Rank a numeric column based on condition-CodePudding

I would like to rank a column called Number after grouping by the column ID but the ranking would be done only if Variable_1 is True. If Variable_1 is False, assign directly to the new column Rank the highest rank : 1

data={'ID':["A","A","B","C","D","D"], 'Variable_1':[True,True,False,False,True,True], 'Number': [10,20,5,6,100,90]}

df=pd.DataFrame(data)

I tried to do the following but did not work:

 df["Rank"] = np.where(df.Variable_1, df.groupby(["ID"])[["Number"]].rank(method='dense', ascending=False), 1)

This is the expected output:

CodePudding user response：

You can first set column Rank to 1 and use pandas.groupby and compute rank. At the end write rank for those rows that have Variable_1 == True with pandas.mask.

df['Rank'] = 1
res = df.groupby('ID')['Number'].rank(ascending=False)
df['Rank'] = df['Rank'].mask(df['Variable_1'], res)
print(df)

  ID  Variable_1  Number  Rank
0  A        True      10     2
1  A        True      20     1
2  B       False       5     1
3  C       False       6     1
4  D        True     100     1
5  D        True      90     2