how to suppress empty list in spark dataframe json output

Question

Here is my pyspark code:

mydf = df.groupby(key1,key2)
           .agg(collect_list(struct(key3)).alias("list2"))
             .groupby(key1)