xupefei commented on code in PR #45701:
URL: https://github.com/apache/spark/pull/45701#discussion_r1572148735
##########
connector/connect/client/jvm/src/main/scala/org/apache/spark/sql/SparkSession.scala:
##########
@@ -813,6 +823,28 @@ class SparkSession private[sql] (
* Set to false to prevent client.releaseSession on close() (testing only)
*/
private[sql] var releaseSessionOnClose = true
+
+ private[sql] def registerObservation(planId: Long, observation:
Observation): Unit = {
+ // makes this class thread-safe:
+ // only the first thread entering this block can set sparkSession
+ // all other threads will see the exception, as it is only allowed to do
this once
+ observation.synchronized {
+ if (observationRegistry.contains(planId)) {
+ throw new IllegalArgumentException("An Observation can be used with a
Dataset only once")
+ }
+ observationRegistry.put(planId, observation)
+ }
+ }
+
+ private[sql] def setMetricsAndUnregisterObservation(
+ planId: Long,
+ metrics: Option[Map[String, Any]]): Unit = {
+ observationRegistry.get(planId).map { observation =>
+ if (observation.setMetricsAndNotify(metrics)) {
+ observationRegistry.remove(planId)
Review Comment:
I had the same question when I looked at the code. In Spark Core we only
de-register the Observation when some non-empty metrics are set, so I decide to
keep it the same in Connect. I am not sure under which circumstance the metrics
can be empty.
--
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]
