gene-db commented on code in PR #48172:
URL: https://github.com/apache/spark/pull/48172#discussion_r1792133835
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileScanRDD.scala:
##########
@@ -72,9 +74,30 @@ case class PartitionedFile(
}
}
+/**
+ * Class used to store statistical data that is collected during a file scan
and could be used to
+ * update the SQL metrics of the scan node. More members could be added to
this class to to collect
+ * metrics related to new features.
+ */
+case class FileScanMetrics(
+ topLevelVariantMetrics: Option[VariantMetrics] = None,
Review Comment:
Why are these optional? It looks like `fileScanMetrics` is already optional
to the rdd, so maybe we don't need these to be optional? Or, what is the
scenario where one or both of these would be optional?
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileScanRDD.scala:
##########
@@ -83,7 +106,9 @@ class FileScanRDD(
val readSchema: StructType,
val metadataColumns: Seq[AttributeReference] = Seq.empty,
metadataExtractors: Map[String, PartitionedFile => Any] = Map.empty,
- options: FileSourceOptions = new
FileSourceOptions(CaseInsensitiveMap(Map.empty)))
+ options: FileSourceOptions = new
FileSourceOptions(CaseInsensitiveMap(Map.empty)),
+ fileScanMetrics: Option[FileScanMetrics] = None,
Review Comment:
Would it make sense that `FileScanMetrics` keeps track of its own
`sqlMetrics: Map[String, SQLMetric]`?
It seems like this `FileScanRDD` does not use `sqlMetrics` independently
from `FileScanMetrics`, and `FileScanMetrics` always needs the `sqlMetrics`.
They seem tightly coupled/related. Also, I would imagine any additional metrics
in the future would also want to update the `sqlMetrics`.
Then, maybe the `sqlMetrics` within the `FileScanMetrics` doesn't have to be
an `Option`, which could simplify things.
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/metric/VariantConstructionMetrics.scala:
##########
@@ -0,0 +1,161 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements. See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.spark.sql.execution.metric
+
+import org.apache.spark.SparkContext
+import org.apache.spark.types.variant.VariantMetrics
+
+case class VariantMetricDescriptor (
+ name: String,
+ metricType: String,
+ description: String
+)
+
+object VariantConstructionMetrics {
+ // Top level variant metrics
+ val VARIANT_BUILDER_TOP_LEVEL_NUMBER_OF_VARIANTS =
"variantBuilderTopLevelNumVariants"
+ val VARIANT_BUILDER_TOP_LEVEL_BYTE_SIZE_BOUND =
"variantBuilderTopLevelByteSizeBound"
+ val VARIANT_BUILDER_TOP_LEVEL_NUM_SCALARS =
"variantBuilderTopLevelNumScalars"
+ val VARIANT_BUILDER_TOP_LEVEL_NUM_PATHS = "variantBuilderTopLevelNumPaths"
+ val VARIANT_BUILDER_TOP_LEVEL_MAX_DEPTH = "variantBuilderTopLevelMaxDepth"
+ // Nested variant metrics
+ val VARIANT_BUILDER_NESTED_NUMBER_OF_VARIANTS =
"variantBuilderNestedNumVariants"
+ val VARIANT_BUILDER_NESTED_BYTE_SIZE_BOUND =
"variantBuilderNestedByteSizeBound"
+ val VARIANT_BUILDER_NESTED_NUM_SCALARS = "variantBuilderNestedNumScalars"
+ val VARIANT_BUILDER_NESTED_NUM_PATHS = "variantBuilderNestedNumPaths"
+ val VARIANT_BUILDER_NESTED_MAX_DEPTH = "variantBuilderNestedMaxDepth"
+
+ final val all: Array[VariantMetricDescriptor] = Array(
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_TOP_LEVEL_NUMBER_OF_VARIANTS,
+ SQLMetrics.SUM_METRIC,
+ "variant top-level - total count"
+ ),
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_TOP_LEVEL_BYTE_SIZE_BOUND,
+ SQLMetrics.SIZE_METRIC,
+ "variant top-level - total byte size"
+ ),
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_TOP_LEVEL_NUM_SCALARS,
+ SQLMetrics.SUM_METRIC,
+ "variant top-level - total number of scalar values"
+ ),
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_TOP_LEVEL_NUM_PATHS,
+ SQLMetrics.SUM_METRIC,
+ "variant top-level - total number of paths"
+ ),
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_TOP_LEVEL_MAX_DEPTH,
+ SQLMetrics.MAX_METRIC,
+ "variant top-level - max depth"
+ ),
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_NESTED_NUMBER_OF_VARIANTS,
+ SQLMetrics.SUM_METRIC,
+ "variant nested - total count"
+ ),
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_NESTED_BYTE_SIZE_BOUND,
+ SQLMetrics.SIZE_METRIC,
+ "variant nested - total byte size"
+ ),
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_NESTED_NUM_SCALARS,
+ SQLMetrics.SUM_METRIC,
+ "variant nested - total number of scalar values"
+ ),
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_NESTED_NUM_PATHS,
+ SQLMetrics.SUM_METRIC,
+ "variant nested - total number of paths"
+ ),
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_NESTED_MAX_DEPTH,
+ SQLMetrics.MAX_METRIC,
+ "variant nested - max depth"
+ )
+ )
+
+ def createSQLMetrics(sparkContext: SparkContext): Map[String, SQLMetric] = {
+ all.map(
+ d => {
+ d.name -> {
+ d.metricType match {
+ case SQLMetrics.SUM_METRIC =>
SQLMetrics.createMetric(sparkContext, d.description)
+ case SQLMetrics.SIZE_METRIC =>
SQLMetrics.createSizeMetric(sparkContext, d.description)
+ case SQLMetrics.MAX_METRIC =>
SQLMetrics.createMaxMetric(sparkContext, d.description)
+ case _ => throw new IllegalArgumentException(s"Unknown metric
type: ${d.metricType}")
+ }
+ }
+ }
+ ).toMap
+ }
+
+ // Update SQL metrics with the data in topLevelVariantMetrics and
nestedVariantMetrics
+ def updateSQLMetrics(
+ topLevelVariantMetrics: Option[VariantMetrics],
+ nestedVariantMetrics: Option[VariantMetrics],
+ sqlMetrics: Option[Map[String, SQLMetric]]): Unit = {
+ sqlMetrics match {
+ case Some(sqlMetrics) =>
+ topLevelVariantMetrics match {
+ case Some(metrics) =>
+ sqlMetrics(VARIANT_BUILDER_TOP_LEVEL_NUMBER_OF_VARIANTS)
+ .add(metrics.variantCount)
+ sqlMetrics(VARIANT_BUILDER_TOP_LEVEL_BYTE_SIZE_BOUND)
+ .add(metrics.byteSize)
+ sqlMetrics(VARIANT_BUILDER_TOP_LEVEL_NUM_SCALARS)
+ .add(metrics.numScalars)
+ sqlMetrics(VARIANT_BUILDER_TOP_LEVEL_NUM_PATHS)
+ .add(metrics.numPaths)
+ sqlMetrics(VARIANT_BUILDER_TOP_LEVEL_MAX_DEPTH)
+ .set(
+ Math.max(
+ sqlMetrics(VARIANT_BUILDER_TOP_LEVEL_MAX_DEPTH).value,
+ metrics.maxDepth
+ )
+ )
+ metrics.reset()
Review Comment:
Ahhh, so the input metrics will be reset after this update. We should make
that clear/explicit in the function comments.
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/json/JacksonParser.scala:
##########
@@ -119,15 +121,20 @@ class JacksonParser(
private val variantAllowDuplicateKeys =
SQLConf.get.getConf(SQLConf.VARIANT_ALLOW_DUPLICATE_KEYS)
- protected final def parseVariant(parser: JsonParser): VariantVal = {
+ // isTopLevel helps in distinguishing between top level variant and nested
variants for metrics.
+ private final def parseVariant(isTopLevel: Boolean)(parser: JsonParser):
VariantVal = {
// Skips `FIELD_NAME` at the beginning. This check is adapted from
`parseJsonToken`, but we
// cannot directly use the function here because it also handles the
`VALUE_NULL` token and
// returns null (representing a SQL NULL). Instead, we want to return a
variant null.
if (parser.getCurrentToken == FIELD_NAME) {
parser.nextToken()
}
try {
- val v = VariantBuilder.parseJson(parser, variantAllowDuplicateKeys)
+ val vm = if (isTopLevel) topLevelVariantMetrics else nestedVariantMetrics
+ val v = vm match {
Review Comment:
This could probably be something like:
```
val v = VariantBuilder.parseJson(parser, variantAllowDuplicateKeys,
vm.getOrElse(new VariantMetrics()))
```
Maybe we won't need the 2 versions of `VariantBuilder.parseJson()`?
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala:
##########
@@ -628,18 +645,39 @@ case class FileSourceScanExec(
}
}
+ val topLevelVariantMetrics: VariantMetrics = new VariantMetrics()
Review Comment:
We should add some comments for these member variables.
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/metric/VariantConstructionMetrics.scala:
##########
@@ -0,0 +1,161 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements. See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.spark.sql.execution.metric
+
+import org.apache.spark.SparkContext
+import org.apache.spark.types.variant.VariantMetrics
+
+case class VariantMetricDescriptor (
+ name: String,
+ metricType: String,
+ description: String
+)
+
+object VariantConstructionMetrics {
+ // Top level variant metrics
+ val VARIANT_BUILDER_TOP_LEVEL_NUMBER_OF_VARIANTS =
"variantBuilderTopLevelNumVariants"
+ val VARIANT_BUILDER_TOP_LEVEL_BYTE_SIZE_BOUND =
"variantBuilderTopLevelByteSizeBound"
+ val VARIANT_BUILDER_TOP_LEVEL_NUM_SCALARS =
"variantBuilderTopLevelNumScalars"
+ val VARIANT_BUILDER_TOP_LEVEL_NUM_PATHS = "variantBuilderTopLevelNumPaths"
+ val VARIANT_BUILDER_TOP_LEVEL_MAX_DEPTH = "variantBuilderTopLevelMaxDepth"
+ // Nested variant metrics
+ val VARIANT_BUILDER_NESTED_NUMBER_OF_VARIANTS =
"variantBuilderNestedNumVariants"
+ val VARIANT_BUILDER_NESTED_BYTE_SIZE_BOUND =
"variantBuilderNestedByteSizeBound"
+ val VARIANT_BUILDER_NESTED_NUM_SCALARS = "variantBuilderNestedNumScalars"
+ val VARIANT_BUILDER_NESTED_NUM_PATHS = "variantBuilderNestedNumPaths"
+ val VARIANT_BUILDER_NESTED_MAX_DEPTH = "variantBuilderNestedMaxDepth"
+
+ final val all: Array[VariantMetricDescriptor] = Array(
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_TOP_LEVEL_NUMBER_OF_VARIANTS,
+ SQLMetrics.SUM_METRIC,
+ "variant top-level - total count"
+ ),
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_TOP_LEVEL_BYTE_SIZE_BOUND,
+ SQLMetrics.SIZE_METRIC,
+ "variant top-level - total byte size"
+ ),
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_TOP_LEVEL_NUM_SCALARS,
+ SQLMetrics.SUM_METRIC,
+ "variant top-level - total number of scalar values"
+ ),
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_TOP_LEVEL_NUM_PATHS,
+ SQLMetrics.SUM_METRIC,
+ "variant top-level - total number of paths"
+ ),
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_TOP_LEVEL_MAX_DEPTH,
+ SQLMetrics.MAX_METRIC,
+ "variant top-level - max depth"
+ ),
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_NESTED_NUMBER_OF_VARIANTS,
+ SQLMetrics.SUM_METRIC,
+ "variant nested - total count"
+ ),
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_NESTED_BYTE_SIZE_BOUND,
+ SQLMetrics.SIZE_METRIC,
+ "variant nested - total byte size"
+ ),
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_NESTED_NUM_SCALARS,
+ SQLMetrics.SUM_METRIC,
+ "variant nested - total number of scalar values"
+ ),
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_NESTED_NUM_PATHS,
+ SQLMetrics.SUM_METRIC,
+ "variant nested - total number of paths"
+ ),
+ VariantMetricDescriptor(
+ VARIANT_BUILDER_NESTED_MAX_DEPTH,
+ SQLMetrics.MAX_METRIC,
+ "variant nested - max depth"
+ )
+ )
+
+ def createSQLMetrics(sparkContext: SparkContext): Map[String, SQLMetric] = {
+ all.map(
+ d => {
+ d.name -> {
+ d.metricType match {
+ case SQLMetrics.SUM_METRIC =>
SQLMetrics.createMetric(sparkContext, d.description)
+ case SQLMetrics.SIZE_METRIC =>
SQLMetrics.createSizeMetric(sparkContext, d.description)
+ case SQLMetrics.MAX_METRIC =>
SQLMetrics.createMaxMetric(sparkContext, d.description)
+ case _ => throw new IllegalArgumentException(s"Unknown metric
type: ${d.metricType}")
+ }
+ }
+ }
+ ).toMap
+ }
+
+ // Update SQL metrics with the data in topLevelVariantMetrics and
nestedVariantMetrics
+ def updateSQLMetrics(
+ topLevelVariantMetrics: Option[VariantMetrics],
+ nestedVariantMetrics: Option[VariantMetrics],
+ sqlMetrics: Option[Map[String, SQLMetric]]): Unit = {
Review Comment:
We should just make this not optional, since there is no reason to update if
there is no `sqlMetrics`.
--
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]
