pyspark - 逐行读取 df 以在另一个 df 中搜索

Question

我是 pyspark 的新手，我需要帮助在 df 中进行搜索。
我有 df1 与学生数据如下

+---------+----------+--------------------+
|studentid|   course |  registration_date |
+---------+----------+--------------------+
|      348|         2|     15-11-2021     |
|      567|         1|     05-11-2021     |
|      595|         3|     15-10-2021     |
|      580|         2|     06-11-2021     |
|      448|         4|     15-09-2021     |
+---------+----------+--------------------+

df2。 有关注册期的信息如下

+--------+------------+------------+
| period | start_date |  end_date  |
+--------+------------+------------+
|       1| 01-09-2021 | 15-09-2021 |
|       2| 16-09-2021 | 30-09-2021 |
|       3| 01-10-2021 | 15-10-2021 |
|       4| 16-10-2021 | 31-10-2021 |
|       5| 01-11-2021 | 15-11-2021 |
|       6| 16-11-2021 | 30-11-2021 |
+--------+------------+------------+

我需要逐行迭代 df1，获取学生注册日期并使用此日期，转到 df2 并获取条件 df2.start_date <= df1.registration_date <= df2.end_date 的期间信息。
结果将是新的 df 如下

+---------+----------+--------------------+--------+------------+------------+
|studentid|   course |  registration_date | period | start_date |  end_date  |
+---------+----------+--------------------+--------+------------+------------+
|      348|         2|     15-11-2021     |       5| 01-11-2021 | 15-11-2021 |
|      567|         1|     05-11-2021     |       5| 01-11-2021 | 15-11-2021 |
|      595|         3|     15-10-2021     |       3| 01-10-2021 | 15-10-2021 |
|      580|         2|     06-11-2021     |       5| 01-11-2021 | 15-11-2021 |
|      448|         4|     15-09-2021     |       1| 01-09-2021 | 15-09-2021 |
+---------+----------+--------------------+--------+------------+------------+

Answer 1

您可以将join条件指定为复杂条件。

工作示例

from datetime import datetime
from pyspark.sql import functions as F


df = spark.createDataFrame([
    (348, 2, datetime.strptime("15-11-2021", "%d-%m-%Y")),
    (567, 1, datetime.strptime("05-11-2021", "%d-%m-%Y")),
    (595, 3, datetime.strptime("15-10-2021", "%d-%m-%Y")),
    (580, 2, datetime.strptime("06-11-2021", "%d-%m-%Y")),
    (448, 4, datetime.strptime("15-09-2021", "%d-%m-%Y")),]
, ("studentid", "course", "registration_date",)).withColumn("registration_date", F.to_date(F.col("registration_date")))

df2 = spark.createDataFrame([
    (1, datetime.strptime("01-09-2021", "%d-%m-%Y"), datetime.strptime("15-09-2021", "%d-%m-%Y")),
    (2, datetime.strptime("16-09-2021", "%d-%m-%Y"), datetime.strptime("30-09-2021", "%d-%m-%Y")),
    (3, datetime.strptime("01-10-2021", "%d-%m-%Y"), datetime.strptime("15-10-2021", "%d-%m-%Y")),
    (4, datetime.strptime("16-10-2021", "%d-%m-%Y"), datetime.strptime("31-10-2021", "%d-%m-%Y")),
    (5, datetime.strptime("01-11-2021", "%d-%m-%Y"), datetime.strptime("15-11-2021", "%d-%m-%Y")),
    (6, datetime.strptime("16-11-2021", "%d-%m-%Y"), datetime.strptime("30-11-2021", "%d-%m-%Y")),]
, ("period", "start_date", "end_date")).withColumn("start_date", F.to_date(F.col("start_date"))).withColumn("end_date", F.to_date(F.col("end_date")))

df.join(df2, (df2["start_date"] <= df["registration_date"]) & (df["registration_date"] <= df2["end_date"])).show()

输出

+---------+------+-----------------+------+----------+----------+
|studentid|course|registration_date|period|start_date|  end_date|
+---------+------+-----------------+------+----------+----------+
|      348|     2|       2021-11-15|     5|2021-11-01|2021-11-15|
|      567|     1|       2021-11-05|     5|2021-11-01|2021-11-15|
|      595|     3|       2021-10-15|     3|2021-10-01|2021-10-15|
|      448|     4|       2021-09-15|     1|2021-09-01|2021-09-15|
|      580|     2|       2021-11-06|     5|2021-11-01|2021-11-15|
+---------+------+-----------------+------+----------+----------+