stats: specially handle unqiue key when estimate (#13354)

statistics/handle/update.go

@@ -941,7 +941,7 @@ func (h *Handle) RecalculateExpectCount(q *statistics.QueryFeedback) error {
 		expected *= idx.GetIncreaseFactor(t.Count)
 	} else {
 		c := t.Columns[id]
-		expected, err = c.GetColumnRowCount(sc, ranges, t.ModifyCount)
+		expected, err = c.GetColumnRowCount(sc, ranges, t.ModifyCount, true)
 		expected *= c.GetIncreaseFactor(t.Count)
 	}
 	if err != nil {

statistics/histogram.go

@@ -727,7 +727,7 @@ func (c *Column) equalRowCount(sc *stmtctx.StatementContext, val types.Datum, mo
 }
 
 // GetColumnRowCount estimates the row count by a slice of Range.
-func (c *Column) GetColumnRowCount(sc *stmtctx.StatementContext, ranges []*ranger.Range, modifyCount int64) (float64, error) {
+func (c *Column) GetColumnRowCount(sc *stmtctx.StatementContext, ranges []*ranger.Range, modifyCount int64, pkIsHandle bool) (float64, error) {
 	var rowCount float64
 	for _, rg := range ranges {
 		cmp, err := rg.LowVal[0].CompareDatum(sc, &rg.HighVal[0])
@@ -737,6 +737,11 @@ func (c *Column) GetColumnRowCount(sc *stmtctx.StatementContext, ranges []*range
 		if cmp == 0 {
 			// the point case.
 			if !rg.LowExclude && !rg.HighExclude {
+				// In this case, the row count is at most 1.
+				if pkIsHandle {
+					rowCount += 1
+					continue
+				}
 				var cnt float64
 				cnt, err = c.equalRowCount(sc, rg.LowVal[0], modifyCount)
 				if err != nil {
@@ -839,6 +844,11 @@ func (idx *Index) GetRowCount(sc *stmtctx.StatementContext, indexRanges []*range
 				continue
 			}
 			if fullLen {
+				// At most 1 in this case.
+				if idx.Info.Unique {
+					totalCount += 1
+					continue
+				}
 				count, err := idx.equalRowCount(sc, lb, modifyCount)
 				if err != nil {
 					return 0, err

statistics/histogram_test.go

@@ -50,9 +50,9 @@ func (s *testStatisticsSuite) TestNewHistogramBySelectivity(c *C) {
 	node.Ranges = append(node.Ranges, &ranger.Range{LowVal: types.MakeDatums(25), HighVal: []types.Datum{types.MaxValueDatum()}})
 	intColResult := `column:1 ndv:16 totColSize:0
 num: 30 lower_bound: 0 upper_bound: 2 repeats: 10
-num: 20 lower_bound: 6 upper_bound: 8 repeats: 0
+num: 11 lower_bound: 6 upper_bound: 8 repeats: 0
 num: 30 lower_bound: 9 upper_bound: 11 repeats: 0
-num: 10 lower_bound: 12 upper_bound: 14 repeats: 0
+num: 1 lower_bound: 12 upper_bound: 14 repeats: 0
 num: 30 lower_bound: 27 upper_bound: 29 repeats: 0`
 
 	stringCol := &Column{}

statistics/selectivity_test.go

@@ -416,6 +416,38 @@ func (s *testStatsSuite) TestEstimationForUnknownValues(c *C) {
 	c.Assert(count, Equals, 0.0)
 }
 
+func (s *testStatsSuite) TestEstimationUniqueKeyEqualConds(c *C) {
+	defer cleanEnv(c, s.store, s.do)
+	testKit := testkit.NewTestKit(c, s.store)
+	testKit.MustExec("use test")
+	testKit.MustExec("drop table if exists t")
+	testKit.MustExec("create table t(a int, b int, c int, unique key(b))")
+	testKit.MustExec("insert into t values (1,1,1),(2,2,2),(3,3,3),(4,4,4),(5,5,5),(6,6,6),(7,7,7)")
+	testKit.MustExec("analyze table t")
+	table, err := s.do.InfoSchema().TableByName(model.NewCIStr("test"), model.NewCIStr("t"))
+	c.Assert(err, IsNil)
+	statsTbl := s.do.StatsHandle().GetTableStats(table.Meta())
+
+	sc := &stmtctx.StatementContext{}
+	idxID := table.Meta().Indices[0].ID
+	count, err := statsTbl.GetRowCountByIndexRanges(sc, idxID, getRange(7, 7))
+	c.Assert(err, IsNil)
+	c.Assert(count, Equals, 1.0)
+
+	count, err = statsTbl.GetRowCountByIndexRanges(sc, idxID, getRange(6, 6))
+	c.Assert(err, IsNil)
+	c.Assert(count, Equals, 1.0)
+
+	colID := table.Meta().Columns[0].ID
+	count, err = statsTbl.GetRowCountByIntColumnRanges(sc, colID, getRange(7, 7))
+	c.Assert(err, IsNil)
+	c.Assert(count, Equals, 1.0)
+
+	count, err = statsTbl.GetRowCountByIntColumnRanges(sc, colID, getRange(6, 6))
+	c.Assert(err, IsNil)
+	c.Assert(count, Equals, 1.0)
+}
+
 func (s *testStatsSuite) TestPrimaryKeySelectivity(c *C) {
 	defer cleanEnv(c, s.store, s.do)
 	testKit := testkit.NewTestKit(c, s.store)

statistics/table.go

@@ -236,7 +236,7 @@ func (coll *HistColl) GetRowCountByIntColumnRanges(sc *stmtctx.StatementContext,
 		}
 		return getPseudoRowCountByUnsignedIntRanges(intRanges, float64(coll.Count)), nil
 	}
-	result, err := c.GetColumnRowCount(sc, intRanges, coll.ModifyCount)
+	result, err := c.GetColumnRowCount(sc, intRanges, coll.ModifyCount, true)
 	result *= c.GetIncreaseFactor(coll.Count)
 	return result, errors.Trace(err)
 }
@@ -247,7 +247,7 @@ func (coll *HistColl) GetRowCountByColumnRanges(sc *stmtctx.StatementContext, co
 	if !ok || c.IsInvalid(sc, coll.Pseudo) {
 		return GetPseudoRowCountByColumnRanges(sc, float64(coll.Count), colRanges, 0)
 	}
-	result, err := c.GetColumnRowCount(sc, colRanges, coll.ModifyCount)
+	result, err := c.GetColumnRowCount(sc, colRanges, coll.ModifyCount, false)
 	result *= c.GetIncreaseFactor(coll.Count)
 	return result, errors.Trace(err)
 }
@@ -366,6 +366,12 @@ func (coll *HistColl) getIndexRowCount(sc *stmtctx.StatementContext, idxID int64
 	totalCount := float64(0)
 	for _, ran := range indexRanges {
 		rangePosition := GetOrdinalOfRangeCond(sc, ran)
+		coverAll := len(ran.LowVal) == len(idx.Info.Columns) && rangePosition == len(ran.LowVal)
+		// // In this case, the row count is at most 1.
+		if coverAll && idx.Info.Unique {
+			totalCount += 1.0
+			continue
+		}
 		// If first one is range, just use the previous way to estimate; if it is [NULL, NULL] range
 		// on single-column index, use previous way as well, because CMSketch does not contain null
 		// values in this case.