问题

我有1000个具有相同结构的数据框，但其中一些可能包含字符串值。我需要对所有这些数据框执行相同的计算，只是排除包含子字符串的行。带有字符串数据集结构的示例如下：

    time    x           y               z
    0.00  run_failed  run_failed   run_failed
    0.02  run_failed  run_failed   run_failed
    0.03  test_failed test_failed  test_failed
    0.04  44            321         644
    0.04  44            321         644
    0.04  44            321         644
    0.03  test_failed test_failed  test_failed
    0.04  44            321         644
    0.04  44            321         644

当使用df.dtypes方法查看时，包含子字符串的数据框将始终是对象类型，而“正常”的数据框将是float64类型。

因此，为了处理这个问题，我创建了以下脚本：

for df in dfs:  
   add = 0
   z = pd.read_csv(df)
   if type(z["x"]) != np.float64:
        bmb = z[z['x'].str.contains('failed')]
        z = z.drop(bmb.index)
        add = len(bmb)
        print(add)
   # 然后是进行计算的代码，假设如果出现字符串，它在if语句内被删除

但当运行代码时，它返回错误："只能对字符串值使用 .str 访问器！"，指向if语句块内部，然而数据集完全是float64类型，为什么它尝试处理这个命令 "bmb = z[z['x'].str.contains('failed')]" 对我来说一点都不清楚。

【注意】：上述内容是您提供的信息的翻译，没有其他内容。

time    x           y               z
0.00  run_failed  run_failed   run_failed
0.02  run_failed  run_failed   run_failed
0.03  test_failed test_failed  test_failed
0.04  44            321         644
0.04  44            321         644
0.04  44            321         644
0.03  test_failed test_failed  test_failed
0.04  44            321         644
0.04  44            321         644

for df in dfs:  
   add = 0
   z = pd.read_csv(df)
   if type(z["x"]) != np.float64:
        bmb = z[z['x'].str.contains('failed')]
        z = z.drop(bmb.index)
        add = len(bmb)
        print(add)
   ....
   and then the code for doing calculations assuming that if string occured, it was dropped inside if statement

答案1

得分: 2

如果 z["x"] 的类型不是 np.float64：

这里你在获取 DataFrame 的*列*的类型，它是一个 `Series`，考虑以下简单的例子：

```python
import pandas as pd
df = pd.DataFrame({"x":[1,2,3]},dtype="int32")
print(type(df["x"]))

输出结果是：

<class 'pandas.core.series.Series'>

因此，你的条件始终成立（即相当于写 if True:）。如果你对持有值的类型感兴趣，请使用 .dtype 属性：

print(df["x"].dtype)

输出结果是：

int32

<details>
<summary>英文:</summary>

    if type(z[&quot;x&quot;]) != np.float64:

Here you are getting type of *column of DataFrame* which is `Series` consider following simple example

    import pandas as pd
    df = pd.DataFrame({&quot;x&quot;:[1,2,3]},dtype=&quot;int32&quot;)
    print(type(df[&quot;x&quot;]))

gives output

    &lt;class &#39;pandas.core.series.Series&#39;&gt;

Therefore your condition always holds (i.e. it is same as writing `if True:`). If you are interested in type of hold values use `.dtype` attribute

    print(df[&quot;x&quot;].dtype)

gives output

    int32





</details>



# 答案2
**得分**: 0

我想我解决了问题。首先，我将以下代码中的部分进行了更改：

if type(z["x"]) != np.float64

改为

z['x'].dtype == 'object'

在后续的计算中，当我将pandas列作为Series获取时，我添加了`.astype(float)`，因为似乎如果通过if语句处理了DataFrame，列中的值变为字符串，如下所示返回：

array(['0.00260045', '0.00257398', '0.00247482', ..., '0.02017634', '0.01997158','0.02019846'])

<details>
<summary>英文:</summary>

I think I figured the problem.
First, I changed 

    if type(z[&quot;x&quot;]) != np.float64
to

    z[&#39;x&#39;].dtype == &#39;object&#39;


and in further calculations, when I get pandas column as series, I added .astype(float) as seems like if dataframe was processed through if statement, values in the column become string, as they were returned as follows:

    array([&#39;0.00260045&#39;, &#39;0.00257398&#39;, &#39;0.00247482&#39;, ..., &#39;0.02017634&#39;, &#39;0.01997158&#39;,&#39;0.02019846&#39;])

</details>