如何使用 pandas.read_csv 将 CSV 文件中的数据插入数据框？

Question

我有一个 csv 文件，例如：

"B/G/213","B/C/208","WW_cis",,
"B/U/215","B/A/206","WW_cis",,
"B/C/214","B/G/207","WW_cis",,
"B/G/217","B/C/204","WW_cis",,
"B/A/216","B/U/205","WW_cis",,
"B/C/219","B/G/202","WW_cis",,
"B/U/218","B/A/203","WW_cis",,
"B/G/201","B/C/220","WW_cis",,
"B/A/203","B/U/218","WW_cis",,

我想将它读入数组或数据框之类的东西，这样我就可以将一列中的元素与另一列中的选定元素进行比较。起初，我使用numpy.genfromtxt 将它直接读入了一个数组，但我在'"B/A/203"' 之类的地方到处都带有额外的引号"。我在某处读到，pandas 允许剥离额外的 " 字符串，所以我尝试了：

class StructureReader(object):
    def __init__(self, filename):
        self.filename=filename
    def read(self):
        self.data=pd.read_csv(StringIO(str("RNA/"+self.filename)), header=None, sep = ",")
        self.data

但我得到了类似的东西：

<class 'pandas.core.frame.DataFrame'> 0 0 RNA/4v6p.csv

如何将我的 CSV 文件转换为某种允许我搜索列和行的数据类型？

Answer 1

数据插入

您将文件名的字符串放入您的DataFrame，即RNA/4v6p.csv 是您在位置row 0, col 0 中的数据。您需要读入文件并存储数据。这可以通过在您的班级中删除 StringIO(str(...)) 来完成

class StructureReader(object):
    def __init__(self, filename):
        self.filename = filename
    def read(self):
        self.data = pd.read_csv("RNA/"+self.filename), header=None, sep = ",")
        self.data

代码结构批判

我还建议将父目录从 硬编码 中删除

1. 总是传入完整的文件路径

   class StructureReader(object):
       def __init__(self, filepath):
           self.filepath = filepath
       def read(self):
           self.data = pd.read_csv(self.filepath), header=None, sep = ",")
           self.data
   

2. 使目录成为__init__() 参数

   class StructureReader(object):
       def __init__(self, directory, filename):
           self.directory = directory
           self.filename = filename
       def read(self):
           self.data=pd.read_csv(self.directory+"/"+self.filename), header=None, sep = ",")
           # or import os and self.data=pd.read_csv(os.path.join(self.directory, self.filename)), header=None, sep = ",")
           self.data
   

3. 使目录成为常量属性

   class StructureReader(object):
       def __init__(self, filename):
           self.directory = "RNA"
           self.filename = filename
       def read(self):
           self.data = pd.read_csv(self.directory+"/"+self.filename), header=None, sep = ",")
           # or import os and self.data=pd.read_csv(os.path.join(self.directory, self.filename)), header=None, sep = ",")
           self.data
   

这与阅读您的数据无关，只是关于构建代码的最佳实践评论（只是我的$0.02）。

Answer 2

IIUC，您可以阅读：

df = pd.read_csv('yourfile.csv', header=None)

对我来说返回：

         0        1       2   3   4
0  B/G/213  B/C/208  WW_cis NaN NaN
1  B/U/215  B/A/206  WW_cis NaN NaN
2  B/C/214  B/G/207  WW_cis NaN NaN
3  B/G/217  B/C/204  WW_cis NaN NaN
4  B/A/216  B/U/205  WW_cis NaN NaN
5  B/C/219  B/G/202  WW_cis NaN NaN
6  B/U/218  B/A/203  WW_cis NaN NaN
7  B/G/201  B/C/220  WW_cis NaN NaN
8  B/A/203  B/U/218  WW_cis NaN NaN

然后您可以只选择您想要的列：

df = df[[0,1,2]]

并像往常一样使用数据帧。

Answer 3

我认为您将 StringIO 与文件名混淆了。您要么将数据作为字符串，然后使用 StringIO，要么简单地指定文件名（not 使用 StringIO）：

In [189]: data="""\
   .....: "B/G/213","B/C/208","WW_cis",,
   .....: "B/U/215","B/A/206","WW_cis",,
   .....: "B/C/214","B/G/207","WW_cis",,
   .....: "B/G/217","B/C/204","WW_cis",,
   .....: "B/A/216","B/U/205","WW_cis",,
   .....: "B/C/219","B/G/202","WW_cis",,
   .....: "B/U/218","B/A/203","WW_cis",,
   .....: "B/G/201","B/C/220","WW_cis",,
   .....: "B/A/203","B/U/218","WW_cis",,
   .....: """

In [190]:

In [190]: df = pd.read_csv(io.StringIO(data), sep=',', header=None, usecols=[0,1,2])

In [191]: df
Out[191]:
         0        1       2
0  B/G/213  B/C/208  WW_cis
1  B/U/215  B/A/206  WW_cis
2  B/C/214  B/G/207  WW_cis
3  B/G/217  B/C/204  WW_cis
4  B/A/216  B/U/205  WW_cis
5  B/C/219  B/G/202  WW_cis
6  B/U/218  B/A/203  WW_cis
7  B/G/201  B/C/220  WW_cis
8  B/A/203  B/U/218  WW_cis

PS 你可以决定你想要解析哪些列（在你的数据框中） - 查看usecols 参数

或者使用文件名

import os

df = pd.read_csv(os.path.join('RNA', self.filename), sep=',', header=None, usecols=[0,1,2])