在 Pandas 中分解多个不均匀的行

Question

我有两列，电话和电子邮件，需要分解成行。我已经弄清楚如何对其中之一执行此操作，但不能同时对两者执行此操作。最大的问题是我可能有0到很多电话和0到很多电子邮件。因此，如果客户有三封电子邮件但没有电话，那么我需要 3 行。如果他们有四部电话和三封电子邮件，那么我需要 4 行。每部电话一个，以及那四行中的三封电子邮件。 示例数据：

| many columns | phones | emails |
|:-------------|:------:|:-------|
| row 1        | A,B,C  | A,B    |
| row 2        |        | D,E,F  |

结果示例：

| many columns | phones | emails |
|:-------------|:------:|:-------|
| row 1        | A      | A      |
| row 1        | B      | B      |
| row 1        | C      |        |
| row 2        |        | D      |
| row 2        |        | E      |
| row 2        |        | F      |

# Convert cell contents into lists rather than strings
df0['phones'] = df0['phones'].str.split(";", expand=False)
df0['emails'] = df0['emails'].str.split(",", expand=False)
df0 = df0.apply(pd.Series.explode) # DOES NOT WORK

当我尝试上面的代码时，出现错误：

ValueError: cannot reindex on an axis with duplicate labels

Answer 1

我假设原始数据帧上的索引是唯一的。如果没有，请在以下代码段之前运行

df = df.reset_index()

：

columns = ["phones", "emails"]

# Explode each column individually, but instead of using `explode`, we will
# use`stack` to give us a second index level
exploded = [
    df[col].str.split(",", expand=True).stack().rename(col)
    for col in columns
]

# Align the exploded columns
exploded = pd.concat(exploded, axis=1).droplevel(-1)

# Merge it with the original data frame
result = pd.concat([df.drop(columns=columns), exploded], axis=1)

Answer 2

import itertools
import pandas as pd
import numpy as np
from pandas import DataFrame as df



df = pd.DataFrame({"x":[1,3,7],"y":["A","B","C"], 
                   "z":["p1,p2,p3","p4","p5,p6"],"package_code":["111,222,333","444","555,666"]})


print(df)
"""
   x  y         z package_code
0  1  A  p1,p2,p3  111,222,333
1  3  B        p4          444
2  7  C     p5,p6      555,666

"""
aa = (
    df.set_index(['x','y'])
    .apply(lambda col : pd.Series(col).str.split(','))
    .explode(['z','package_code'])
    .reset_index()
    .reindex(df.columns,axis=1)
    )
print(aa)

"""
   x  y   z package_code
0  1  A  p1          111
1  1  A  p2          222
2  1  A  p3          333
3  3  B  p4          444
4  7  C  p5          555
5  7  C  p6          666
"""