This is an automated email from the ASF dual-hosted git repository.
xxyu pushed a commit to branch kylin5
in repository https://gitbox.apache.org/repos/asf/kylin.git
commit 8f38936d58f198d60f7bde8ff233b26b015754f8
Author: sibingzhang <74443791+sibingzh...@users.noreply.github.com>
AuthorDate: Fri Jan 27 16:42:20 2023 +0800
KYLIN-5496 Fix the query result is incorrect after the character type in
the 'yyyy-mm-dd' format is converted to the timestamp type and then filtered
---
.../kylin/query/routing/RealizationPruner.java | 2 +-
.../sql/execution/datasource/FilePruner.scala | 31 +++++++++++++---------
2 files changed, 20 insertions(+), 13 deletions(-)
diff --git
a/src/query-common/src/main/java/org/apache/kylin/query/routing/RealizationPruner.java
b/src/query-common/src/main/java/org/apache/kylin/query/routing/RealizationPruner.java
index b429cf0407..dc54aae001 100644
---
a/src/query-common/src/main/java/org/apache/kylin/query/routing/RealizationPruner.java
+++
b/src/query-common/src/main/java/org/apache/kylin/query/routing/RealizationPruner.java
@@ -268,7 +268,7 @@ public class RealizationPruner {
private static String normalization(String dateFormat, RexLiteral
rexLiteral) {
RelDataTypeFamily typeFamily = rexLiteral.getType().getFamily();
- if (SqlTypeFamily.DATE == typeFamily) {
+ if (SqlTypeFamily.DATE == typeFamily || SqlTypeFamily.TIMESTAMP ==
typeFamily) {
// Calendar uses UTC timezone, just to keep RexLiteral's value(an
instanceof DateString)
long timeInMillis = ((Calendar)
rexLiteral.getValue()).getTimeInMillis();
String dateStr = DateFormat.formatToDateStr(timeInMillis,
dateFormat, RealizationPruner.UTC_ZONE);
diff --git
a/src/spark-project/spark-common/src/main/scala/org/apache/spark/sql/execution/datasource/FilePruner.scala
b/src/spark-project/spark-common/src/main/scala/org/apache/spark/sql/execution/datasource/FilePruner.scala
index b658a0bd7e..b97d402217 100644
---
a/src/spark-project/spark-common/src/main/scala/org/apache/spark/sql/execution/datasource/FilePruner.scala
+++
b/src/spark-project/spark-common/src/main/scala/org/apache/spark/sql/execution/datasource/FilePruner.scala
@@ -26,6 +26,7 @@ import org.apache.kylin.common.util.{DateFormat, HadoopUtil}
import org.apache.kylin.common.{KapConfig, KylinConfig, QueryContext}
import org.apache.kylin.engine.spark.utils.{LogEx, LogUtils}
import org.apache.kylin.metadata.cube.model.{DimensionRangeInfo, LayoutEntity,
NDataflow, NDataflowManager}
+import org.apache.kylin.metadata.datatype.DataType
import org.apache.kylin.metadata.model.{PartitionDesc, TblColRef}
import org.apache.kylin.metadata.project.NProjectManager
import org.apache.spark.internal.Logging
@@ -731,9 +732,9 @@ abstract class PushableColumnBase {
case class SegDimFilters(dimRange: java.util.Map[String, DimensionRangeInfo],
dimCols: java.util.Map[Integer, TblColRef]) extends Logging {
private def insurance(id: String, value: Any)
- (func: String => Filter): Filter = {
+ (func: Any => Filter): Filter = {
if (dimRange.containsKey(id) && dimCols.containsKey(id.toInt)) {
- func(value.toString)
+ func(value)
} else {
Trivial(true)
}
@@ -749,23 +750,29 @@ case class SegDimFilters(dimRange: java.util.Map[String,
DimensionRangeInfo], di
}
def foldFilter(filter: Filter): Filter = {
+
+ def getDataType(col: String, value: Any): DataType = {
+ if (value.isInstanceOf[Date] || value.isInstanceOf[Timestamp]) return
DataType.getType("date")
+ dimCols.get(col.toInt).getType
+ }
+
filter match {
case EqualTo(id, value: Any) =>
val col = escapeQuote(id)
insurance(col, value) {
ts => {
- val dataType = dimCols.get(col.toInt).getType
- Trivial(dataType.compare(ts, dimRange.get(col).getMin) >= 0
- && dataType.compare(ts, dimRange.get(col).getMax) <= 0)
+ val dataType = getDataType(col, value)
+ Trivial(dataType.compare(ts.toString, dimRange.get(col).getMin) >= 0
+ && dataType.compare(ts.toString, dimRange.get(col).getMax) <= 0)
}
}
case In(id, values: Array[Any]) =>
val col = escapeQuote(id)
val satisfied = values.map(v => insurance(col, v) {
ts => {
- val dataType = dimCols.get(col.toInt).getType
- Trivial(dataType.compare(ts, dimRange.get(col).getMin) >= 0
- && dataType.compare(ts, dimRange.get(col).getMax) <= 0)
+ val dataType = getDataType(col, v)
+ Trivial(dataType.compare(ts.toString, dimRange.get(col).getMin) >= 0
+ && dataType.compare(ts.toString, dimRange.get(col).getMax) <= 0)
}
}).exists(_.equals(Trivial(true)))
Trivial(satisfied)
@@ -777,22 +784,22 @@ case class SegDimFilters(dimRange: java.util.Map[String,
DimensionRangeInfo], di
case GreaterThan(id, value: Any) =>
val col = escapeQuote(id)
insurance(col, value) {
- ts => Trivial(dimCols.get(col.toInt).getType.compare(ts,
dimRange.get(col).getMax) < 0)
+ ts => Trivial(getDataType(col, value).compare(ts.toString,
dimRange.get(col).getMax) < 0)
}
case GreaterThanOrEqual(id, value: Any) =>
val col = escapeQuote(id)
insurance(col, value) {
- ts => Trivial(dimCols.get(col.toInt).getType.compare(ts,
dimRange.get(col).getMax) <= 0)
+ ts => Trivial(getDataType(col, value).compare(ts.toString,
dimRange.get(col).getMax) <= 0)
}
case LessThan(id, value: Any) =>
val col = escapeQuote(id)
insurance(col, value) {
- ts => Trivial(dimCols.get(col.toInt).getType.compare(ts,
dimRange.get(col).getMin) > 0)
+ ts => Trivial(getDataType(col, value).compare(ts.toString,
dimRange.get(col).getMin) > 0)
}
case LessThanOrEqual(id, value: Any) =>
val col = escapeQuote(id)
insurance(col, value) {
- ts => Trivial(dimCols.get(col.toInt).getType.compare(ts,
dimRange.get(col).getMin) >= 0)
+ ts => Trivial(getDataType(col, value).compare(ts.toString,
dimRange.get(col).getMin) >= 0)
}
case And(left: Filter, right: Filter) =>
And(foldFilter(left), foldFilter(right)) match {
