如何折叠 pandas 中的空值列?

how to collapse columns in pandas on null values?

假设我有以下数据框:

pd.DataFrame({'col1':    ["a", "a", np.nan, np.nan, np.nan],
            'override1': ["b", np.nan, "b", np.nan, np.nan],
            'override2': ["c", np.nan, np.nan, "c", np.nan]})


    col1    override1   override2
0     a        b          c
1     a       NaN        NaN
2     NaN      b         NaN
3     NaN     NaN         c
4     NaN     NaN         NaN

有没有办法将 3 列合并为一列,其中 override2 覆盖 override1,后者覆盖 col1,但是,如果存在 NaN,则值bofore 是要保留的吗?另外,我主要是在寻找一种方法,这样我就不必再增加一个专栏了。我真的在寻找内置的 pandas 解决方案。

这是我正在寻找的输出:

 collapsed
0  c
1  a
2  b
3  c
4  NaN

这是一种方法:

df.lookup(df.index , df.notna().cumsum(1).idxmax(1))
# array(['c', 'a', 'b', 'c', nan], dtype=object)

或者等效地使用底层 numpy 数组,并更改 idxmax with ndarray.argmax:

df.values[df.index, df.notna().cumsum(1).values.argmax(1)]
# array(['c', 'a', 'b', 'c', nan], dtype=object)
import pandas as pd
import numpy as np
df=pd.DataFrame({'col1':    ["a", "a", np.nan, np.nan, np.nan],
            'override1': ["b", np.nan, "b", np.nan, np.nan],
            'override2': ["c", np.nan, np.nan, "c", np.nan]})

print(df)
df=df['col1'].fillna('') + df['override1'].fillna('')+ df['override2'].fillna('')
print(df)

一个简单的解决方案涉及向前填充和选择最后一列。评论中提到了这一点。

df.ffill(1).iloc[:,-1].to_frame(name='collapsed')

  collapsed
0         c
1         a
2         b
3         c
4       NaN

如果您对性能感兴趣,我们可以使用 Divakar 的 justify 函数的修改版本:

pd.DataFrame({'collapsed': justify(
    df.values, invalid_val=np.nan, axis=1, side='right')[:,-1]
})

  collapsed
0         c
1         a
2         b
3         c
4       NaN

def justify(a, invalid_val=0, axis=1, side='left'):    
    """
    Justifies a 2D array

    Parameters
    ----------
    A : ndarray
        Input array to be justified
    axis : int
        Axis along which justification is to be made
    side : str
        Direction of justification. It could be 'left', 'right', 'up', 'down'
        It should be 'left' or 'right' for axis=1 and 'up' or 'down' for axis=0.

    """

    if invalid_val is np.nan:
        mask = pd.notna(a)   # modified for strings
    else:
        mask = a!=invalid_val
    justified_mask = np.sort(mask,axis=axis)
    if (side=='up') | (side=='left'):
        justified_mask = np.flip(justified_mask,axis=axis)
    out = np.full(a.shape, invalid_val, dtype=a.dtype) 
    if axis==1:
        out[justified_mask] = a[mask]
    else:
        out.T[justified_mask.T] = a.T[mask.T]
    return out

关注性能,这里是 NumPy -

In [106]: idx = df.shape[1] - 1 - df.notnull().to_numpy()[:,::-1].argmax(1)

In [107]: pd.Series(df.to_numpy()[np.arange(len(df)),idx])
Out[107]: 
0      c
1      a
2      b
3      c
4    NaN
dtype: object

性能 不是 而是美观和优雅 (-:

df.stack().groupby(level=0).last().reindex(df.index)

0      c
1      a
2      b
3      c
4    NaN
dtype: object

使用 ffill

df.ffill(1).iloc[:,-1]