Question

我在Python Pandas中有两个DataFrame。

存储在单元格中的数据如下：

DF1
- DatabaseId    Integer
- DatabaseName  String

DF2
- CreateString  String

我想将列DataBaseID应用于DF2中的任何记录，其中DF1.DatabaseName存在于Create String的上下文中。

Example:
DatabaseName = "UserDB"        CreateString = "This create string would fail"
DatabaseName = "UserDB"        CreateString = "This create string has UserDB in it"

第一条记录将失败，并且不会包含在结果集中。第二条记录会成功，并且会出现在结果集中。

我研究过各种选项，包括.isin和.contains，但这些选项都没有用。这似乎是一个“受控制的”笛卡尔联合，具有'如果匹配成功'的条件。但我无法找到一种方法来做到这一点，而且效率很高。

需要评估的总列表大小分别在100K到500K之间。

更新添加了更多示例数据：

>>> DF1.head(10)
DatabaseID     DatabaseName
0              DB1
1              DB2
2              DB3
3              DB4
...

>>> DF2.head(10)
CreateString
None
None
None
CREATE VIEW DB1.Table1 AS LOC…
None
REPLACE VIEW DB3.Table3...
CREATE VIEW DB3.Table10 AS SELE...
CREATE VIEW DB55.Table999 AS SELEC...
...

Desired Result
DatabaseID      DatabaseName        CreateText
0               DB1                 CREATE VIEW DB1.Table1 AS LOC…
2               DB3                 REPLACE VIEW DB3.Table3...
2               DB3                 CREATE VIEW DB3.Table10 AS SELE...
...
etc...
...

Answer 1

更新：如何解析表名：

In [100]: df2['TableName'] = df2.CreateString.str.extract('\s+(\w+\.\w+)\s+', expand=True)

In [101]: df2
Out[101]:
                            CreateString DatabaseName      TableName
0                                   None          NaN            NaN
1                                   None          NaN            NaN
2                                   None          NaN            NaN
3         CREATE VIEW DB1.Table1 AS LOC…          DB1     DB1.Table1
4                                   None          NaN            NaN
5            REPLACE VIEW DB3.Table3 ...          DB3     DB3.Table3
6     CREATE VIEW DB3.Table10 AS SELE...          DB3    DB3.Table10
7  CREATE VIEW DB55.Table999 AS SELEC...         DB55  DB55.Table999

原始回答：

你可以这样做：

In [83]: df2['DatabaseName'] = df2.CreateString.str.extract('\s+(\w+)\.\w+\s+', expand=True)

In [84]: pd.merge(df2, df1, on='DatabaseName', how='left')
Out[84]:
                            CreateString DatabaseName  DatabaseID
0                                   None          NaN         NaN
1                                   None          NaN         NaN
2                                   None          NaN         NaN
3         CREATE VIEW DB1.Table1 AS LOC…          DB1         0.0
4                                   None          NaN         NaN
5            REPLACE VIEW DB3.Table3 ...          DB3         2.0
6     CREATE VIEW DB3.Table10 AS SELE...          DB3         2.0
7  CREATE VIEW DB55.Table999 AS SELEC...         DB55         NaN

Pandas数据帧 - 基于“InStr”合并两个数据帧＆gt; 0

1 个答案: