python - pandas dataframe，按另一列中的值排名，不使用 Python FOR 循环

Question

对于一列的每个值(在我下面的示例中为“日期”)，我想按另一列中的值(在我的示例中为“值”列)进行排名

我的代码可以运行，但我想知道是否可以在没有 Python for 循环的情况下完成

data = {'code': ['AAA', 'AAA', 'AAA', 'BBB', 'BBB', 'BBB', 'CCC', 'CCC', 'CCC'],
        'date': ['2001-01-01', '2001-01-02', '2001-01-03', '2001-01-01', '2001-01-02', '2001-01-03', '2001-01-01', '2001-01-02', '2001-01-03'],
        'value': [32, 23, 34, 23, 34, 12, 28, 39, 40]}
df = pd.DataFrame(data)
print(df)

pd.set_option('mode.chained_assignment', None)
result = pd.DataFrame()
for date in df['date'].unique():
    sub = df[df['date'] == date]
    sub['rank'] = len(sub) - sub['value'].rank() + 1
    result = result.append(sub[['code', 'date', 'rank']])
pd.set_option('mode.chained_assignment', 'warn')
df2 = pd.merge(df, result, on=['code', 'date'])

print(df2.sort_values(['date', 'code']))  # within each date, rows are ranked by value

  code        date  value  rank
0  AAA  2001-01-01     32   1.0
3  BBB  2001-01-01     23   3.0
6  CCC  2001-01-01     28   2.0
1  AAA  2001-01-02     23   3.0
4  BBB  2001-01-02     34   2.0
7  CCC  2001-01-02     39   1.0
2  AAA  2001-01-03     34   2.0
5  BBB  2001-01-03     12   3.0
8  CCC  2001-01-03     40   1.0

我能否在不通过 Python for 循环迭代的情况下获得相同的结果？

Answer 1

然后让我们用groupby做rank

df['rank'] = df.groupby('code')['value'].rank()
df
Out[491]: 
  code        date  value  rank
0  AAA  2001-01-01     32   2.0
1  AAA  2001-01-02     23   1.0
2  AAA  2001-01-03     34   3.0
3  BBB  2001-01-01     23   2.0
4  BBB  2001-01-02     34   3.0
5  BBB  2001-01-03     12   1.0
6  CCC  2001-01-01     28   1.0
7  CCC  2001-01-02     39   2.0
8  CCC  2001-01-03     40   3.0