“总线错误:10"；尝试设置 pandas 列名称时 [英] "Bus error: 10" when trying to set pandas column name

问题描述

这是我正在尝试通过索引重命名熊猫列的代码:

Here is the code I'm executing trying to rename pandas column by index:

import pandas as pd

df = pd.read_csv('input.csv', dtype='unicode', delim_whitespace=True)
df.columns.values[2] = "id"
print(df)

我很确定这不是最好的方法，但是当我使用Python 3.5运行它时，我得到了:

I'm pretty sure this is not the best approach, but when I run this with Python 3.5 I get:

$ python3.5 test.py
Bus error: 10

这是我第一次看到这样的错误.没有回溯，只有这个输出字符串.

This is the first time I see an error like this. There is no traceback, just this output string.

Bus error: 10是什么意思?

这是input.csv的内容:

visitIp        userId   idSite
128.227.50.161   a        35
24.222.206.154   a        35
10.12.0.1        a        35
10.12.0.1        a        35
10.12.0.1        a        35
24.222.206.154   a        35

(使用pandas 0.17.1)

(Using pandas 0.17.1)

推荐答案

公共汽车错误发生于以下情况处理器无法访问无效的内存地址.

Bus error occurs when a processor can't access an invalid memory address.

df.columns是实例.任何更改它的操作实际上都会返回一个新对象.修改其元素是非法的，例如df.columns[2] = 'id'会引发异常.

df.columns is an instance if Index which is an immutable object in pandas. Any operation changing it, returns in fact a new object. Modifying its elements is illegal, for example df.columns[2] = 'id' would raise an exception.

您正在访问和修改索引的基础数据.实际上，不是直接数据，而是数据的numpy view，它可能是一个临时对象. (内部，Index.values是一个返回self._data.view(ndarray)的属性.)

You were accessing and modifying an underlying data of the index. Actually, not the data directly but a numpy view of the data, which could have been a temporary object. (Internally, Index.values is a property that returns self._data.view(ndarray).)

我也无法重现此行为，我也不知道到底发生了什么以及现在为什么起作用.在numpy C/cython代码中，这很可能是未定义的行为.

I couldn't reproduce this behaviour either and I don't know exactly what happened and why it now works. It can very well be an undefined behaviour in numpy C/cython code.

这篇关于“总线错误:10"；尝试设置 pandas 列名称时的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！