Python：用for循环创建二维NumPy数组时遇IndexError问题求助

Let's break down why you're hitting this IndexError and how to fix it step by step.

Why the Error Happens #

• You initialize result_array = np.array([]) as an empty NumPy array. On the first iteration of your outer loop (h=0), you try to access result_array[h]—but an empty array has zero elements, so index 0 is completely out of bounds. That's exactly what the error message is flagging.
• Your use of np.append here is also flawed: np.append expects the original array and the element/array to add, but trying to index into an empty array first breaks the entire operation before the append can even run.

Fix Steps & Corrected Code #

Here's how to adjust your code to build the 2D array correctly, plus some cleanup to remove redundant work:

1. Use a list for iterative appends instead of an empty NumPy array
   Lists are far more efficient for repeated appends, and we can convert to a NumPy array once all data is collected.
2. Eliminate redundant re-initializations
   Variables like xp and zp don't need to be redefined in every inner loop—this just wastes processing time.
3. Simplify inner loop logic
   You don't need the separate inter_result array; just append each calculated nanmax(gz) directly to inter2d_result.
4. Append full 1D arrays to the result list
   Skip indexing into the result during appends—just add each completed inter2d_result to the list, then convert to a 2D NumPy array at the end.

Corrected Code #

import numpy as np
# Assuming mesher and talwani modules are imported correctly

density = 1000
result_list = []  # Use list for efficient iterative appends
visual_x = np.array([])
# Pre-define xp and zp once instead of reinitializing every loop
xp = np.arange(-10000, 10000, 10.0)
zp = np.zeros_like(xp)

for h in range(3):
    # Reset polygon points for each outer loop iteration
    current_points = np.array([[-100, 0], [100, 0], [100, 300],[-100,300]])
    inter2d_result = np.array([])
    
    for i in range(10):
        poly = [mesher.Polygon(current_points, {'density': density})]
        gz = talwani.gz(xp, zp, poly)
        # Append the max value directly to our 1D result array
        inter2d_result = np.append(inter2d_result, np.nanmax(gz))
        visual_x = np.append(visual_x, current_points[1][0] * 2)
        
        # Update polygon x-coordinates for next inner loop iteration
        current_points[0][0] -= 10
        current_points[1][0] += 10
        current_points[2][0] += 10
        current_points[3][0] -= 10
    
    # Add the completed 1D array to our result collection list
    result_list.append(inter2d_result)

# Convert the list of 1D arrays to a single 2D NumPy array
result_array = np.array(result_list)

Key Notes #

• The final result_array will have a shape of (3, 10), exactly matching your desired structure [[inter2d_resultFIRST],[inter2d_resultSECOND],[inter2d_resultTHIRD]].
• Using a list for result_list avoids the overhead of repeated NumPy array concatenations, which are less efficient than list appends.
• We use current_points inside the outer loop to reset the polygon coordinates cleanly without modifying a global variable repeatedly.

内容的提问来源于stack exchange，提问作者alx