spark dataframe - combine text from multiple rows in pyspark

Question

I created a PySpark dataframe using the following code

testlist = [
             {"category":"A","name":"A1"}, 
             {"category":"A","name":"A2"}, 
             {"category":"B","name":"B1"},
             {"category":"B","name":"B2"}
]

spark_df = spark.createDataFrame(testlist)

Result:

category    name
A           A1
A           A2
B           B1
B           B2

I want to make it appear as follows:

category    name
A           A1, A2
B           B1, B2

I tried the following code which does not work

spark_df.groupby('category').agg('name', lambda x:x + ', ')

Can anyone help identify what I am doing wrong and the best way to make this happen ?

Answer 1

One option is to use pyspark.sql.functions.collect_list() as the aggregate function.

from pyspark.sql.functions import collect_list
grouped_df = spark_df.groupby('category').agg(collect_list('name').alias("name"))

This will collect the values for name into a list and the resultant output will look like:

grouped_df.show()
#+---------+---------+
#|category |name     |
#+---------+---------+
#|A        |[A1, A2] |
#|B        |[B1, B2] |
#+---------+---------+

Update 2019-06-10: If you wanted your output as a concatenated string, you can use pyspark.sql.functions.concat_ws to concatenate the values of the collected list, which will be better than using a udf:

from pyspark.sql.functions import concat_ws

grouped_df.withColumn("name", concat_ws(", ", "name")).show()
#+---------+-------+
#|category |name   |
#+---------+-------+
#|A        |A1, A2 |
#|B        |B1, B2 |
#+---------+-------+

---

Original Answer: If you wanted your output as a concatenated string, you'd have to can use a udf. For example, you can first do the groupBy() as above and the apply a udf to join the collected list:

from pyspark.sql.functions import udf
concat_list = udf(lambda lst: ", ".join(lst), StringType())

grouped_df.withColumn("name", concat_list("name")).show()
#+---------+-------+
#|category |name   |
#+---------+-------+
#|A        |A1, A2 |
#|B        |B1, B2 |
#+---------+-------+