Question

我有2个大型数组（65000,5）和（65000,2），但出于某种原因，当我在内部访问它们的循环或函数时，尺寸会随机变化。下面是一个示例，我在for循环之前和for循环内部测试数组，结果不同。

for test_i in range(len(norm_data)):
    if len(norm_data[test_i]) != norm_row_length:
        print("len(norm_data[", test_i, "]) = ", len(norm_data[test_i]), " should be ", norm_row_length)
    if len(vector_data[test_i]) != vector_row_length:
        print("len(vector_data[", test_i, "]) = ", len(vector_data[test_i]), " should be ", vector_row_length)


for index in range(start, training_len):
    x_rows = []
    for prev_i in range(n_prev):

        for test_i in range(len(norm_data)):
            if len(norm_data[test_i]) != norm_row_length:
                print("in loop len(norm_data[", test_i,"]) = ", len(norm_data[test_i]), " should be ", norm_row_length)
            if len(vector_data[test_i]) != vector_row_length:
                print("in loop len(vector_data[", test_i,"]) = ", len(vector_data[test_i]), " should be ", vector_row_length)

这是输出的一部分：

in loop len(norm_data[ 0 ]) =  7  should be  5
in loop len(norm_data[ 1 ]) =  9  should be  5
in loop len(norm_data[ 2 ]) =  11  should be  5
in loop len(norm_data[ 3 ]) =  13  should be  5
in loop len(norm_data[ 4 ]) =  15  should be  5
in loop len(norm_data[ 5 ]) =  15  should be  5
in loop len(norm_data[ 6 ]) =  15  should be  5

编辑：我发现了这个问题。循环中的其余代码是修改数组。

row_index = index - (n_prev-prev_i)
x_row = norm_data[row_index]
x_row += vector_data[row_index]
x_rows.append(x_row)

这是通过引用从norm_data中获取行而不是复制它们并在之后修改它们。解决方案是复制行。

row_index = index - (n_prev-prev_i)
x_row = list(norm_data[row_index])
# or x_row = norm_data[row_index][:]
x_row += vector_data[row_index]
x_rows.append(x_row)

Answer 1

跟进我的评论，这是使用enumerate()的第一个循环的示例：

for i, row in enumerate(norm_data):
    if len(row) != norm_row_length:
        print("len(norm_data[{}]) = {} should be {}".format(i, len(row), norm_row_length))
    if len(vector_data[i]) != vector_row_length:
        print("len(vector_data[{}]) = {} should be {}".format(i, len(vector_data[i]), vector_row_length))

它还用format()替换了这些字符串组合。它更清晰，更Pythonic。

Answer 2

我发现了问题。循环中的其余代码是修改数组。

row_index = index - (n_prev-prev_i)
x_row = norm_data[row_index]
x_row += vector_data[row_index]
x_rows.append(x_row)

这是通过引用从norm_data中获取行而不是复制它们并在之后修改它们。解决方案是复制行。

row_index = index - (n_prev-prev_i)
x_row = list(norm_data[row_index])
# or x_row = norm_data[row_index][:]
x_row += vector_data[row_index]
x_rows.append(x_row)

数组更改循环或函数内的维度

2 个答案: