如何在添加新项目时使用熊猫面板等xarray

Question

我已将pandas面板转换为xarray，但无法像熊猫面板一样轻松地添加新项目，主轴和副轴。代码如下：

import numpy as np 

import pandas as pd 

import xarray as xr 


panel = pd.Panel(np.random.randn(3, 4, 5), items=['one', 'two', 'three'], 
       major_axis=pd.date_range('1/1/2000', periods=4), 
       minor_axis=['a', 'b', 'c', 'd','e']) 

如果我想添加一个新的项目，例如

，我可以：

panel.four=pd.DataFrame(np.ones((4,5)),index=pd.date_range('1/1/2000', periods=4), columns=['a', 'b', 'c', 'd','e']) 

panel.four 

      a b c d e 
2000-01-01 1.0 1.0 1.0 1.0 1.0 

2000-01-02 1.0 1.0 1.0 1.0 1.0 

2000-01-03 1.0 1.0 1.0 1.0 1.0 

2000-01-04 1.0 1.0 1.0 1.0 1.0 

我很难在xarray增加项目，长轴/短轴

px=panel.to_xarray() 

#px gives me 
<xarray.DataArray (items: 3, major_axis: 5, minor_axis: 4)> 

array([[[-0.440081, -0.888226, 0.158702, 2.107577], 
     [ 0.917835, -0.174557, 0.501626, 0.116761], 
     [ 0.406988, 1.95184 , -1.345948, 2.960774], 
     [-1.905529, 0.25793 , 0.076162, 1.954012], 
     [ 0.499675, 1.87567 , -1.698771, -1.143766]], 


     [[ 0.070269, -1.151737, -0.344155, -0.506383], 
     [-2.199357, -0.040909, 0.491984, -0.333431], 
     [-0.113155, -0.668475, 2.366683, -0.421863], 
     [-0.567336, -0.302224, 1.638386, -0.038545], 
     [ 0.55067 , -0.409266, -0.27916 , -0.942144]], 


     [[ 1.269171, -0.151471, -0.664072, 0.269168], 
     [-0.486492, 0.59632 , -0.191977, 0.22537 ], 
     [ 0.069231, -0.345793, -0.450797, -2.982 ], 
     [-0.42338 , -0.849736, 0.965738, -0.544596], 
     [-1.455378, -0.256441, -1.204572, -0.347749]]]) 

Coordinates: 

    * items  (items) object 'one' 'two' 'three' 

    * major_axis (major_axis) datetime64[ns] 2000-01-01 2000-01-02 2000-01-03 ... 

    * minor_axis (minor_axis) object 'a' 'b' 'c' 'd' 


#how should I add a fourth item, increase/delete major axis, minor axis? 

Answer 1

xarray任务并不像熊猫小组那样优雅。假设我们想在上面的数据数组中添加第四项。下面是它如何工作的：

four=xr.DataArray(np.ones((1,4,5)), coords=[['four'],pd.date_range('1/1/2000', periods=4),['a', 'b', 'c', 'd','e']], 
        dims=['items','major_axis','minor_axis']) 

pxc=xr.concat([px,four],dim='items') 

无论是操作上的项目或长轴/短轴，类似的逻辑盛行。用于删除使用

pxc.drop(['four'], dim='items') 

Answer 2

xarray.DataArray是基于单个NumPy的阵列上在内部，因此它不能被有效地调整大小或附加。您最好的选择是用xarray.concat制作一个新的，更大的DataArray。

如果您想将项目添加到pd.Panel，那么您可能正在查看的数据结构为xarray.Dataset。这些是最容易从多索引数据帧相当于以构建一个面板：

# First, make a DataFrame with a MultiIndex 
>>> df = panel.to_frame() 

>>> df.head() 
         one  two  three 
major  minor 
2000-01-01 a  0.278958 0.676034 -1.544726 
      b  -0.918150 -2.707339 -0.552987 
      c  0.023479 0.175528 -0.817556 
      d  1.798001 -0.142016 1.390834 
      e  0.256575 0.265369 -1.829766 

# Now, convert the DataFrame with a MultiIndex to xarray 
>>> ds = df.to_xarray() 

>>> ds 
<xarray.Dataset> 
Dimensions: (major: 4, minor: 5) 
Coordinates: 
    * major (major) datetime64[ns] 2000-01-01 2000-01-02 2000-01-03 2000-01-04 
    * minor (minor) object 'a' 'b' 'c' 'd' 'e' 
Data variables: 
    one  (major, minor) float64 0.279 -0.9182 0.02348 1.798 0.2566 2.41 ... 
    two  (major, minor) float64 0.676 -2.707 0.1755 -0.142 0.2654 ... 
    three (major, minor) float64 -1.545 -0.553 -0.8176 1.391 -1.83 ... 

# You can assign a DataFrame if it has the right column/index names 
>>> ds['four'] = pd.DataFrame(np.ones((4,5)), 
...       index=pd.date_range('1/1/2000', periods=4, name='major'), 
...       columns=pd.Index(['a', 'b', 'c', 'd', 'e'], name='minor')) 

# or just pass a tuple directly: 
>>> ds['five'] = (('major', 'minor'), np.zeros((4, 5))) 

>>> ds 
<xarray.Dataset> 
Dimensions: (major: 4, minor: 5) 
Coordinates: 
    * major (major) datetime64[ns] 2000-01-01 2000-01-02 2000-01-03 2000-01-04 
    * minor (minor) object 'a' 'b' 'c' 'd' 'e' 
Data variables: 
    one  (major, minor) float64 0.279 -0.9182 0.02348 1.798 0.2566 2.41 ... 
    two  (major, minor) float64 0.676 -2.707 0.1755 -0.142 0.2654 ... 
    three (major, minor) float64 -1.545 -0.553 -0.8176 1.391 -1.83 ... 
    four  (major, minor) float64 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 ... 
    five  (major, minor) float64 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 

更多关于从pandas.Panel过渡到xarray，在xarray文档阅读本节： http://xarray.pydata.org/en/stable/pandas.html#transitioning-from-pandas-panel-to-xarray