我有一个表，其中包含错误数据需要修复。
df = pd.DataFrame({
'store_id' : list('aaaabbbbcccc'),
'product_id' : list('111122223333'),
'time_create' : (1,1,1,3,1,1,2,2,10,11,12,13),
'store_product_quantity_old' : (0,0,0,3,0,0,5,5, 0,1,2,3),
'store_product_quantity_new' : (1,1,1,5,2,3,4,10,1,2,3,4)
})
dups = df[df.duplicated(subset=['store_id', 'product_id', 'time_create'], keep=False)].copy()
dups.loc[:, 'quantity_diff'] = dups.store_product_quantity_new - dups.store_product_quantity_old
a = dups.groupby(['store_id', 'product_id', 'time_create']).agg({'quantity_diff': 'sum'} )
x = df.drop(df[df.duplicated(subset=['store_id', 'product_id', 'time_create'])].index)
x = x.set_index(['store_id', 'product_id', 'time_create'])
x.iloc[a.index].store_product_quantity_new = x.iloc[a.index].store_product_quantity_old + a.quantity_diff
但是在最后一步失败了：
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-129-0183d1586485> in <module>()
----> 1 x.iloc[a.index].store_product_quantity_new = x.store_product_quantity_old + a.quantity_diff
C:\Anaconda3\lib\site-packages\pandas\core\indexing.py in __getitem__(self, key)
1476
1477 maybe_callable = com._apply_if_callable(key, self.obj)
-> 1478 return self._getitem_axis(maybe_callable, axis=axis)
1479
1480 def _is_scalar_access(self, key):
C:\Anaconda3\lib\site-packages\pandas\core\indexing.py in _getitem_axis(self, key, axis)
2089 # a list of integers
2090 elif is_list_like_indexer(key):
-> 2091 return self._get_list_axis(key, axis=axis)
2092
2093 # a single integer
C:\Anaconda3\lib\site-packages\pandas\core\indexing.py in _get_list_axis(self, key, axis)
2068 axis = self.axis or 0
2069 try:
-> 2070 return self.obj._take(key, axis=axis)
2071 except IndexError:
2072 # re-raise with different error message
C:\Anaconda3\lib\site-packages\pandas\core\generic.py in _take(self, indices, axis, is_copy)
2787 new_data = self._data.take(indices,
2788 axis=self._get_block_manager_axis(axis),
-> 2789 verify=True)
2790 result = self._constructor(new_data).__finalize__(self)
2791
C:\Anaconda3\lib\site-packages\pandas\core\internals.py in take(self, indexer, axis, verify, convert)
4524 dtype='int64')
4525 if isinstance(indexer, slice)
-> 4526 else np.asanyarray(indexer, dtype='int64'))
4527
4528 n = self.shape[axis]
C:\Anaconda3\lib\site-packages\numpy\core\numeric.py in asanyarray(a, dtype, order)
542
543 """
--> 544 return array(a, dtype, copy=False, order=order, subok=True)
545
546
ValueError: setting an array element with a sequence.
答案 0 :(得分：1)
使用drop_duplicates并创建不包含iloc的新列：
dups = df[df.duplicated(subset=['store_id', 'product_id', 'time_create'], keep=False)].copy()
dups['quantity_diff'] = dups.store_product_quantity_new - dups.store_product_quantity_old
a = dups.groupby(['store_id', 'product_id', 'time_create']).agg({'quantity_diff': 'sum'})
x = df.drop_duplicates(subset=['store_id', 'product_id', 'time_create'])
x = x.set_index(['store_id', 'product_id', 'time_create'])
x['store_product_quantity_new'] = x.store_product_quantity_old + a.quantity_diff
print (x)
store_product_quantity_old \
store_id product_id time_create
a 1 1 0
3 3
b 2 1 0
2 5
c 3 10 0
11 1
12 2
13 3
store_product_quantity_new
store_id product_id time_create
a 1 1 3.0
3 NaN
b 2 1 5.0
2 5.0
c 3 10 NaN
11 NaN
12 NaN
13 NaN
为避免使用NaN，add与参数fill_value=0一起使用：
x['store_product_quantity_new'] = (x.store_product_quantity_old
.add(a.quantity_diff, fill_value=0))
print (x)
store_product_quantity_old \
store_id product_id time_create
a 1 1 0
3 3
b 2 1 0
2 5
c 3 10 0
11 1
12 2
13 3
store_product_quantity_new
store_id product_id time_create
a 1 1 3.0
3 3.0
b 2 1 5.0
2 5.0
c 3 10 0.0
11 1.0
12 2.0
13 3.0
答案 1 :(得分：0)
使用x.loc[a.index, 'store_product_quantity_new'] = x.store_product_quantity_old + a.quantity_diff
我尝试使用以下代码并成功运行。
import pandas as pd
df = pd.DataFrame(columns = ['store_id','product_id','time_create', 'store_product_quantity_old', 'store_product_quantity_new'])
df.loc[len(df)] = ['5aab11da-5dd2-477f-916c-3ed7e81ec03f', '460a49f8-7a8a-426d-8dec-d650d28035ee' , 1540305578301, 215,186]
df.loc[len(df)] = ['901b87fe-7a33-49ae-8730-de6f72167c8d' , '347ed0f2-423d-43b5-864a-654ebfab88e0' , 1540036103826, 10, 9]
df.loc[len(df)] = ['831b632e-12bd-4c23-a6fd-a18749d8d508' , 'c0e48f01-7d37-433e-8c82-66621a83be58' , 1540770907795, 0 , 20]
df.loc[len(df)] = ['31919fe6-bcef-483b-bc44-0fb2360993b2' , '2444245c-69d5-43ca-9138-2428acb368e0' , 1539659604914 , 90, 89]
df.loc[len(df)] = ['901b87fe-7a33-49ae-8730-de6f72167c8d' , '4614ce9a-52a2-42c4-9fda-2f200231531e' , 1538523837324 , 115, 114]
df.loc[len(df)] = ['8f3dfc01-3a82-4fbf-a681-94cc807b41a1' , '484f911a-f0d9-43a9-bcaf-ef5e67c8f64c' , 1539388385358 , 5 , 6]
df.loc[len(df)] = ['1148a913-7860-4525-b9c9-06c428baea4e' , '0e6ee8aa-f9f1-4541-ade7-04f3df6cbf71' ,1538355213073 , 171, 170]
df.loc[len(df)] = ['8d3527f7-8c25-4e47-8a8d-ddd6dcde439b' , 'cdfaa486-75ed-4be6-8457-057d8708142a' ,1539499795481 , 156, 155]
df.loc[len(df)] = ['9e7ff8bf-6aa4-4731-96d3-68c875f843f6' , '6f157811-0b18-4144-8400-311a299e2386' ,1538389333436 , 37 , 36]
df.loc[len(df)] = ['8dc2853b-ffe9-4dc8-9ad6-85622eed0c74' , 'f6b11510-f719-44d8-91a3-1b5cab1011ad' ,1539735238709 , 6 , 5]
df.loc[len(df)] = ['05a2d25c-d04f-4b32-8678-7c3b31d45fbb' , 'b7a2019a-eac1-4900-9c51-9576a77a0711' ,1538540129711 , 128, 127]
df.loc[len(df)] = ['05a2d25c-d04f-4b32-8678-7c3b31d45fbb' , 'b7a2019a-eac1-4900-9c51-9576a77a0711' ,1538540129711 , 129, 128]
df.loc[len(df)] = ['05a2d25c-d04f-4b32-8678-7c3b31d45fbb' , 'b7a2019a-eac1-4900-9c51-9576a77a0711' ,1538540129711 , 130, 129]
dups = df[df.duplicated(subset=['store_id', 'product_id', 'time_create'], keep=False)].copy()
dups['quantity_diff'] = dups['store_product_quantity_old'] - dups['store_product_quantity_new']
a = dups.groupby(['store_id', 'product_id', 'time_create']).agg({'quantity_diff': 'sum'} )
dups.loc[:, 'quantity_diff'] = dups.store_product_quantity_new - dups.store_product_quantity_old
x = df.drop(df[df.duplicated(subset=['store_id', 'product_id', 'time_create'])].index)
x = x.set_index(['store_id', 'product_id', 'time_create'])
x.loc[a.index, 'store_product_quantity_new'] = x.store_product_quantity_old + a.quantity_diff
print(x)
答案 2 :(得分：0)
哦，这是我的粗心。完成代码后，忘记的iloc用于选择整数索引，而整数索引不能用于多索引。因为loc在我的代码完成之前失败。这使我认为我的代码是错误的。
发布之前，我真的需要复习问题。
@jezrael的回答让我震惊，但是直接将x.store_product_quantity_old和a.quantity_diff加到x['store_product_quantity_new']上并不是我的好主意。所以我不接受他就投票。
更好的方法是将iloc更改为loc：
x.iloc[a.index].store_product_quantity_new = x.iloc[a.index].store_product_quantity_old + a.quantity_diff
到
x.loc[a.index].store_product_quantity_new = x.loc[a.index].store_product_quantity_old + a.quantity_diff
会工作的。
再次敲响自己，然后看到这个问题的人：
基于整数的索引，可以按位置进行选择。
通过标签或布尔数组访问一组行和列。 .loc []主要基于标签，但也可以与布尔数组一起使用。
如果您的熊猫代码有任何错误，请在发布问题之前先进行检查。