Python 如何在pyspark中创建包含两个数据帧列的字典？

我有一个包含两列的数据框架，如下所示：

    df = spark.createDataFrame([('A', 'Science'),
 ('A', 'Math'),
 ('A', 'Physics'),
 ('B', 'Science'),
 ('B', 'English'),
 ('C', 'Math'),
 ('C', 'English'),
 ('C', 'Latin')],
 ['Group', 'Subjects'])


Group   Subjects
A       Science
A       Math
A       Physics
B       Science
B       English
C       Math
C       English
C       Latin

我需要为Group列中的每个唯一值遍历这些数据，并执行一些处理。我正在考虑创建一个字典，每个组的名称作为键，相应的主题列表作为值

因此，我的预期输出如下所示：

{A:['Science', 'Math', 'Physics'], B:['Science', 'English'], C:['Math', 'English', 'Latin']}

---

如何在pyspark中实现这一点？

请查看：您可以执行

groupBy

并使用

collect\u list

    #Input DF
    # +-----+-------+
    # |group|subject|
    # +-----+-------+
    # |    A|   Math|
    # |    A|Physics|
    # |    B|Science|
    # +-----+-------+

    df1 = df.groupBy("group").agg(F.collect_list("subject").alias("subject")).orderBy("group")

    df1.show(truncate=False)

    # +-----+---------------+
    # |group|subject        |
    # +-----+---------------+
    # |A    |[Math, Physics]|
    # |B    |[Science]      |
    # +-----+---------------+

    dict = {row['group']:row['subject'] for row in df1.collect()}

    print(dict)

    # {'A': ['Math', 'Physics'], 'B': ['Science']}

---

如果您需要唯一的主题，则可以使用collect_set，否则请使用collect_list

import pyspark.sql.functions as F
 df = spark.createDataFrame([('A', 'Science'),
 ('A', 'Math'),
 ('A', 'Physics'),
 ('B', 'Science'),
 ('B', 'English'),
 ('C', 'Math'),
 ('C', 'English'),
 ('C', 'Latin')],
 ['Group', 'Subjects'])
 
 df_tst=df.groupby('Group').agg(F.collect_set("Subjects").alias('Subjects')).withColumn("dict",F.create_map('Group',"Subjects"))

结果:

+-----+------------------------+-------------------------------+
|Group|Subjects                |dict                           |
+-----+------------------------+-------------------------------+
|C    |[Math, Latin, English]  |[C -> [Math, Latin, English]]  |
|B    |[Science, English]      |[B -> [Science, English]]      |
|A    |[Math, Physics, Science]|[A -> [Math, Physics, Science]]|
+-----+------------------------+-------------------------------+