Manya0407 commented on code in PR #6746:
URL: https://github.com/apache/hive/pull/6746#discussion_r4025147361
##########
ql/src/test/org/apache/hadoop/hive/ql/optimizer/calcite/stats/TestFilterSelectivityEstimator.java:
##########
@@ -1202,4 +1210,184 @@ private static long timestampMillis(String timestamp) {
private static long timestamp(String timestamp) {
return timestampMillis(timestamp) / 1000;
}
+
+ private static final int INTEGER_FIELD_INDEX = 6; // f_integer
+ private static final int DATE_FIELD_INDEX = 9; // f_date
+
+ private void setupMinMaxNoHistogram(float min, float max) {
+ setupMinMaxNoHistogram(min, max, 0);
+ }
+
+ private void setupMinMaxNoHistogram(float min, float max, long numNulls) {
+ stats = new ColStatistics();
+ stats.setHistogram(null);
+ stats.setRange(min, max);
+ stats.setNumNulls(numNulls);
+ currentInputRef = REX_BUILDER.makeInputRef(scan, INTEGER_FIELD_INDEX);
+ doReturn(Collections.singletonList(stats)).when(tableMock)
+ .getColStat(Collections.singletonList(INTEGER_FIELD_INDEX));
+ }
+
+ private RelNode createScanWithPlanner(HiveConf conf) {
+ RelOptPlanner planner = CalcitePlanner.createPlanner(conf);
+ RelOptCluster cluster = RelOptCluster.create(planner, REX_BUILDER);
+ RelBuilder relBuilder = HiveRelFactories.HIVE_BUILDER.create(cluster,
schemaMock);
+ HiveTableScan tableScan =
+ new HiveTableScan(cluster, cluster.traitSetOf(HiveRelNode.CONVENTION),
tableMock, "table", null, false, false);
+ return relBuilder.push(tableScan).build();
+ }
+
+ @Test
+ public void testComparisonMinMaxNoHistogram() {
+ setupMinMaxNoHistogram(0, 100);
+ RexNode int50 = REX_BUILDER.makeLiteral(50,
TYPE_FACTORY.createSqlType(INTEGER), true);
+ RexNode filter =
REX_BUILDER.makeCall(SqlStdOperatorTable.LESS_THAN_OR_EQUAL, currentInputRef,
int50);
+ FilterSelectivityEstimator estimator = new
FilterSelectivityEstimator(scan, mq);
+ Assert.assertEquals(0.5, estimator.estimateSelectivity(filter), DELTA);
Review Comment:
Thanks for spelling out the discrete interpretation-The test currently
expects 0.5 because the implementation uses continuous uniform width over [min,
max], same as EvaluateComparatorWithRange in the annotation path: (50 ā 0) /
(100 ā 0).
Your 51/101 assumes discrete integer counting on [0, 100] ā 51 values
satisfy x <= 50.
I believe discrete should be more intuitive for integer types from a user
perspective. Iām happy to change the expected value (and the estimator for
integer types if we align on that model) once confirmed we want discrete
semantics in CBO and how that should match the annotation path.
--
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]
