cloud-fan commented on a change in pull request #24715: [SPARK-25474][SQL] Data
source tables support fallback to HDFS for size estimation
URL: https://github.com/apache/spark/pull/24715#discussion_r315036601
##########
File path:
sql/core/src/test/scala/org/apache/spark/sql/StatisticsCollectionSuite.scala
##########
@@ -650,4 +650,46 @@ class StatisticsCollectionSuite extends
StatisticsCollectionTestBase with Shared
}
}
}
+
+ test("Non-partitioned data source table support fallback to HDFS for size
estimation") {
+ withTempDir { dir =>
+ Seq(false, true).foreach { fallBackToHDFS =>
+ withSQLConf(SQLConf.ENABLE_FALL_BACK_TO_HDFS_FOR_STATS.key ->
s"$fallBackToHDFS") {
+ withTable("spark_25474") {
+ sql(s"CREATE TABLE spark_25474 (c1 BIGINT) USING PARQUET LOCATION
'${dir.toURI}'")
+
spark.range(5).write.mode(SaveMode.Overwrite).parquet(dir.getCanonicalPath)
+
+ assert(getCatalogTable("spark_25474").stats.isEmpty)
+ val relation =
spark.table("spark_25474").queryExecution.analyzed.children.head
+ // fallBackToHDFS = true: The table stats will be recalculated by
DetermineTableStats
+ // fallBackToHDFS = false: The table stats will be recalculated by
FileIndex
Review comment:
no we can't do that, as it will be an regression (image a query picks
broadcast join in Spark 2.4 and SMJ in Spark 3.0).
Since partitioned table usually has many many files, I'm OK with the current
behavior.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
