我有一个csv文件,其中包含两个列的日期和时间戳。我正在使用pandas read_csv
将内容读入数据帧。我的最终目标是根据数据绘制时间序列图。
!head vmstat.csv
wait_proc,sleep_proc,swapped_memory,free_memory,buffered_memory,cached_memory,swapped_in,swapped_out,received_block,sent_block,interrups,context_switches,user_time,sys_time,idle_time,wait_io_time,stolen_time,date,time
0,0,10896,3776872,380028,10284052,0,0,6,16,7716,4755,3,1,96,0,0,2012-11-01,08:59:27
0,0,10896,3776500,380028,10284208,0,0,0,40,7471,4620,0,0,99,0,0,2012-11-01,08:59:32
0,0,10896,3749840,380028,10286864,0,0,339,19,7479,4704,20,2,77,1,0,2012-11-01,08:59:37
0,0,10896,3747536,380028,10286964,0,0,17,118,7488,4638,0,0,99,0,0,2012-11-01,08:59:42
0,0,10896,3747452,380028,10287148,0,0,0,24,7489,4676,0,0,99,0,0,2012-11-01,08:59:47
df = read_csv("vmstat.csv", parse_dates=[['date','time']])
f = DataFrame(df, columns=[ 'date_time', 'user_time', 'sys_time', 'wait_io_time'])
In [3]: f
Out[3]:
date_time user_time sys_time wait_io_time
0 2012-11-01 08:59:27 3 1 0
1 2012-11-01 08:59:32 0 0 0
2 2012-11-01 08:59:37 20 2 1
3 2012-11-01 08:59:42 0 0 0
4 2012-11-01 08:59:47 0 0 0
到目前为止,我们可以正确读取数据,并在DataFrame中合并date_time
。如果我尝试使用date_time
中的df
作为索引,则会出现问题。指定index = df.date_time
会提供所有NaN
值:
dindex = f['date_time']
print dindex
g = DataFrame(f, columns=[ 'user_time', 'sys_time', 'wait_io_time'], index=dindex)
In [7]: g
Out[7]:
0 2012-11-01 08:59:27
1 2012-11-01 08:59:32
2 2012-11-01 08:59:37
3 2012-11-01 08:59:42
4 2012-11-01 08:59:47
Name: date_time <---- dindex
g:
user_time sys_time wait_io_time
date_time
2012-11-01 08:59:27 NaN NaN NaN
2012-11-01 08:59:32 NaN NaN NaN
2012-11-01 08:59:37 NaN NaN NaN
2012-11-01 08:59:42 NaN NaN NaN
2012-11-01 08:59:47 NaN NaN NaN
如您所见,列值以NaN
为单位显示。如何在中间f
框架中获得正确的值?
答案 0 :(得分:3)
您想使用set_index
:
df1 = df.set_index('date_time')
选择列'date_time'
作为新DataFrame的索引。
注意:您在DataFrame构造函数中遇到的行为演示如下:
df = pd.DataFrame([[1,2],[3,4]])
df1 = pd.DataFrame(df, index=[1,2])
In [3]: df1
Out[3]:
0 1
1 3 4
2 NaN NaN
答案 1 :(得分:0)
我可以通过以下代码获得解决方法:
up = f.pivot_table('user_time', rows='date_time')
sp = f.pivot_table('sys_time', rows='date_time')
wp = f.pivot_table('wait_io_time', rows='date_time')
u=pandas.DataFrame(up)
u['sys_time']=sp
u['wait_io_time']=wp
my_colors = ["#FF6666", "#00CC33", "#44EEEE"]
print u
输出:
user_time sys_time wait_io_time
date_time
2012-11-01 08:59:27 3 1 0
2012-11-01 08:59:32 0 0 0
2012-11-01 08:59:37 20 2 1
2012-11-01 08:59:42 0 0 0
2012-11-01 08:59:47 0 0 0
应该有更直接的方法来实现这一目标,但我是熊猫中的新人。
此外,u.plot()函数在绘制时间序列图时失败。 “AttributeError:'numpy.int64'对象没有属性'序数'”所以等待别人听取更好的解决方案。