python - 读取 csv(逗号分隔文件)，值在引号中，逗号作为小数点

Question

我有一个包含这样数据的文件:

    2.10.2014 23:30:00,"25,1",nan,nan,nan
    2.10.2014 23:30:00,nan,"15,2",nan,nan
    2.10.2014 23:30:00,nan,nan,"125,14",nan
    2.10.2014 23:45:00,nan,0,nan,nan

我想阅读此文件。期望的输出:

    2.10.2014 23:30:00 25.1  nan   nan     nan
    2.10.2014 23:30:00 nan   15.2  nan     nan
    2.10.2014 23:30:00 nan   nan   125.14  nan
    2.10.2014 23:45:00 nan   0     nan     nan

重要的是要注意，如果值 0 出现，引号就消失了。

此时我的代码如下所示:

     import pandas as pd
     import csv

     df=pd.read_csv("file.csv",
                    sep=',\s+',
                    quoting=csv.QUOTE_NONE, 
                    header=None, 
                    encoding="mbcs")

结果:

     "2.10.2014 23:30:00,""25,1"",nan,nan,nan"

代替 quoting=csv.QUOTE_NONE 我还尝试使用 escapechar='"'

Answer 1

将 decimal=',' 传递给 read_csv :

In [28]:
import io
import pandas as pd
t="""2.10.2014 23:30:00,"25,1",nan,nan,nan
    2.10.2014 23:30:00,nan,"15,2",nan,nan
    2.10.2014 23:30:00,nan,nan,"125,14",nan
    2.10.2014 23:45:00,nan,0,nan,nan"""
pd.read_csv(io.StringIO(t), decimal=',', header=None)

Out[28]:
                        0     1     2       3   4
0      2.10.2014 23:30:00  25.1   NaN     NaN NaN
1      2.10.2014 23:30:00   NaN  15.2     NaN NaN
2      2.10.2014 23:30:00   NaN   NaN  125.14 NaN
3      2.10.2014 23:45:00   NaN   0.0     NaN NaN

此外，您可以传递 parse_dates=[0] 将第一列解释为 datetime:

In [31]:
pd.read_csv(io.StringIO(t), decimal=',', header=None, parse_dates=[0])

Out[31]:
                    0     1     2       3   4
0 2014-02-10 23:30:00  25.1   NaN     NaN NaN
1 2014-02-10 23:30:00   NaN  15.2     NaN NaN
2 2014-02-10 23:30:00   NaN   NaN  125.14 NaN
3 2014-02-10 23:45:00   NaN   0.0     NaN NaN

在你的情况下忽略 io.StringIO 位，这只是为了让我从文本字符串加载你的数据，只需执行以下操作:

df=pd.read_csv("file.csv", sep=',\s+', quoting=csv.QUOTE_NONE, header=None, decimal=',', parse_dates=[0], encoding="mbcs")