如何在添加新項目時使用熊貓面板等xarray

Question

我已將pandas面板轉換爲xarray，但無法像熊貓面板一樣輕鬆地添加新項目，主軸和副軸。代碼如下：

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']) 

如果我想添加一個新的項目，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 

我很難在xarray增加項目，長軸/短軸

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任務並不像熊貓小組那樣優雅。假設我們想在上面的數據數組中添加第四項。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') 

無論是操作上的項目或長軸/短軸，類似的邏輯盛行。用於刪除使用

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

Answer 2

xarray.DataArray是基於單個NumPy的陣列上在內部，因此它不能被有效地調整大小或附加。您最好的選擇是用xarray.concat製作一個新的，更大的DataArray。

如果您想將項目添加到pd.Panel，那麼您可能正在查看的數據結構爲xarray.Dataset。這些是最容易從多索引數據幀相當於以構建一個面板：

# 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 ... 

更多關於從pandas.Panel過渡到xarray，在xarray文檔閱讀本節： http://xarray.pydata.org/en/stable/pandas.html#transitioning-from-pandas-panel-to-xarray