解决Pandas中使用ffill填充多事件值时的代码覆盖问题

我有一个可简化为如下形式的DataFrame：

import numpy as np
import pandas as pd

TR = [17,18,19,20,21,22,23,24,25,26,27,28,29]
Event = ['SRNeg', np.NaN, np.NaN, np.NaN, np.NaN, np.NaN, np.NaN, np.NaN, 'Rating', np.NaN, np.NaN, 'ArrowTask', np.NaN]
df = pd.DataFrame({'Event':Event,'TR':TR})

需要将事件值后的部分NaN填充为该事件值（并非所有NaN都需填充），理想输出如下：

TR = [17,18,19,20,21,22,23,24,25,26,27,28,29]
Event = ['SRNeg', np.NaN, np.NaN, np.NaN, np.NaN, np.NaN, np.NaN, np.NaN, 'Rating', np.NaN, np.NaN, 'ArrowTask', np.NaN]
Event_BOLD_Duration = ['SRNeg', 'SRNeg', 'SRNeg', 'SRNeg', 'SRNeg', 'SRNeg', 'SRNeg', 'SRNeg', 'Rating', 'Rating', np.NaN, 'ArrowTask', 'ArrowTask']
df = pd.DataFrame({'Event':Event,'Event_BOLD_Duration':Event_BOLD_Duration,'TR':TR})

我原本编写了如下代码，但每行代码都会覆盖前一行的输出结果：

# line 1
scores_full['Event_BOLD_Duration'] = scores_full['Event'].where(scores_full['Event'].isin(['SRNeg', 'SINeg'])).ffill(limit=7).fillna(scores_full['Event']) 
# line 2
scores_full['Event_BOLD_Duration'] = scores_full['Event'].where(scores_full['Event'].isin(['ArrowTask'])).ffill(limit=4).fillna(scores_full['Event']) 
# line 3
scores_full['Event_BOLD_Duration'] = scores_full['Event'].where(scores_full['Event'].isin(['Rating'])).ffill(limit=1).fillna(scores_full['Event'])

你的问题出在每次赋值都直接从原始Event列重新生成数据，完全覆盖了之前的处理结果。正确的做法是先初始化目标列，再针对不同事件类型逐个填充后续指定数量的NaN。

实现代码 #

import numpy as np
import pandas as pd

# 初始化DataFrame
TR = [17,18,19,20,21,22,23,24,25,26,27,28,29]
Event = ['SRNeg', np.NaN, np.NaN, np.NaN, np.NaN, np.NaN, np.NaN, np.NaN, 'Rating', np.NaN, np.NaN, 'ArrowTask', np.NaN]
df = pd.DataFrame({'Event':Event,'TR':TR})

# 初始化目标列，先复制原始Event值
df['Event_BOLD_Duration'] = df['Event'].copy()

# 1. 处理SRNeg/SINeg：每个事件后填充7行
for event in ['SRNeg', 'SINeg']:
    # 获取所有该事件的索引
    event_idx_list = df[df['Event'] == event].index
    for idx in event_idx_list:
        # 计算填充的结束位置，避免超出DataFrame范围
        fill_end = min(idx + 7, df.index[-1])
        # 填充从idx+1到fill_end的行
        df.loc[idx+1:fill_end, 'Event_BOLD_Duration'] = event

# 2. 处理Rating：每个事件后填充1行
event_idx_list = df[df['Event'] == 'Rating'].index
for idx in event_idx_list:
    fill_end = min(idx + 1, df.index[-1])
    if fill_end > idx:  # 确保存在可填充的行
        df.loc[idx+1:fill_end, 'Event_BOLD_Duration'] = 'Rating'

# 3. 处理ArrowTask：每个事件后填充1行（匹配你的理想输出）
event_idx_list = df[df['Event'] == 'ArrowTask'].index
for idx in event_idx_list:
    fill_end = min(idx + 1, df.index[-1])
    if fill_end > idx:
        df.loc[idx+1:fill_end, 'Event_BOLD_Duration'] = 'ArrowTask'

代码说明 #

• 先复制原始Event列到Event_BOLD_Duration，保留所有原始非NaN值
• 针对每个事件类型，找到其出现的索引，然后向后填充指定数量的行
• 使用min(idx + N, df.index[-1])避免超出DataFrame的最后一行索引，防止报错

执行后即可得到你想要的理想输出。

内容的提问来源于stack exchange，提问作者jo_