python - 如果一个列字符串包含在另一个 Python 中，则一对多合并两个数据框

Question

我有两个数据框，我想根据 words 的 if 列值进行合并来自 df1包含 keywords 的列值来自 df2 .我一直在尝试使用 str.extract .但到目前为止还没有运气获得预期的结果。示例如下:

df1:

[{'id': 1, 'words': 'chellomedia', 'languages': nan},
 {'id': 2, 'words': 'Moien Welt!', 'languages': 'Luxemburgish'},
 {'id': 3, 'words': 'Ahoj světe!', 'languages': 'Czech'},
 {'id': 4, 'words': 'hello world', 'languages': nan},
 {'id': 5, 'words': '¡Hola Mundo!', 'languages': 'Spanish'},
 {'id': 6, 'words': 'hello kitty', 'languages': 'English'},
 {'id': 7, 'words': 'Ciao mondo!', 'languages': 'Italian'},
 {'id': 8, 'words': 'hola world', 'languages': nan}]

df2:

[{'code': 1, 'keywords': 'Hello'},
 {'code': 2, 'keywords': 'hola'},
 {'code': 3, 'keywords': 'world'}]

我的试用代码:

df1['words'] = df1['words'].str.lower()
df2['keywords'] = df2['keywords'].str.lower()

pat = '|'.join([re.escape(x) for x in df2.keywords])
df1.insert(0, 'keywords', df1['words'].str.extract('(' + pat + ')', expand=False))

pd.merge(df1, df2, on='keywords', how='left')

输出:

  keywords  id         words     languages  code
0    hello   1   chellomedia           NaN   1.0
1      NaN   2   moien welt!  Luxemburgish   NaN
2      NaN   3   ahoj světe!         Czech   NaN
3    hello   4   hello world           NaN   1.0
4     hola   5  ¡hola mundo!       Spanish   2.0
5    hello   6   hello kitty       English   1.0
6      NaN   7   ciao mondo!       Italian   NaN
7     hola   8    hola world           NaN   2.0

但是想要的应该是这样的:

  keywords  id         words     languages  code
0    hello   1   chellomedia           NaN   1.0
1      NaN   2   moien welt!  Luxemburgish   NaN
2      NaN   3   ahoj světe!         Czech   NaN
3    hello   4   hello world           NaN   1.0
4    world   4   hello world           NaN   3.0  ---> should be generated in df
5     hola   5  ¡hola mundo!       Spanish   2.0
6    hello   6   hello kitty       English   1.0
7      NaN   7   ciao mondo!       Italian   NaN
8     hola   8    hola world           NaN   2.0
9    world   8    hola world           NaN   3.0  ---> should be generated in df

我怎样才能产生预期的结果？谢谢。

Answer 1

你必须使用 findall 和 explode 而不是 extract，例如:

df1.insert(0, 'keywords', df1['words'].str.findall('(' + pat + ')'))
print(pd.merge(df1.explode('keywords'), df2, on='keywords', how='left')
        .sort_values('id').reset_index(drop=True))

输出:

  keywords  id         words     languages  code
0    hello   1   chellomedia           NaN   1.0
1      NaN   2   moien welt!  Luxemburgish   NaN
2      NaN   3   ahoj světe!         Czech   NaN
3    hello   4   hello world           NaN   1.0
4    world   4   hello world           NaN   3.0
5     hola   5  ¡hola mundo!       Spanish   2.0
6    hello   6   hello kitty       English   1.0
7      NaN   7   ciao mondo!       Italian   NaN
8    world   8    hola world           NaN   3.0
9     hola   8    hola world           NaN   2.0

和你需要的完全一样:)