Github user BryanCutler commented on a diff in the pull request: https://github.com/apache/spark/pull/18659#discussion_r139261343 --- Diff: python/pyspark/serializers.py --- @@ -199,6 +211,33 @@ def __repr__(self): return "ArrowSerializer" +class ArrowPandasSerializer(ArrowSerializer): + + def __init__(self): + super(ArrowPandasSerializer, self).__init__() + + def dumps(self, series): + """ + Make an ArrowRecordBatch from a Pandas Series and serialize + """ + import pyarrow as pa --- End diff -- Yeah, it would probably be best to handle it the same way as in `toPandas()`. That got me thinking that it is a little weird to have an SQLConf "spark.sql.execution.arrow.enable" that is set for `toPandas()` but has no bearing with `pandas_udf`. It doesn't need to since it is an explicit call but seems a little contradictory, what do you think?
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org