This is an automated email from the ASF dual-hosted git repository.
morrysnow pushed a commit to branch branch-2.1
in repository https://gitbox.apache.org/repos/asf/doris.git
The following commit(s) were added to refs/heads/branch-2.1 by this push:
new e64f2e68e05 [opt](nereids) refine stats derive (#40654) (#40698)
(#42050)
e64f2e68e05 is described below
commit e64f2e68e051d9f59624fc0e22659d287f353582
Author: xzj7019 <[email protected]>
AuthorDate: Fri Oct 18 16:18:10 2024 +0800
[opt](nereids) refine stats derive (#40654) (#40698) (#42050)
pick from master #40654 #40698
---
.../doris/nereids/stats/ExpressionEstimation.java | 23 +++++---
.../doris/nereids/stats/StatsCalculator.java | 4 +-
.../org/apache/doris/nereids/types/StringType.java | 5 --
.../nereids/types/coercion/CharacterType.java | 3 ++
.../doris/statistics/ColumnStatisticBuilder.java | 18 +++++++
.../org/apache/doris/statistics/Statistics.java | 5 +-
.../nereids/stats/ExpressionEstimationTest.java | 63 ++++++++++++++++++++++
7 files changed, 106 insertions(+), 15 deletions(-)
diff --git
a/fe/fe-core/src/main/java/org/apache/doris/nereids/stats/ExpressionEstimation.java
b/fe/fe-core/src/main/java/org/apache/doris/nereids/stats/ExpressionEstimation.java
index 14d5ae8b63d..780c5922c6a 100644
---
a/fe/fe-core/src/main/java/org/apache/doris/nereids/stats/ExpressionEstimation.java
+++
b/fe/fe-core/src/main/java/org/apache/doris/nereids/stats/ExpressionEstimation.java
@@ -128,6 +128,10 @@ public class ExpressionEstimation extends
ExpressionVisitor<ColumnStatistic, Sta
@Override
public ColumnStatistic visit(Expression expr, Statistics context) {
+ ColumnStatistic stats = context.findColumnStatistics(expr);
+ if (stats != null) {
+ return stats;
+ }
List<Expression> childrenExpr = expr.children();
if (CollectionUtils.isEmpty(childrenExpr)) {
return ColumnStatistic.UNKNOWN;
@@ -135,26 +139,28 @@ public class ExpressionEstimation extends
ExpressionVisitor<ColumnStatistic, Sta
return expr.child(0).accept(this, context);
}
- //TODO: case-when need to re-implemented
@Override
public ColumnStatistic visitCaseWhen(CaseWhen caseWhen, Statistics
context) {
double ndv = caseWhen.getWhenClauses().size();
+ double width = 1;
if (caseWhen.getDefaultValue().isPresent()) {
ndv += 1;
}
for (WhenClause clause : caseWhen.getWhenClauses()) {
ColumnStatistic colStats =
ExpressionEstimation.estimate(clause.getResult(), context);
ndv = Math.max(ndv, colStats.ndv);
+ width = Math.max(width, clause.getResult().getDataType().width());
}
if (caseWhen.getDefaultValue().isPresent()) {
ColumnStatistic colStats =
ExpressionEstimation.estimate(caseWhen.getDefaultValue().get(), context);
ndv = Math.max(ndv, colStats.ndv);
+ width = Math.max(width,
caseWhen.getDefaultValue().get().getDataType().width());
}
return new ColumnStatisticBuilder()
.setNdv(ndv)
.setMinValue(Double.NEGATIVE_INFINITY)
.setMaxValue(Double.POSITIVE_INFINITY)
- .setAvgSizeByte(8)
+ .setAvgSizeByte(width)
.setNumNulls(0)
.build();
}
@@ -162,15 +168,20 @@ public class ExpressionEstimation extends
ExpressionVisitor<ColumnStatistic, Sta
@Override
public ColumnStatistic visitIf(If ifClause, Statistics context) {
double ndv = 2;
+ double width = 1;
ColumnStatistic colStatsThen =
ExpressionEstimation.estimate(ifClause.child(1), context);
ndv = Math.max(ndv, colStatsThen.ndv);
+ width = Math.max(width, ifClause.child(1).getDataType().width());
+
ColumnStatistic colStatsElse =
ExpressionEstimation.estimate(ifClause.child(2), context);
ndv = Math.max(ndv, colStatsElse.ndv);
+ width = Math.max(width, ifClause.child(2).getDataType().width());
+
return new ColumnStatisticBuilder()
.setNdv(ndv)
.setMinValue(Double.NEGATIVE_INFINITY)
.setMaxValue(Double.POSITIVE_INFINITY)
- .setAvgSizeByte(8)
+ .setAvgSizeByte(width)
.setNumNulls(0)
.build();
}
@@ -242,9 +253,9 @@ public class ExpressionEstimation extends
ExpressionVisitor<ColumnStatistic, Sta
return new ColumnStatisticBuilder()
.setMaxValue(literalVal)
.setMinValue(literalVal)
- .setNdv(1)
- .setNumNulls(1)
- .setAvgSizeByte(1)
+ .setNdv(literal.isNullLiteral() ? 0 : 1)
+ .setNumNulls(literal.isNullLiteral() ? 1 : 0)
+ .setAvgSizeByte(literal.getDataType().width())
.setMinExpr(literal.toLegacyLiteral())
.setMaxExpr(literal.toLegacyLiteral())
.build();
diff --git
a/fe/fe-core/src/main/java/org/apache/doris/nereids/stats/StatsCalculator.java
b/fe/fe-core/src/main/java/org/apache/doris/nereids/stats/StatsCalculator.java
index 1a983532a94..a7540622303 100644
---
a/fe/fe-core/src/main/java/org/apache/doris/nereids/stats/StatsCalculator.java
+++
b/fe/fe-core/src/main/java/org/apache/doris/nereids/stats/StatsCalculator.java
@@ -825,9 +825,7 @@ public class StatsCalculator extends
DefaultPlanVisitor<Statistics, Void> {
cache = getColumnStatistic(table, colName, idxId);
}
ColumnStatisticBuilder colStatsBuilder = new
ColumnStatisticBuilder(cache);
- if (cache.avgSizeByte <= 0) {
-
colStatsBuilder.setAvgSizeByte(slotReference.getColumn().get().getType().getSlotSize());
- }
+ colStatsBuilder.normalizeAvgSizeByte(slotReference);
if (!cache.isUnKnown) {
rowCount = Math.max(rowCount, cache.count + deltaRowCount);
} else {
diff --git
a/fe/fe-core/src/main/java/org/apache/doris/nereids/types/StringType.java
b/fe/fe-core/src/main/java/org/apache/doris/nereids/types/StringType.java
index 935716e42bf..8e92f83274e 100644
--- a/fe/fe-core/src/main/java/org/apache/doris/nereids/types/StringType.java
+++ b/fe/fe-core/src/main/java/org/apache/doris/nereids/types/StringType.java
@@ -31,11 +31,6 @@ public class StringType extends CharacterType {
super(-1);
}
- @Override
- public int width() {
- return len;
- }
-
@Override
public Type toCatalogDataType() {
return Type.STRING;
diff --git
a/fe/fe-core/src/main/java/org/apache/doris/nereids/types/coercion/CharacterType.java
b/fe/fe-core/src/main/java/org/apache/doris/nereids/types/coercion/CharacterType.java
index c02ea39e39a..153147f8bca 100644
---
a/fe/fe-core/src/main/java/org/apache/doris/nereids/types/coercion/CharacterType.java
+++
b/fe/fe-core/src/main/java/org/apache/doris/nereids/types/coercion/CharacterType.java
@@ -27,8 +27,11 @@ import org.apache.doris.nereids.types.StringType;
public abstract class CharacterType extends PrimitiveType {
private static final int WIDTH = 16;
+ public static final int DEFAULT_WIDTH = WIDTH;
protected final int len;
+ // When defining SQL schemas, users often tend to set the length of string
+ // fields much longer than actually needed for storage.
public CharacterType(int len) {
this.len = len;
diff --git
a/fe/fe-core/src/main/java/org/apache/doris/statistics/ColumnStatisticBuilder.java
b/fe/fe-core/src/main/java/org/apache/doris/statistics/ColumnStatisticBuilder.java
index a512fbadbda..85daa8ea0fe 100644
---
a/fe/fe-core/src/main/java/org/apache/doris/statistics/ColumnStatisticBuilder.java
+++
b/fe/fe-core/src/main/java/org/apache/doris/statistics/ColumnStatisticBuilder.java
@@ -18,6 +18,8 @@
package org.apache.doris.statistics;
import org.apache.doris.analysis.LiteralExpr;
+import org.apache.doris.nereids.trees.expressions.SlotReference;
+import org.apache.doris.nereids.types.coercion.CharacterType;
public class ColumnStatisticBuilder {
private double count;
@@ -170,4 +172,20 @@ public class ColumnStatisticBuilder {
isUnknown, updatedTime);
return colStats;
}
+
+ public void normalizeAvgSizeByte(SlotReference slot) {
+ if (isUnknown) {
+ return;
+ }
+ if (avgSizeByte > 0) {
+ return;
+ }
+ avgSizeByte = slot.getDataType().toCatalogDataType().getSlotSize();
+ // When defining SQL schemas, users often tend to set the length of
string \
+ // fields much longer than actually needed for storage.
+ if (slot.getDataType() instanceof CharacterType) {
+ avgSizeByte = Math.min(avgSizeByte,
+ CharacterType.DEFAULT_WIDTH);
+ }
+ }
}
diff --git
a/fe/fe-core/src/main/java/org/apache/doris/statistics/Statistics.java
b/fe/fe-core/src/main/java/org/apache/doris/statistics/Statistics.java
index a907f6412f1..8a42245e895 100644
--- a/fe/fe-core/src/main/java/org/apache/doris/statistics/Statistics.java
+++ b/fe/fe-core/src/main/java/org/apache/doris/statistics/Statistics.java
@@ -20,6 +20,7 @@ package org.apache.doris.statistics;
import org.apache.doris.nereids.stats.StatsMathUtil;
import org.apache.doris.nereids.trees.expressions.Expression;
import org.apache.doris.nereids.trees.expressions.Slot;
+import org.apache.doris.nereids.types.coercion.CharacterType;
import java.text.DecimalFormat;
import java.util.HashMap;
@@ -143,7 +144,9 @@ public class Statistics {
if (tupleSize <= 0) {
double tempSize = 0.0;
for (ColumnStatistic s : expressionToColumnStats.values()) {
- tempSize += s.avgSizeByte;
+ if (s != null) {
+ tempSize += Math.max(1,
Math.min(CharacterType.DEFAULT_WIDTH, s.avgSizeByte));
+ }
}
tupleSize = Math.max(1, tempSize);
}
diff --git
a/fe/fe-core/src/test/java/org/apache/doris/nereids/stats/ExpressionEstimationTest.java
b/fe/fe-core/src/test/java/org/apache/doris/nereids/stats/ExpressionEstimationTest.java
index 1748802e4dd..32e7504a535 100644
---
a/fe/fe-core/src/test/java/org/apache/doris/nereids/stats/ExpressionEstimationTest.java
+++
b/fe/fe-core/src/test/java/org/apache/doris/nereids/stats/ExpressionEstimationTest.java
@@ -31,7 +31,14 @@ import org.apache.doris.nereids.trees.expressions.WhenClause;
import org.apache.doris.nereids.trees.expressions.functions.agg.Max;
import org.apache.doris.nereids.trees.expressions.functions.agg.Min;
import org.apache.doris.nereids.trees.expressions.functions.scalar.If;
+import org.apache.doris.nereids.trees.expressions.literal.BigIntLiteral;
import org.apache.doris.nereids.trees.expressions.literal.BooleanLiteral;
+import org.apache.doris.nereids.trees.expressions.literal.DateTimeLiteral;
+import org.apache.doris.nereids.trees.expressions.literal.DateV2Literal;
+import org.apache.doris.nereids.trees.expressions.literal.DecimalLiteral;
+import org.apache.doris.nereids.trees.expressions.literal.DoubleLiteral;
+import org.apache.doris.nereids.trees.expressions.literal.NullLiteral;
+import org.apache.doris.nereids.trees.expressions.literal.VarcharLiteral;
import org.apache.doris.nereids.types.DateType;
import org.apache.doris.nereids.types.DoubleType;
import org.apache.doris.nereids.types.IntegerType;
@@ -44,6 +51,7 @@ import org.apache.commons.math3.util.Precision;
import org.junit.jupiter.api.Assertions;
import org.junit.jupiter.api.Test;
+import java.math.BigDecimal;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.List;
@@ -357,6 +365,7 @@ class ExpressionEstimationTest {
CaseWhen caseWhen = new CaseWhen(whens);
ColumnStatistic est = ExpressionEstimation.estimate(caseWhen, stats);
Assertions.assertEquals(est.ndv, 100);
+ Assertions.assertEquals(est.avgSizeByte, 16);
}
@Test
@@ -383,5 +392,59 @@ class ExpressionEstimationTest {
If ifClause = new If(BooleanLiteral.TRUE, a, b);
ColumnStatistic est = ExpressionEstimation.estimate(ifClause, stats);
Assertions.assertEquals(est.ndv, 100);
+ Assertions.assertEquals(est.avgSizeByte, 16);
+ }
+
+ @Test
+ public void testLiteral() {
+ Statistics stats = new Statistics(1000, new HashMap<>());
+
+ BigIntLiteral l1 = new BigIntLiteral(1000000);
+ ColumnStatistic est = ExpressionEstimation.estimate(l1, stats);
+ Assertions.assertEquals(est.ndv, 1);
+ Assertions.assertEquals(est.avgSizeByte, 8);
+ Assertions.assertEquals(est.numNulls, 0);
+
+ VarcharLiteral l2 = new VarcharLiteral("abcdefghij");
+ est = ExpressionEstimation.estimate(l2, stats);
+ Assertions.assertEquals(est.ndv, 1);
+ Assertions.assertEquals(est.avgSizeByte, 10);
+ Assertions.assertEquals(est.numNulls, 0);
+
+ DoubleLiteral l3 = new DoubleLiteral(0.01);
+ est = ExpressionEstimation.estimate(l3, stats);
+ Assertions.assertEquals(est.ndv, 1);
+ Assertions.assertEquals(est.avgSizeByte, 8);
+ Assertions.assertEquals(est.numNulls, 0);
+
+ DateV2Literal l4 = new DateV2Literal("2024-09-10");
+ est = ExpressionEstimation.estimate(l4, stats);
+ Assertions.assertEquals(est.ndv, 1);
+ Assertions.assertEquals(est.avgSizeByte, 4);
+ Assertions.assertEquals(est.numNulls, 0);
+
+ DateTimeLiteral l5 = new DateTimeLiteral("2024-09-10 00:00:00");
+ est = ExpressionEstimation.estimate(l5, stats);
+ Assertions.assertEquals(est.ndv, 1);
+ Assertions.assertEquals(est.avgSizeByte, 16);
+ Assertions.assertEquals(est.numNulls, 0);
+
+ BooleanLiteral l6 = BooleanLiteral.TRUE;
+ est = ExpressionEstimation.estimate(l6, stats);
+ Assertions.assertEquals(est.ndv, 1);
+ Assertions.assertEquals(est.avgSizeByte, 1);
+ Assertions.assertEquals(est.numNulls, 0);
+
+ DecimalLiteral l7 = new DecimalLiteral(BigDecimal.valueOf(2024.0928));
+ est = ExpressionEstimation.estimate(l7, stats);
+ Assertions.assertEquals(est.ndv, 1);
+ Assertions.assertEquals(est.avgSizeByte, 16);
+ Assertions.assertEquals(est.numNulls, 0);
+
+ NullLiteral l8 = new NullLiteral();
+ est = ExpressionEstimation.estimate(l8, stats);
+ Assertions.assertEquals(est.ndv, 0);
+ Assertions.assertEquals(est.avgSizeByte, 1);
+ Assertions.assertEquals(est.numNulls, 1);
}
}
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]