将 pandas 数据框列导入为字符串而不是 int

问题描述

我想将以下 csv 作为字符串而不是 int64 导入.Pandas read_csv 自动将其转换为 int64，但我需要此列作为字符串.

I would like to import the following csv as strings not as int64. Pandas read_csv automatically converts it to int64, but I need this column as string.

ID
00013007854817840016671868
00013007854817840016749251
00013007854817840016754630
00013007854817840016781876
00013007854817840017028824
00013007854817840017963235
00013007854817840018860166


df = read_csv('sample.csv')

df.ID
>>

0   -9223372036854775808
1   -9223372036854775808
2   -9223372036854775808
3   -9223372036854775808
4   -9223372036854775808
5   -9223372036854775808
6   -9223372036854775808
Name: ID

不幸的是，使用转换器会得到相同的结果.

Unfortunately using converters gives the same result.

df = read_csv('sample.csv', converters={'ID': str})
df.ID
>>

0   -9223372036854775808
1   -9223372036854775808
2   -9223372036854775808
3   -9223372036854775808
4   -9223372036854775808
5   -9223372036854775808
6   -9223372036854775808
Name: ID

推荐答案

只是想重申这将适用于 pandas >= 0.9.1:

Just want to reiterate this will work in pandas >= 0.9.1:

In [2]: read_csv('sample.csv', dtype={'ID': object})
Out[2]: 
                           ID
0  00013007854817840016671868
1  00013007854817840016749251
2  00013007854817840016754630
3  00013007854817840016781876
4  00013007854817840017028824
5  00013007854817840017963235
6  00013007854817840018860166

我也在创建一个关于检测整数溢出的问题.

I'm creating an issue about detecting integer overflows also.

在此处查看解决方案:https://github.com/pydata/pandas/issues/2247

更新，因为它可以帮助他人:

Update as it helps others:

要将所有列作为str，可以这样做(来自评论):

To have all columns as str, one can do this (from the comment):

pd.read_csv('sample.csv', dtype = str)

要将大多数或选择性列作为str，可以这样做:

To have most or selective columns as str, one can do this:

# lst of column names which needs to be string
lst_str_cols = ['prefix', 'serial']
# use dictionary comprehension to make dict of dtypes
dict_dtypes = {x : 'str'  for x in lst_str_cols}
# use dict on dtypes
pd.read_csv('sample.csv', dtype=dict_dtypes)