如何用列表正确替换pandas DataFrame中的元素？

问题场景 #

我有一个简单的Pandas DataFrame：

import pandas as pd

test_df = pd.DataFrame(
    {
        0: [
            "Setup",
            "Dissection",
            "Specimen Removal",
            "Closure",
        ]
    }
)

希望将列0中的文本替换为对应的RGBA颜色值，于是写了如下代码：

colours = {
    "Setup": [0.12156862745098039, 0.4666666666666667, 0.7058823529411765, 1.0],
    "Dissection": [0.8392156862745098, 0.15294117647058825, 0.1568627450980392, 1.0],
    "Specimen Removal": [
        0.9686274509803922,
        0.7137254901960784,
        0.8235294117647058,
        1.0,
    ],
    "Closure": [0.6196078431372549, 0.8549019607843137, 0.8980392156862745, 1.0],
}

test_df.replace(colours)

运行后触发如下错误：

File ~/.venv/fastai/lib/python3.11/site-packages/pandas/core/internals/blocks.py:729, in Block.replace(self, to_replace, value, inplace, mask, using_cow)
    725 elif self._can_hold_element(value):
    726     # TODO(CoW): Maybe split here as well into columns where mask has True
    727     # and rest?
    728     blk = self._maybe_copy(using_cow, inplace)
--> 729     putmask_inplace(blk.values, mask, value)
    730     if not (self.is_object and value is None):
    731         # if the user *explicitly* gave None, we keep None, otherwise
    732         #  may downcast to NaN
    733     blocks = blk.convert(copy=False, using_cow=using_cow)

File ~/.venv/fastai/lib/python3.11/site-packages/pandas/core/array_algos/putmask.py:56, in putmask_inplace(values, mask, value)
     54         values[mask] = value[mask]
     55     else:
--> 56         values[mask] = value
     57 else:
     58     # GH#37833 np.putmask is more performant than __setitem__
     59     np.putmask(values, mask, value)

ValueError: NumPy boolean array indexing assignment cannot assign 4 input values to the 1 output values where the mask is true

错误原因 #

pd.DataFrame.replace()的设计逻辑是单个值替换为单个值，原DataFrame的列0是一维的字符串列，每个单元格只能存储单个元素；而你要替换的是长度为4的列表，相当于试图把4个值塞进原本只能放1个值的位置，因此触发维度不匹配的报错。

解决方法 #

方法1：用map()直接映射（保留列表格式） #

使用Series.map()方法可以直接将字典的键映射为对应的值，每个单元格会存储完整的颜色列表：

test_df[0] = test_df[0].map(colours)

处理后的DataFrame结构示例：

0
0  [0.12156862745098039, 0.4666666666666667, 0.70...
1  [0.8392156862745098, 0.15294117647058825, 0.15...
2  [0.9686274509803922, 0.7137254901960784, 0.823...
3  [0.6196078431372549, 0.8549019607843137, 0.898...

方法2：拆分颜色通道为独立列 #

如果后续需要单独处理RGBA的每个通道，可以用apply(pd.Series)把列表拆分成多列：

# 映射后拆分列
color_df = test_df[0].map(colours).apply(pd.Series)
# 给列命名（可选）
color_df.columns = ['R', 'G', 'B', 'A']

处理后的DataFrame结构示例：

R         G         B    A
0  0.121569  0.466667  0.705882  1.0
1  0.839216  0.152941  0.156863  1.0
2  0.968627  0.713725  0.823529  1.0
3  0.619608  0.854902  0.898039  1.0

内容的提问来源于stack exchange，提问作者K G