Question

我有一个pyspark数据框，其中已对数据进行分组以用collect_list列出。

from pyspark.sql.functions import udf, collect_list
from itertools import combinations, chain

#Create Dataframe
df = spark.createDataFrame( [(1,'a'), (1,'b'), (2,'c')] , ["id", "colA"])   

df.show()
>>>
+---+----+
| id|colA|
+---+----+
|  1|   a|
|  1|   b|
|  2|   c|
+---+----+

#Group by and collect to list
df = df.groupBy(df.id).agg(collect_list("colA").alias("colAlist"))

df.show()
>>>
+---+--------+
| id|colAList|
+---+--------+
|  1|  [a, b]|
|  2|     [c]|
+---+--------+

然后我使用一个函数来查找列表元素到新列表的所有组合

allsubsets = lambda l: list(chain(*[combinations(l , n) for n in range(1,len(l)+1)]))
df = df.withColumn('colAsubsets',udf(allsubsets)(df['colAList']))

所以我会期待

+---+--------------------+
| id| colAsubsets        |
+---+--------------------+
|  1|  [[a], [b], [a,b]] |
|  2|  [[b]]             |
+---+--------------------+

但是我得到了

df.show()
>>>
 +---+--------+-----------------------------------------------------------------------------------------+
|id |colAList|colAsubsets                                                                              |
+---+--------+-----------------------------------------------------------------------------------------+
|1  |[a, b]  |[[Ljava.lang.Object;@75e2d657, [Ljava.lang.Object;@7f662637, [Ljava.lang.Object;@b572639]|
|2  |[c]     |[[Ljava.lang.Object;@26f67148]                                                           |
+---+--------+-----------------------------------------------------------------------------------------+

任何想法怎么办？然后也许如何将列表展平到不同的行？

Answer 1

您需要做的就是以扁平的方式从chain和combinations创建的对象中提取元素

正在更改

allsubsets = lambda l: list(chain(*[combinations(l , n) for n in range(1,len(l)+1)]))

到以下

allsubsets = lambda l: [[z for z in y] for y in chain(*[combinations(l , n) for n in range(1,len(l)+1)])]

应该给您

+---+---------+------------------+
|id |colA_list|colAsubsets       |
+---+---------+------------------+
|1  |[a, b]   |[[a], [b], [a, b]]|
|2  |[c]      |[[c]]             |
+---+---------+------------------+

我希望答案会有所帮助

Answer 2

改进@RameshMaharjan答案，以将列表展平到不同的行：

您必须在数组上使用explode。您必须先指定udf的类型，以便它不返回StringType。

from pyspark.sql.functions import explode
from pyspark.sql.types import ArrayType, StringType

allsubsets = lambda l: [[z for z in y] for y in chain(*[combinations(l , n) for n in range(1,len(l)+1)])]
df = df.withColumn('colAsubsets', udf(allsubsets, ArrayType(ArrayType(StringType())))(df['colAList']))
df = df.select('id', explode('colAsubsets'))

结果：

+---+------+
| id|   col|
+---+------+
|  1|   [a]|
|  1|   [b]|
|  1|[a, b]|
|  2|   [c]|
+---+------+

列表上的Spark数据帧操作返回[Ljava.lang.Object; @]

2 个答案: