Python 使用Pandas读取csv时间数据时数据类型不一致
我正在使用Pandas读取带有时间数据的csv文件。我注意到时间戳的数据格式因时区而异。我不是这里的专家,所以我可能犯了一个错误。这里有一个简单的例子来说明我的意思 我有两个csv文件:data1.csv:Python 使用Pandas读取csv时间数据时数据类型不一致,python,pandas,Python,Pandas,我正在使用Pandas读取带有时间数据的csv文件。我注意到时间戳的数据格式因时区而异。我不是这里的专家,所以我可能犯了一个错误。这里有一个简单的例子来说明我的意思 我有两个csv文件:data1.csv: Timestamp,State 2020-05-26T10:00:00+01:00,3 2020-05-26T10:10:00+00:00,1 和data2.csv: Timestamp,State 2020-05-26T10:00:00+00:00,3 2020-05-26T10:10:
Timestamp,State
2020-05-26T10:00:00+01:00,3
2020-05-26T10:10:00+00:00,1
和data2.csv:
Timestamp,State
2020-05-26T10:00:00+00:00,3
2020-05-26T10:10:00+00:00,1
请注意,唯一的区别是第一行中的时区。当我读取第一个csv文件时,我得到的时间戳是Python datetimes(注意,我只查看最后一行,在这两种情况下时间戳是相同的):
所以这很好。但是,当我对data2.csv执行相同操作时,我得到
In [5]: df_2 = pd.read_csv('data2.csv', parse_dates=['Timestamp'])
In [6]: df_2['Timestamp'].values[1]
Out[6]: numpy.datetime64('2020-05-26T10:10:00.000000000')
In [7]: df_2.iloc[1].Timestamp
Out[7]: Timestamp('2020-05-26 10:10:00+0000', tz='UTC')
现在我们有了Numpy datetime64或时间戳,这取决于我们如何从数据帧中提取它们
格式不一致令人恼火。这是一个bug还是我做错了什么?这是pandas的一个弱点:它不能在本地表示带有混合时区的列。看见 详情请参阅。 与那里写的相反,我得到了混合时区列的python
datetime
类型(而不是string
),但它应该回答您的问题
import pandas as pd
import io
print(pd.__version__)
s1 = """Timestamp,State
2020-05-26T10:00:00+01:00,3
2020-05-26T10:10:00+00:00,1"""
s2 = """Timestamp,State
2020-05-26T10:00:00+00:00,3
2020-05-26T10:10:00+00:00,1"""
print('\n----- default:')
df1 = pd.read_csv(io.StringIO(s1), parse_dates=['Timestamp'])
print(df1, '\n', df1.applymap(type))
df2 = pd.read_csv(io.StringIO(s2), parse_dates=['Timestamp'])
print(df2, '\n', df2.applymap(type))
print('\n----- with date_parser:')
df1 = pd.read_csv(io.StringIO(s1), parse_dates=['Timestamp'], date_parser=lambda col: pd.to_datetime(col, utc=True))
print(df1, '\n', df1.applymap(type))
df2 = pd.read_csv(io.StringIO(s2), parse_dates=['Timestamp'], date_parser=lambda col: pd.to_datetime(col, utc=True))
print(df2, '\n', df2.applymap(type))
输出:
1.0.3
----- default:
Timestamp State
0 2020-05-26 10:00:00+01:00 3
1 2020-05-26 10:10:00+00:00 1
Timestamp State
0 <class 'datetime.datetime'> <class 'int'>
1 <class 'datetime.datetime'> <class 'int'>
Timestamp State
0 2020-05-26 10:00:00+00:00 3
1 2020-05-26 10:10:00+00:00 1
Timestamp State
0 <class 'pandas._libs.tslibs.timestamps.Timesta... <class 'int'>
1 <class 'pandas._libs.tslibs.timestamps.Timesta... <class 'int'>
----- with date_parser:
Timestamp State
0 2020-05-26 09:00:00+00:00 3
1 2020-05-26 10:10:00+00:00 1
Timestamp State
0 <class 'pandas._libs.tslibs.timestamps.Timesta... <class 'int'>
1 <class 'pandas._libs.tslibs.timestamps.Timesta... <class 'int'>
Timestamp State
0 2020-05-26 10:00:00+00:00 3
1 2020-05-26 10:10:00+00:00 1
Timestamp State
0 <class 'pandas._libs.tslibs.timestamps.Timesta... <class 'int'>
1 <class 'pandas._libs.tslibs.timestamps.Timesta... <class 'int'>
1.0.3
-----默认值:
时间戳状态
0 2020-05-26 10:00:00+01:00 3
1 2020-05-26 10:10:00+00:00 1
时间戳状态
0
1.
时间戳状态
0 2020-05-26 10:00:00+00:00 3
1 2020-05-26 10:10:00+00:00 1
时间戳状态
0
1.0.3
----- default:
Timestamp State
0 2020-05-26 10:00:00+01:00 3
1 2020-05-26 10:10:00+00:00 1
Timestamp State
0 <class 'datetime.datetime'> <class 'int'>
1 <class 'datetime.datetime'> <class 'int'>
Timestamp State
0 2020-05-26 10:00:00+00:00 3
1 2020-05-26 10:10:00+00:00 1
Timestamp State
0 <class 'pandas._libs.tslibs.timestamps.Timesta... <class 'int'>
1 <class 'pandas._libs.tslibs.timestamps.Timesta... <class 'int'>
----- with date_parser:
Timestamp State
0 2020-05-26 09:00:00+00:00 3
1 2020-05-26 10:10:00+00:00 1
Timestamp State
0 <class 'pandas._libs.tslibs.timestamps.Timesta... <class 'int'>
1 <class 'pandas._libs.tslibs.timestamps.Timesta... <class 'int'>
Timestamp State
0 2020-05-26 10:00:00+00:00 3
1 2020-05-26 10:10:00+00:00 1
Timestamp State
0 <class 'pandas._libs.tslibs.timestamps.Timesta... <class 'int'>
1 <class 'pandas._libs.tslibs.timestamps.Timesta... <class 'int'>