed rows (1,3,5) with y's rows

Method 3: Tile + Stack + Reshape (Vectorized Operation)

Repeat x to match the number of rows in y, stack it with y along a new axis, then reshape to flatten the pairs into rows:

repeated_x = np.tile(x, (len(y), 1))  # Shape: (3,5)
combined = np.stack([repeated_x, y], axis=1)  # Shape: (3,2,5)
result = combined.reshape(-1, x.size)  # Shape: (6,5)

All three methods will produce your desired output:

array([[1, 2, 3, 4, 5],
       [4, 6, 2, 6, 9],
       [1, 2, 3, 4, 5],
       [5, 9, 8, 7, 4],
       [1, 2, 3, 4, 5],
       [3, 2, 5, 4, 9]])

Content of the question来源于stack exchange，提问作者mocs