添加新项目时如何使用像熊猫面板一样的xarray

Question

我已将 pandas 面板转换为 xarray，但无法像使用 pandas 面板那样轻松地添加新项目、长轴和短轴。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'])

例如，如果我想添加一个新项目，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 中增加项目、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 分配不如 pandas 面板优雅。假设我们要在上面的数据数组中添加第四项。以下是它的工作原理：

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')

无论是在item还是major/minor轴上的操作，都是类似的逻辑。用于删除使用

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

Answer 2

xarray.DataArray 在内部基于单个 NumPy 数组，因此无法有效地调整大小或附加到它。您最好的选择是使用 xarray.concat 创建一个新的、更大的 DataArray。

如果您想将项目添加到pd.Panel，您可能正在寻找的数据结构是xarray.Dataset。这些是从相当于面板的多索引 DataFrame 中最容易构建的：

# 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 ...

有关从 pandas.Panel 过渡到 xarray 的更多信息，请阅读 xarray 文档中的此部分： http://xarray.pydata.org/en/stable/pandas.html#transitioning-from-pandas-panel-to-xarray