How to save a dataframe into a json file with multiline option in pyspark

Question

In Pyspark I want to save a dataframe as a json file but in the below format

Say this is my dataframe

>>> rdd1.show()
+----------+-----+
|        f1|   f2|
+----------+-----+
|AAAAAAAAAA|99999|
| BBBBBBBBB|99999|
| CCCCCCCCC|99999|
+----------+-----+

if I save the above dataframe as a json file it gives an output like below

>>>rdd1.coalesce(1).write.json("file:///test_directory/sample4")
{"f1":"AAAAAAAAAA","f2":"99999"}
{"f1":"BBBBBBBBB","f2":"99999"}
{"f1":"CCCCCCCCC","f2":"99999"}

But I want it like the below

[{"f1":"AAAAAAAAAA","f2":"99999"},{"f1":"BBBBBBBBB","f2":"99999"},{"f1":"CCCCCCCCC","f2":"99999"}]

I have tried option("multiLine", "true") and lineSep="," none seems to be working, these options are working only for read not write. Please suggest a solution for this problem

Answer 1

Use to_json with collect_list function and write as .text().

Example:

df.show()
#+-----+-----+
#|   f1|   f2|
#+-----+-----+
#|AAAAA| 9999|
#|  BBB|99999|
#| CCCC| 9999|
#+-----+-----+

from pyspark.sql.functions import *

df.agg(to_json(collect_list(struct(col("f1"),col("f2")))).alias("d")).\
write.\
mode("overwrite").\
text("<path>")

#output
#[{"f1":"AAAAA","f2":"9999"},{"f1":"BBB","f2":"99999"},{"f1":"CCCC","f2":"9999"}]