Merge #98194

98194: opt: add setting to always use histograms to calculate stats r=rytaft a=rytaft Informs #64570 Release note (sql change): Added a new session setting, `optimizer_always_use_histograms`, which ensures that the optimizer always uses histograms when available to calculate the statistics of every plan that it explores. Enabling this setting can prevent the optimizer from choosing a suboptimal index when statistics for a table are stale. Co-authored-by: Rebecca Taft <[email protected]>
cockroachdb · Mar 8, 2023 · 329a232 · 329a232
2 parents 1b162d1 + ef1604f
commit 329a232
Show file tree

Hide file tree

Showing 10 changed files with 369 additions and 1 deletion.
diff --git a/pkg/sql/exec_util.go b/pkg/sql/exec_util.go
@@ -3466,6 +3466,10 @@ func (m *sessionDataMutator) SetEnforceHomeRegionFollowerReadsEnabled(val bool)
 	m.data.EnforceHomeRegionFollowerReadsEnabled = val
 }
 
+func (m *sessionDataMutator) SetOptimizerAlwaysUseHistograms(val bool) {
+	m.data.OptimizerAlwaysUseHistograms = val
+}
+
 // Utility functions related to scrubbing sensitive information on SQL Stats.
 
 // quantizeCounts ensures that the Count field in the

diff --git a/pkg/sql/logictest/testdata/logic_test/information_schema b/pkg/sql/logictest/testdata/logic_test/information_schema
@@ -5016,6 +5016,7 @@ null_ordered_last                                     off
 on_update_rehome_row_enabled                          on
 opt_split_scan_limit                                  2048
 optimizer                                             on
+optimizer_always_use_histograms                       off
 optimizer_use_forecasts                               on
 optimizer_use_histograms                              on
 optimizer_use_improved_disjunction_stats              on

diff --git a/pkg/sql/logictest/testdata/logic_test/pg_catalog b/pkg/sql/logictest/testdata/logic_test/pg_catalog
@@ -2647,6 +2647,7 @@ node_id                                               1                   NULL
 null_ordered_last                                     off                 NULL      NULL        NULL        string
 on_update_rehome_row_enabled                          on                  NULL      NULL        NULL        string
 opt_split_scan_limit                                  2048                NULL      NULL        NULL        string
+optimizer_always_use_histograms                       off                 NULL      NULL        NULL        string
 optimizer_use_forecasts                               on                  NULL      NULL        NULL        string
 optimizer_use_histograms                              on                  NULL      NULL        NULL        string
 optimizer_use_improved_disjunction_stats              on                  NULL      NULL        NULL        string
@@ -2794,6 +2795,7 @@ node_id                                               1                   NULL
 null_ordered_last                                     off                 NULL  user     NULL      off                 off
 on_update_rehome_row_enabled                          on                  NULL  user     NULL      on                  on
 opt_split_scan_limit                                  2048                NULL  user     NULL      2048                2048
+optimizer_always_use_histograms                       off                 NULL  user     NULL      off                 off
 optimizer_use_forecasts                               on                  NULL  user     NULL      on                  on
 optimizer_use_histograms                              on                  NULL  user     NULL      on                  on
 optimizer_use_improved_disjunction_stats              on                  NULL  user     NULL      on                  on
@@ -2940,6 +2942,7 @@ null_ordered_last                                     NULL    NULL     NULL
 on_update_rehome_row_enabled                          NULL    NULL     NULL     NULL        NULL
 opt_split_scan_limit                                  NULL    NULL     NULL     NULL        NULL
 optimizer                                             NULL    NULL     NULL     NULL        NULL
+optimizer_always_use_histograms                       NULL    NULL     NULL     NULL        NULL
 optimizer_use_forecasts                               NULL    NULL     NULL     NULL        NULL
 optimizer_use_histograms                              NULL    NULL     NULL     NULL        NULL
 optimizer_use_improved_disjunction_stats              NULL    NULL     NULL     NULL        NULL

diff --git a/pkg/sql/logictest/testdata/logic_test/show_source b/pkg/sql/logictest/testdata/logic_test/show_source
@@ -109,6 +109,7 @@ node_id                                               1
 null_ordered_last                                     off
 on_update_rehome_row_enabled                          on
 opt_split_scan_limit                                  2048
+optimizer_always_use_histograms                       off
 optimizer_use_forecasts                               on
 optimizer_use_histograms                              on
 optimizer_use_improved_disjunction_stats              on

diff --git a/pkg/sql/opt/memo/memo.go b/pkg/sql/opt/memo/memo.go
@@ -162,6 +162,7 @@ type Memo struct {
 	useImprovedDisjunctionStats            bool
 	useLimitOrderingForStreamingGroupBy    bool
 	useImprovedSplitDisjunctionForJoins    bool
+	alwaysUseHistograms                    bool
 
 	// curRank is the highest currently in-use scalar expression rank.
 	curRank opt.ScalarRank
@@ -219,6 +220,7 @@ func (m *Memo) Init(ctx context.Context, evalCtx *eval.Context) {
 		useImprovedDisjunctionStats:            evalCtx.SessionData().OptimizerUseImprovedDisjunctionStats,
 		useLimitOrderingForStreamingGroupBy:    evalCtx.SessionData().OptimizerUseLimitOrderingForStreamingGroupBy,
 		useImprovedSplitDisjunctionForJoins:    evalCtx.SessionData().OptimizerUseImprovedSplitDisjunctionForJoins,
+		alwaysUseHistograms:                    evalCtx.SessionData().OptimizerAlwaysUseHistograms,
 	}
 	m.metadata.Init()
 	m.logPropsBuilder.init(ctx, evalCtx, m)
@@ -359,7 +361,8 @@ func (m *Memo) IsStale(
 		m.allowOrdinalColumnReferences != evalCtx.SessionData().AllowOrdinalColumnReferences ||
 		m.useImprovedDisjunctionStats != evalCtx.SessionData().OptimizerUseImprovedDisjunctionStats ||
 		m.useLimitOrderingForStreamingGroupBy != evalCtx.SessionData().OptimizerUseLimitOrderingForStreamingGroupBy ||
-		m.useImprovedSplitDisjunctionForJoins != evalCtx.SessionData().OptimizerUseImprovedSplitDisjunctionForJoins {
+		m.useImprovedSplitDisjunctionForJoins != evalCtx.SessionData().OptimizerUseImprovedSplitDisjunctionForJoins ||
+		m.alwaysUseHistograms != evalCtx.SessionData().OptimizerAlwaysUseHistograms {
 		return true, nil
 	}
 

diff --git a/pkg/sql/opt/memo/memo_test.go b/pkg/sql/opt/memo/memo_test.go
@@ -348,6 +348,12 @@ func TestMemoIsStale(t *testing.T) {
 	evalCtx.SessionData().OptimizerUseImprovedDisjunctionStats = false
 	notStale()
 
+	// Stale optimizer_always_use_histograms.
+	evalCtx.SessionData().OptimizerAlwaysUseHistograms = true
+	stale()
+	evalCtx.SessionData().OptimizerAlwaysUseHistograms = false
+	notStale()
+
 	// Stale data sources and schema. Create new catalog so that data sources are
 	// recreated and can be modified independently.
 	catalog = testcat.New()

diff --git a/pkg/sql/opt/memo/statistics_builder.go b/pkg/sql/opt/memo/statistics_builder.go
@@ -2815,6 +2815,9 @@ func (sb *statisticsBuilder) finalizeFromRowCountAndDistinctCounts(
 }
 
 func (sb *statisticsBuilder) shouldUseHistogram(relProps *props.Relational) bool {
+	if sb.evalCtx.SessionData().OptimizerAlwaysUseHistograms {
+		return true
+	}
 	// If we know that the cardinality is below a certain threshold (e.g., due to
 	// a constraint on a key column), don't bother adding the overhead of
 	// creating a histogram.