Question

我尝试使用np.insert在给定索引处的数组中插入特定值。在我使用Numpy 1.12之前，代码运行正常但是使用新的Numpy 1.13.3时出现以下错误

ValueError: shape mismatch: value array of shape () could not be broadcast to indexing result of shape ()

我的代码：

intial_array= 1D numpy array
indices= 1D numpy array
values_to_insert= 1D numpy array

mt_new2=np.insert(intial_array, indices,values_to_insert)

此问题是否已知或有人知道如何解决此问题？

Answer 1

早期numpy可以根据需要复制values以符合索引大小：

>>> x = numpy.arange(10)
>>> numpy.insert(x,[1,3,4,5],[10,20])
array([ 0, 10,  1,  2, 20,  3, 10,  4, 20,  5,  6,  7,  8,  9])
>>> numpy.__version__
'1.12.0'

New numpy期望匹配尺寸：

In [81]: x = np.arange(10)

In [82]: np.insert(x, [1,3,4,5],[10,20])
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-82-382864de5db0> in <module>()
----> 1 np.insert(x, [1,3,4,5],[10,20])

/usr/local/lib/python3.5/dist-packages/numpy/lib/function_base.py in insert(arr, obj, values, axis)
   5085     slobj[axis] = indices
   5086     slobj2[axis] = old_mask
-> 5087     new[slobj] = values
   5088     new[slobj2] = arr
   5089 

ValueError: shape mismatch: value array of shape (2,) could not be broadcast to indexing result of shape (4,)

In [83]: np.insert(x, [1,3,4,5],[10,20,10,20])
Out[83]: array([ 0, 10,  1,  2, 20,  3, 10,  4, 20,  5,  6,  7,  8,  9])

看起来早期版本明确或隐含地使用resize

In [85]: np.insert(x, [1,3,4,5],np.resize([10,20,30],4))
Out[85]: array([ 0, 10,  1,  2, 20,  3, 30,  4, 10,  5,  6,  7,  8,  9])

numpy版本中的np.insert错误＆＃39; 1.13.3＆＃39;

1 个答案: