如何通过循环基于DataFrame特定列值识别实例并截取'Shot'对应行及前4行生成序列

Got it, let's tackle this problem step by step using pandas—perfect for your 1000-row DataFrame. Here's how to efficiently get the rows for each 'Shot' event plus the 4 preceding rows:

Step 1: Locate all 'Shot' event indices #

First, grab the index positions of every row where eventName equals 'Shot':

import pandas as pd
import numpy as np

# Assume your DataFrame is named df
shot_indices = df[df['eventName'] == 'Shot'].index

Step 2: Gather all relevant indices #

We need to include each Shot's index, plus the 4 indices before it (or start from row 0 if there aren't 4 prior rows). Using numpy keeps this clean and automatically handles overlapping ranges to avoid duplicate rows:

# Generate an array of all indices we need to keep
all_relevant_indices = np.unique(
    np.concatenate([np.arange(max(0, idx - 4), idx + 1) for idx in shot_indices])
)

The max(0, idx -4) prevents us from trying to access negative indices for shots that occur in the first 4 rows. np.unique removes duplicates that might come from overlapping ranges (like two consecutive shots sharing some preceding rows).

Step 3: Slice the DataFrame for your final result #

Now just extract the rows using the collected indices:

result_df = df.loc[all_relevant_indices]

Example with your sample data #

In your provided sample, the 'Shot' is at index 4. The code will select indices 0-4, returning all 5 rows of your sample—exactly what you need.

Alternative loop-based approach (for clarity) #

If you prefer a more explicit, beginner-friendly loop, this works too:

selected_rows = []
for idx in shot_indices:
    # Ensure we don't go below row 0
    start_idx = max(0, idx - 4)
    # Append the slice from start_idx to idx (inclusive)
    selected_rows.append(df.loc[start_idx:idx])

# Combine slices and remove duplicate rows
result_df = pd.concat(selected_rows).drop_duplicates()

Both methods will give you the sequence of rows you need for further processing.

内容的提问来源于stack exchange，提问作者Stuart Macfarlane