如何在Pandas DataFrame中实现事件倒计时列？

Got it, let's figure out how to build that countdown column [B] based on your event column [A]. Here are a couple of practical approaches, depending on whether you prefer readability or efficiency for larger datasets.

Approach 1: Readable Loop-Based Method (Great for Small to Medium Data) #

This method is straightforward and easy to follow, perfect if you're working with a dataset where performance isn't the top priority.

import pandas as pd

# Your sample DataFrame
df = pd.DataFrame({'A': [0, 0, 0, 0, 0, 0, 1, 0, 0, 0]})

# Initialize column B with all 0s
df['B'] = 0

# Get the indices where the event occurs (A == 1)
event_rows = df[df['A'] == 1].index

# Iterate over each event to set the countdown values
for event_idx in event_rows:
    # Set values for 4 days before the event: 5, 4, 3, 2
    for days_before, countdown_val in zip(range(4, 0, -1), [5, 4, 3, 2]):
        target_row = event_idx - days_before
        # Make sure we don't go out of bounds (before the start of the DataFrame)
        if target_row >= 0:
            df.loc[target_row, 'B'] = countdown_val
    # Set the event day itself to 1
    df.loc[event_idx, 'B'] = 1

print(df)

How this works: #

• We start by setting all values in column B to 0 (our default state).
• We find every row where column A is 1—these are our event dates.
• For each event, we go back 4 days, 3 days, etc., and assign the corresponding countdown number. We add a check to avoid trying to modify rows that don't exist (like if an event is in the first 4 rows of your DataFrame).
• Finally, we set the event day's B value to 1.

Approach 2: Vectorized Numpy Method (Better for Large Datasets) #

If you're working with a big dataset, loops can get slow. This vectorized approach uses numpy to handle the operations more efficiently:

import pandas as pd
import numpy as np

df = pd.DataFrame({'A': [0, 0, 0, 0, 0, 0, 1, 0, 0, 0]})

# Create an array of zeros to hold our B values
b_values = np.zeros(len(df), dtype=int)

# Get the indices of all event rows
event_indices = df['A'].values.nonzero()[0]

for idx in event_indices:
    # Generate the range of rows to update: 4 days before event to event day
    rows_to_update = np.arange(idx - 4, idx + 1)
    # The corresponding countdown values: 5,4,3,2,1
    countdown_values = np.array([5, 4, 3, 2, 1])
    # Filter out any rows that are before the start of the DataFrame
    valid_rows = rows_to_update >= 0
    # Assign the values to our array
    b_values[rows_to_update[valid_rows]] = countdown_values[valid_rows]

# Assign the array to column B
df['B'] = b_values

print(df)

Key Notes: #

• Both methods handle multiple events (multiple rows where A=1) automatically. If events are close together (less than 5 days apart), the later event's countdown will overwrite the earlier one's values—which makes sense, since the most recent event's countdown should take priority.
• If an event occurs in the first 4 rows of your DataFrame, the code will skip any "days before" that don't exist (e.g., if the event is at index 1, only index 0 will get a value of 2, and the event row gets 1).

内容的提问来源于stack exchange，提问作者J Ng