grundprinzip commented on code in PR #40827:
URL: https://github.com/apache/spark/pull/40827#discussion_r1181011537
##########
connector/connect/client/jvm/src/main/scala/org/apache/spark/sql/SparkSession.scala:
##########
@@ -119,12 +119,23 @@ class SparkSession private[sql] (
private def createDataset[T](encoder: AgnosticEncoder[T], data:
Iterator[T]): Dataset[T] = {
newDataset(encoder) { builder =>
- val localRelationBuilder = builder.getLocalRelationBuilder
- .setSchema(encoder.schema.json)
if (data.nonEmpty) {
val timeZoneId = conf.get("spark.sql.session.timeZone")
- val arrowData = ConvertToArrow(encoder, data, timeZoneId, allocator)
- localRelationBuilder.setData(arrowData)
+ val (arrowData, arrowDataSize) = ConvertToArrow(encoder, data,
timeZoneId, allocator)
+ if (arrowDataSize <=
conf.get(SQLConf.LOCAL_RELATION_CACHE_THRESHOLD.key).toInt) {
Review Comment:
Yes absolutely, your're right this needs to be configurable. My point is
mostly that we don't have Spark Confs on the client. In Python we don't have
the JVM to parse them on startup for example, you can set them via
`spark.conf.set` but that's it.
My general recommendation would be to avoid pulling in additional SQL/Core
dependencies in the client.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
