Python xarray.dataset.sel返回不需要的值

嗯，这很奇怪。我只希望得到第一个值，而不是一秒钟间隔内的所有值。我很想知道如何只选择想要的点

>>> wind.W_C.sel(Time = '2017-08-19T10:41:02.000000000', tolerance=0.0000000001).Time
<xarray.DataArray 'Time' (Time: 32)>
array(['2017-08-19T10:41:02.000000000', '2017-08-19T10:41:02.031250000',
   '2017-08-19T10:41:02.062500000', '2017-08-19T10:41:02.093750000',
   '2017-08-19T10:41:02.125000000', '2017-08-19T10:41:02.156250000',
   '2017-08-19T10:41:02.187500000', '2017-08-19T10:41:02.218750000',
   '2017-08-19T10:41:02.250000000', '2017-08-19T10:41:02.281250000',
   '2017-08-19T10:41:02.312500000', '2017-08-19T10:41:02.343750000',
   '2017-08-19T10:41:02.375000000', '2017-08-19T10:41:02.406250000',
   '2017-08-19T10:41:02.437500000', '2017-08-19T10:41:02.468750000',
   '2017-08-19T10:41:02.500000000', '2017-08-19T10:41:02.531250000',
   '2017-08-19T10:41:02.562500000', '2017-08-19T10:41:02.593750000',
   '2017-08-19T10:41:02.625000000', '2017-08-19T10:41:02.656250000',
   '2017-08-19T10:41:02.687500000', '2017-08-19T10:41:02.718750000',
   '2017-08-19T10:41:02.750000000', '2017-08-19T10:41:02.781250000',
   '2017-08-19T10:41:02.812500000', '2017-08-19T10:41:02.843750000',
   '2017-08-19T10:41:02.875000000', '2017-08-19T10:41:02.906250000',
   '2017-08-19T10:41:02.937500000', '2017-08-19T10:41:02.968750000'], 
dtype='datetime64[ns]')
Coordinates:
  * Time     (Time) datetime64[ns] 2017-08-19T10:41:02 ...
>>>

风力发电机（时间='2017-08-19T10:41:02.000000000'，公差=0.000000000 1）。时间 数组（['2017-08-19T10:41:02.000000000'，'2017-08-19T10:41:02.031250000'， “2017-08-19T10:41:02.062500000”，“2017-08-19T10:41:02.093750000”， “2017-08-19T10:41:02.125000000”、“2017-08-19T10:41:02.15650000”， “2017-08-19T10:41:02.187500000”，“2017-08-19T10:41:02.218750000”， “2017-08-19T10:41:02.250000000”、“2017-08-19T10:41:02.281250000”， “2017-08-19T10:41:02.312500000”，“2017-08-19T10:41:02.343750000”， “2017-08-19T10:41:02.375000000”，“2017-08-19T10:41:02.406250000”， “2017-08-19T10:41:02.437500000”、“2017-08-19T10:41:02.468750000”， “2017-08-19T10:41:02.500000000”、“2017-08-19T10:41:02.531250000”， “2017-08-19T10:41:02.562500000”，“2017-08-19T10:41:02.593750000”， “2017-08-19T10:41:02.62500000”，“2017-08-19T10:41:02.656250000”， “2017-08-19T10:41:02.687500000”、“2017-08-19T10:41:02.718750000”， “2017-08-19T10:41:02.75000000”，“2017-08-19T10:41:02.781250000”， “2017-08-19T10:41:02.812500000”，“2017-08-19T10:41:02.843750000”， “2017-08-19T10:41:02.875000000”，“2017-08-19T10:41:02.906250000”， “2017-08-19T10:41:02.937500000”、“2017-08-19T10:41:02.968750000”， dtype='datetime64[ns]'） 协调： *时间日期时间64[ns]2017-08-19T10:41:02。。。 >>> 编辑：对于任何不完全为x.0000000的时间值，它都能按预期工作（例如，使用time='2017-08-19T10:41:02.062500000'返回一个点）

编辑2：切片执行相同的操作：

>>> wind.W_C.sel(Time = slice('2017-08-19T10:41:01.999999999','2017-08-19T10:41:02.000000001'))
<xarray.DataArray 'W_C' (Time: 32)>
array([-0.365306, -0.351855, -0.301889, -0.315359, -0.340766, -0.366259,
   -0.361642, -0.318217, -0.346406, -0.310716, -0.188696, -0.206533,
   -0.228649, -0.219716, -0.259015, -0.348025, -0.251485, -0.251169,
   -0.29893 , -0.419135, -0.397326, -0.329589, -0.325526, -0.286838,
   -0.295504, -0.242547, -0.294295, -0.274641, -0.299403, -0.25828 ,
   -0.16487 , -0.317597], dtype=float32)
Coordinates:
  * Time     (Time) datetime64[ns] 2017-08-19T10:41:02 ...
>>>

>>风力发电选择（时间=slice（'2017-08-19T10:41:01.99999999'，'2017-08-19T10:41:02.00000000 1'））
数组（[-0.365306，-0.351855，-0.301889，-0.315359，-0.340766，-0.366259，
-0.361642, -0.318217, -0.346406, -0.310716, -0.188696, -0.206533,
-0.228649, -0.219716, -0.259015, -0.348025, -0.251485, -0.251169,
-0.29893 , -0.419135, -0.397326, -0.329589, -0.325526, -0.286838,
-0.295504, -0.242547, -0.294295, -0.274641, -0.299403, -0.25828 ,
-0.16487，-0.317597]，数据类型=32）
协调：
*时间日期时间64[ns]2017-08-19T10:41:02。。。
>>>

编辑3：必须是xarray中的错误。即使在指定公差时要小心：

>>> wind.W_C.sel(Time = '2017-08-19T10:41:02.0', tolerance = np.timedelta64(1,'us')).Time
<xarray.DataArray 'Time' (Time: 32)>
array(['2017-08-19T10:41:02.000000000', '2017-08-19T10:41:02.031250000',
   '2017-08-19T10:41:02.062500000', '2017-08-19T10:41:02.093750000',
   '2017-08-19T10:41:02.125000000', '2017-08-19T10:41:02.156250000',
   '2017-08-19T10:41:02.187500000', '2017-08-19T10:41:02.218750000',
   '2017-08-19T10:41:02.250000000', '2017-08-19T10:41:02.281250000',
   '2017-08-19T10:41:02.312500000', '2017-08-19T10:41:02.343750000',
   '2017-08-19T10:41:02.375000000', '2017-08-19T10:41:02.406250000',
   '2017-08-19T10:41:02.437500000', '2017-08-19T10:41:02.468750000',
   '2017-08-19T10:41:02.500000000', '2017-08-19T10:41:02.531250000',
   '2017-08-19T10:41:02.562500000', '2017-08-19T10:41:02.593750000',
   '2017-08-19T10:41:02.625000000', '2017-08-19T10:41:02.656250000',
   '2017-08-19T10:41:02.687500000', '2017-08-19T10:41:02.718750000',
   '2017-08-19T10:41:02.750000000', '2017-08-19T10:41:02.781250000',
   '2017-08-19T10:41:02.812500000', '2017-08-19T10:41:02.843750000',
   '2017-08-19T10:41:02.875000000', '2017-08-19T10:41:02.906250000',
   '2017-08-19T10:41:02.937500000', '2017-08-19T10:41:02.968750000'], dtype='datetime64[ns]')
Coordinates:
  * Time     (Time) datetime64[ns] 2017-08-19T10:41:02 ...
>>> 

风力发电机（时间='2017-08-19T10:41:02.0'，公差=np.timedelta64（1，'us'））。时间 数组（['2017-08-19T10:41:02.000000000'，'2017-08-19T10:41:02.031250000'， “2017-08-19T10:41:02.062500000”，“2017-08-19T10:41:02.093750000”， “2017-08-19T10:41:02.125000000”、“2017-08-19T10:41:02.15650000”， “2017-08-19T10:41:02.187500000”，“2017-08-19T10:41:02.218750000”， “2017-08-19T10:41:02.250000000”、“2017-08-19T10:41:02.281250000”， “2017-08-19T10:41:02.312500000”，“2017-08-19T10:41:02.343750000”， “2017-08-19T10:41:02.375000000”，“2017-08-19T10:41:02.406250000”， “2017-08-19T10:41:02.437500000”、“2017-08-19T10:41:02.468750000”， “2017-08-19T10:41:02.500000000”、“2017-08-19T10:41:02.531250000”， “2017-08-19T10:41:02.562500000”，“2017-08-19T10:41:02.593750000”， “2017-08-19T10:41:02.62500000”，“2017-08-19T10:41:02.656250000”， “2017-08-19T10:41:02.687500000”、“2017-08-19T10:41:02.718750000”， “2017-08-19T10:41:02.75000000”，“2017-08-19T10:41:02.781250000”， “2017-08-19T10:41:02.812500000”，“2017-08-19T10:41:02.843750000”， “2017-08-19T10:41:02.875000000”，“2017-08-19T10:41:02.906250000”， ‘2017-08-19T10:41:02.937500000’、‘2017-08-19T10:41:02.968750000’、dtype='datetime64[ns].） 协调： *时间日期时间64[ns]2017-08-19T10:41:02。。。 >>> 最终编辑：似乎是关于xarray解码datetime64字符串的方式，也许？这项工作：

>>> wind.W_C.sel(Time = np.datetime64('2017-08-19T10:41:02.000000000')).Time
<xarray.DataArray 'Time' ()>
array(1503139262000000000L, dtype='datetime64[ns]')
Coordinates:
    Time     datetime64[ns] 2017-08-19T10:41:02
>>>

风力发电机组（时间=np.datetime64（'2017-08-19T10:41:02.000000000'））。时间 数组（1503139262000000000L，dtype='datetime64[ns]'） 协调： 时间日期时间64[ns]2017-08-19T10:41:02 >>>

---

我认为这是一个上游问题。熊猫确实返回了同样的结果，

wind.W_C.to_Pandas.loc['2017-08-19T10:41:02.0']

给出了32个项目，但

wind.W_C.to_Pandas.loc[np.datetime64（'2017-08-19T10:41:02.0'）]

给出了一个项目。思考得很好。我在问题上加了一个熊猫标签。