求助：为DataFrame多列新增行差计算列（row[n+1]-row[n]）的实现

Hey there, this is a super common task when working with sequential financial data (like your stock earnings use case), and pandas has a built-in tool that makes this straightforward. Let me walk you through two practical approaches to get exactly what you need—adding 6 new columns following the stock_name + "_Earning" rule, while keeping all your original data intact.

Approach 1: Loop Through Columns (Great for Step-by-Step Clarity) #

If you prefer explicit, easy-to-follow code, looping through each column lets you control every part of the process:

Step-by-Step Code #

First, let's create a sample DataFrame to simulate your stock data:

import pandas as pd
import numpy as np

# 模拟6只股票的原始数据（10行示例）
np.random.seed(42)
df = pd.DataFrame({
    'StockA': np.random.randint(100, 200, 10),
    'StockB': np.random.randint(80, 180, 10),
    'StockC': np.random.randint(120, 220, 10),
    'StockD': np.random.randint(90, 190, 10),
    'StockE': np.random.randint(110, 210, 10),
    'StockF': np.random.randint(70, 170, 10)
})

Now calculate the row-to-row differences and add the new columns:

# 遍历每一列，计算当前行与前一行的差值
for col in df.columns:
    # 按照规则命名新列
    new_col_name = f"{col}_Earning"
    # 使用diff()：默认就是计算row[n+1] - row[n]的差值
    df[new_col_name] = df[col].diff()

Approach 2: Batch Processing (Cleaner & More Efficient) #

For a concise, one-liner-style solution, you can generate all difference columns at once and merge them with the original DataFrame:

# 批量生成所有差值列，并按规则重命名
earning_columns = df.diff().rename(columns=lambda x: f"{x}_Earning")
# 合并原数据和新列（axis=1表示按列拼接）
df = pd.concat([df, earning_columns], axis=1)

Key Notes #

• NaN Handling: The first row of each new _Earning column will be NaN (since there's no previous row to compare). If you need to fill this gap (e.g., with 0), just add .fillna(0) to the diff() call:

  df[new_col_name] = df[col].diff().fillna(0)
  

• Preserve Original Data: Both methods keep all your original columns intact—they only append the new difference columns to the end of the DataFrame.
• Naming Accuracy: Using f-strings (Python 3.6+) or lambda functions ensures your new columns strictly follow the stock_name + "_Earning" convention.

Verify the Result #

To confirm all 6 new columns were added, print the DataFrame's column list:

print(df.columns)
# Output: Index(['StockA', 'StockB', 'StockC', 'StockD', 'StockE', 'StockF',
#                'StockA_Earning', 'StockB_Earning', 'StockC_Earning',
#                'StockD_Earning', 'StockE_Earning', 'StockF_Earning'],
#               dtype='object')

内容的提问来源于stack exchange，提问作者3kstc