带有lambda的熊猫groupby在列表中

Question

I have following dataframe

df = pd.DataFrame({'ItemType': ['Red', 'White', 'Red', 'Blue', 'White', 'White', 'White', 'Green'], 
               'ItemPrice': [10, 11, 12, 13, 14, 15, 16, 17], 
               'ItemID': ['A', 'A', 'B', 'B', 'C', 'C', 'D', 'D']})

I would like get records (rows) with ItemIDs that contain only "White" ItemType in a form of a DataFrame

I have attempted following solution:

types = ['Red','Blue','Green']

~df.groupby('ItemID')['ItemType'].any().apply(lambda u: u in(types))

But this gives me an incorrect result (D should be False) and in a form of a series.

A False
B False
C True
D True

Thank you!

Answer 1

You should avoid using apply here, as it is usually quite slow. Instead, assign a flag column before you groupby , and then use all to assert that none of a groups values are in types :

df.assign(flag=~df.ItemType.isin(types)).groupby('ItemID').flag.all()

ItemID
A    False
B    False
C     True
D    False
Name: flag, dtype: bool

---

However, just to demonstrate the logic of the operation, and show what was incorrect about your approach, here is a working version using apply :

~df.groupby('ItemID').ItemType.apply(lambda x: any(i in types for i in x))

You need to use any inside the lambda, as opposed to on the Series before using apply .

---

To access rows where this condition is met, you may use transform :

df[df.assign(flag=~df.ItemType.isin(types)).groupby('ItemID').flag.transform('all')]

  ItemType  ItemPrice ItemID
4    White         14      C
5    White         15      C

Answer 2

An alternative method is to calculate an array of non-white ItemID values. Then filter your dataframe:

non_whites = df.loc[df['ItemType'].ne('White'), 'ItemID'].unique()

res = df[~df['ItemID'].isin(non_whites)]

print(res)

  ItemType  ItemPrice ItemID
4    White         14      C
5    White         15      C

You can also use GroupBy , but it's not absolutely necessary.