python - 如何在pyspark中将2列聚合到地图中

Question

我有一个像这样的数据框

a = spark.createDataFrame([['Alice', '2020-03-03', '1'], ['Bob', '2020-03-03', '1'], ['Bob', '2020-03-05', '2']], ['name', 'dt', 'hits'])
a.show()
+-----+----------+----+
| name|        dt|hits|
+-----+----------+----+
|Alice|2020-03-03|   1|
|  Bob|2020-03-03|   1|
|  Bob|2020-03-05|   2|
+-----+----------+----+

我想聚合 dt 并点击 Columns 到地图中 -

+-----+------------------------------------+
| name|    map                             |
+-----+------------------------------------+
|Alice|   {'2020-03-03': 1, '2020-03-05':2}|
|  Bob|   {'2020-03-03': 1}                |
+-----+------------------------------------+

但是这段代码抛出了一个异常：

from pyspark.sql import functions as F
a = a.groupBy(F.col('name')).agg(F.create_map(F.col('dt'), F.col('hits')))

Py4JJavaError: An error occurred while calling o2920.agg.
: org.apache.spark.sql.AnalysisException: expression '`dt`' is neither present in the group by, nor is it an aggregate function. Add to group by or wrap in first() (or first_value) if you don't care which value you get.;;
Aggregate [name#1329], [name#1329, map(dt#1330, hits#1331) AS map(dt, hits)#1361]
+- LogicalRDD [name#1329, dt#1330, hits#1331], false

我究竟做错了什么？

标签： pythonpysparkpyspark-dataframes

Answer 1

对于spark2.4+，您可以map_from_arrays这样使用：

from pyspark.sql import functions as F

a.groupBy("name").agg(F.map_from_arrays(F.collect_list("dt"),\
                     F.collect_list("hits")).alias("map")).show(truncate=False)

#+-----+----------------------------------+
#|name |map                               |
#+-----+----------------------------------+
#|Bob  |[2020-03-03 -> 1, 2020-03-05 -> 2]|
#|Alice|[2020-03-03 -> 1]                 |
#+-----+----------------------------------+