如何编写代码将多维数组中满足条件的索引及后续元素置零并解决numpy属性错误

First, let's break down the issues in your original code:

1. numpy.ndarray (even with dtype="object") doesn't have an index() method—this is exactly why you're hitting the AttributeError.
2. Using l.index(i) is bug-prone: if a row has duplicate values, it will return the first occurrence of i instead of the current position you're checking.
3. You weren't breaking the inner loop after finding the first condition match, which leads to unnecessary (and potentially incorrect) repeated zeroing.

Here's the corrected code that addresses all these issues and produces your expected output:

import numpy as np

a = np.array([[12,45,50,-600,30,23.2,232,-44,12], 
              [45,50,60,30,23.2,232,-44,12,23], 
              [50,60,30,23.2,232,-44,12,34,12], 
              [60,30,23.2,232,-44,12,120,3,4], 
              [-300,23.2,232,-44,12,23,23,2,12], 
              [23.2,-232,-44,12,34,1,2,300.3,5]], dtype='object')
b = np.array([-12.4,-13.4,-44,-100,-6.3,-11,-10,-1.1,-0.3])

# Iterate over each row in array a
for row_idx, row in enumerate(a):
    # Check each element and its index in the current row
    for counter, val in enumerate(row):
        # Match the two conditions from your requirement
        if val <= 0 or b[counter] <= -100:
            # Zero out all elements from current index onwards
            row[counter:] = [0] * len(row[counter:])
            # Break the inner loop immediately—we only need the first matching position
            break

# Print the result as a list (matches your expected output format)
print(a.tolist())

Key Fixes & Improvements: #

• Replaced l.index(i) with counter: Since we're already using enumerate() to get the current element's index, we can directly use counter to target the correct position—no need to search for the index.
• Added break after zeroing: Once we find the first position that meets either condition, we stop checking the rest of the row (since those elements will already be zeroed out).
• Adjusted condition to val <= 0: Your requirement specifies "negative or zero"—the original code only checked for negative values (i < 0), so we fixed that to include zero.

Bonus: Faster Vectorized Approach (For Uniform Row Lengths) #

Since your rows all have the same length, you can ditch the object dtype and use numpy's vectorized operations for better performance (especially with large datasets):

import numpy as np

# Convert to float64 array (works because all rows are same length)
a = np.array([[12,45,50,-600,30,23.2,232,-44,12], 
              [45,50,60,30,23.2,232,-44,12,23], 
              [50,60,30,23.2,232,-44,12,34,12], 
              [60,30,23.2,232,-44,12,120,3,4], 
              [-300,23.2,232,-44,12,23,23,2,12], 
              [23.2,-232,-44,12,34,1,2,300.3,5]], dtype=np.float64)
b = np.array([-12.4,-13.4,-44,-100,-6.3,-11,-10,-1.1,-0.3], dtype=np.float64)

for row_idx in range(a.shape[0]):
    # Create a boolean mask for the two conditions
    condition_mask = (a[row_idx] <= 0) | (b <= -100)
    # Find the first position where the condition is True
    first_zero_pos = np.argmax(condition_mask)
    # If any position meets the condition, zero out from that point onward
    if condition_mask[first_zero_pos]:
        a[row_idx, first_zero_pos:] = 0

print(a.tolist())

This approach leverages numpy's optimized C-backed operations instead of Python loops, making it much faster for large arrays.

内容的提问来源于stack exchange，提问作者georgehere