从DataFrame筛选InvoiceDate在指定日期范围±7天内的行

首先确认所有日期列均为datetime64类型，这是日期运算的基础。若读取CSV时未指定类型，先执行转换：

import pandas as pd

# 转换日期列格式
df["InvoiceDate"] = pd.to_datetime(df["InvoiceDate"])
df["CalcStartDate"] = pd.to_datetime(df["CalcStartDate"])
df["CalcEndDate"] = pd.to_datetime(df["CalcEndDate"])

接下来用pd.Timedelta做日期偏移——Pandas的向量化操作对大数据量效率极高，远优于逐行循环，完全不用顾虑性能问题：

# 筛选InvoiceDate在[CalcStartDate-7天, CalcEndDate+7天]范围内的行
filtered_df = df[
    (df["InvoiceDate"] >= df["CalcStartDate"] - pd.Timedelta(days=7)) &
    (df["InvoiceDate"] <= df["CalcEndDate"] + pd.Timedelta(days=7))
]

额外优化建议 #

• 读取CSV时直接指定日期列，避免后续二次转换：

  df = pd.read_csv(
      "your_file.csv",
      parse_dates=["InvoiceDate", "CalcStartDate", "CalcEndDate"]
  )
  

• 若数据量过大导致内存不足，可使用chunksize分块读取处理，再合并结果：

  chunk_list = []
  for chunk in pd.read_csv("your_file.csv", parse_dates=["InvoiceDate", "CalcStartDate", "CalcEndDate"], chunksize=10000):
      filtered_chunk = chunk[
          (chunk["InvoiceDate"] >= chunk["CalcStartDate"] - pd.Timedelta(days=7)) &
          (chunk["InvoiceDate"] <= chunk["CalcEndDate"] + pd.Timedelta(days=7))
      ]
      chunk_list.append(filtered_chunk)
  filtered_df = pd.concat(chunk_list)
  

内容的提问来源于stack exchange，提问作者5122014009