zhli1142015 commented on code in PR #5740:
URL: https://github.com/apache/incubator-gluten/pull/5740#discussion_r1600885835
##########
gluten-ut/spark34/src/test/scala/org/apache/spark/sql/execution/datasources/parquet/GlutenParquetRowIndexSuite.scala:
##########
@@ -16,6 +16,337 @@
*/
package org.apache.spark.sql.execution.datasources.parquet
+import org.apache.gluten.execution.{BatchScanExecTransformer,
FileSourceScanExecTransformer}
+
import org.apache.spark.sql.GlutenSQLTestsBaseTrait
+import org.apache.spark.sql.execution.FileSourceScanExec
+import org.apache.spark.sql.execution.datasources.FileFormat
+import org.apache.spark.sql.execution.datasources.v2.BatchScanExec
+import
org.apache.spark.sql.execution.datasources.v2.parquet.ParquetDataSourceV2
+import org.apache.spark.sql.functions.{col, max, min}
+import org.apache.spark.sql.internal.SQLConf
+import org.apache.spark.sql.types.{LongType, StringType}
+
+import org.apache.hadoop.fs.Path
+import org.apache.parquet.column.ParquetProperties._
+import org.apache.parquet.format.converter.ParquetMetadataConverter
+import org.apache.parquet.hadoop.ParquetOutputFormat
+import org.apache.parquet.hadoop.ParquetWriter.DEFAULT_BLOCK_SIZE
+
+import java.io.File
+
+import scala.collection.JavaConverters._
+
+class GlutenParquetRowIndexSuite extends ParquetRowIndexSuite with
GlutenSQLTestsBaseTrait {
+ import testImplicits._
+ override def beforeAll(): Unit = {
+ super.beforeAll()
+ sparkContext.setLogLevel("info")
+ }
+
+ private def readRowGroupRowCounts(path: String): Seq[Long] = {
+ ParquetFooterReader
+ .readFooter(
+ spark.sessionState.newHadoopConf(),
+ new Path(path),
+ ParquetMetadataConverter.NO_FILTER)
+ .getBlocks
+ .asScala
+ .map(_.getRowCount)
+ }
+
+ private def readRowGroupRowCounts(dir: File): Seq[Seq[Long]] = {
+ assert(dir.isDirectory)
+ dir
+ .listFiles()
+ .filter(f => f.isFile && f.getName.endsWith("parquet"))
+ .map(f => readRowGroupRowCounts(f.getAbsolutePath))
+ }
+
+ /** Do the files contain exactly one row group? */
+ private def assertOneRowGroup(dir: File): Unit = {
+ readRowGroupRowCounts(dir).foreach {
+ rcs => assert(rcs.length == 1, "expected one row group per file")
+ }
+ }
+
+ /**
+ * Do the files have a good layout to test row group skipping (both range
metadata filter, and by
+ * using min/max).
+ */
+ private def assertTinyRowGroups(dir: File): Unit = {
+ readRowGroupRowCounts(dir).foreach {
+ rcs =>
+ assert(rcs.length > 1, "expected multiple row groups per file")
+ assert(rcs.last <= DEFAULT_MINIMUM_RECORD_COUNT_FOR_CHECK)
+ assert(
+ rcs.reverse.tail.distinct ==
Seq(DEFAULT_MINIMUM_RECORD_COUNT_FOR_CHECK),
+ "expected row groups with minimal row count")
+ }
+ }
+
+ /**
+ * Do the files have a good layout to test a combination of page skipping
and row group skipping?
+ */
+ private def assertIntermediateRowGroups(dir: File): Unit = {
+ readRowGroupRowCounts(dir).foreach {
+ rcs =>
+ assert(rcs.length >= 3, "expected at least 3 row groups per file")
+ rcs.reverse.tail.foreach {
+ rc =>
+ assert(
+ rc > DEFAULT_MINIMUM_RECORD_COUNT_FOR_CHECK,
+ "expected row groups larger than minimal row count")
+ }
+ }
+ }
+
+ case class GlutenRowIndexTestConf(
Review Comment:
@gaoyangxiaozhu please open an issue for above. Thanks.
--
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]
