我有一个Log df，其中df有Description列。看起来像。
Description
Machine x : Turn off
Another action here
Another action here
Machine y : Turn off
Machine x : Turn on
Another action here
我只需要用"："拆分行
类似：
Description Machine Action
Machine x : Turn off Machine x Turn off
Another action here
Another action here
Machine y : Turn off Machine y Turn off
Machine x : Turn on Machine x Turn on
Another action here
我已经试过了：
s = df["Description"].apply(lambda x:x.split(":"))
df["Action"] = s.apply(lambda x: x[1])
df["Machine"] = s.apply(lambda x: x[0])
还有一些带有"startswith"的东西。
您可以将str.extract与合适的regex一起使用。这将查找:周围的所有值(同时剥离冒号周围的空格(：
df[['Machine', 'Action']] = df.Description.str.extract('(.*) : (.*)',expand=True)
>>> df
Description Machine Action
0 Machine x : Turn off Machine x Turn off
1 Another action here NaN NaN
2 Another action here NaN NaN
3 Machine y : Turn off Machine y Turn off
4 Machine x : Turn on Machine x Turn on
5 Another action here NaN NaN
# df[['Machine', 'Action']] = df.Description.str.extract('(.*) : (.*)',expand=True).fillna('')
给定一个数据帧
>>> df
Description
0 Machine x : Turn off
1 Another action here
2 Another action here
3 Machine y : Turn off
4 Machine x : Turn on
5 Another action here
我会通过Series.str.split(splitter, expand=True)来处理这个问题。
>>> has_colon = df['Description'].str.contains(':')
>>> df[['Machine', 'Action']] = df.loc[has_colon, 'Description'].str.split('s*:s*', expand=True)
>>> df
Description Machine Action
0 Machine x : Turn off Machine x Turn off
1 Another action here NaN NaN
2 Another action here NaN NaN
3 Machine y : Turn off Machine y Turn off
4 Machine x : Turn on Machine x Turn on
5 Another action here NaN NaN
如果您喜欢空字符串，可以通过替换NaN单元格
>>> df.fillna('')
Description Machine Action
0 Machine x : Turn off Machine x Turn off
1 Another action here
2 Another action here
3 Machine y : Turn off Machine y Turn off
4 Machine x : Turn on Machine x Turn on
5 Another action here
仅使用split和expand=True
df[['Machine', 'Action']] =df.Description.str.split(':',expand=True).dropna()
df
Description Machine Action
0 Machine x : Turn off Machine x Turn off
1 Another action here NaN NaN
2 Another action here NaN NaN
3 Machine y : Turn off Machine y Turn off
4 Machine x : Turn on Machine x Turn on
5 Another action here NaN NaN
使用pd.Series.str.extract函数和特定的正则表达式模式(覆盖:分隔符周围的潜在多个空格(：
In [491]: df
Out[491]:
Description
0 Machine x : Turn off
1 Another action here
2 Another action here
3 Machine y : Turn off
4 Machine x : Turn on
5 Another action here
In [492]: pd.concat([df, df.Description.str.extract('(?P<Machine>[^:]+)s+:s+(?P<Action>[^:]+)').fillna('')], axis=1)
Out[492]:
Description Machine Action
0 Machine x : Turn off Machine x Turn off
1 Another action here
2 Another action here
3 Machine y : Turn off Machine y Turn off
4 Machine x : Turn on Machine x Turn on
5 Another action here
StringMethods既有用又方便，但通常性能不佳
我建议使用默认构造函数和纯python字符串处理
df[['Machine', 'Action']] = pd.DataFrame([x.split(':') for x in df.Description]).dropna()
定时比.str访问器选项更好。
df = pd.concat([df]*1000)
%timeit pd.DataFrame([x.split(':') for x in df.Description]).dropna()
4.47 ms ± 252 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)
%timeit df.Description.str.split(':',expand=True).dropna()
14.9 ms ± 323 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)
%timeit df.Description.str.extract('(.*) : (.*)',expand=True)
16.6 ms ± 393 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)
%timeit pd.concat([df, df.Description.str.extract('(?P<Machine>[^:]+)s+:s+(?P<Action>[^:]+)').fillna('')], axis=1)
22.5 ms ± 448 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)
我的主张是：
msk = df.Description.str.contains(':')
df[['Machine', 'Action']] = df.Description.str.split(':', 1, expand=True).where(msk, '')
首先创建一个掩码-行可以接收非空值。
然后只对掩码为true的行执行实际替换。其他行(实际上是所有新列(接收一个空字符串。