Github user wzhfy commented on a diff in the pull request: https://github.com/apache/spark/pull/15090#discussion_r80626981 --- Diff: sql/core/src/main/scala/org/apache/spark/sql/execution/command/AnalyzeColumnCommand.scala --- @@ -0,0 +1,179 @@ +/* + * 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.command + +import scala.collection.mutable + +import org.apache.spark.sql._ +import org.apache.spark.sql.catalyst.{InternalRow, TableIdentifier} +import org.apache.spark.sql.catalyst.analysis.EliminateSubqueryAliases +import org.apache.spark.sql.catalyst.catalog.{CatalogRelation, CatalogTable} +import org.apache.spark.sql.catalyst.expressions._ +import org.apache.spark.sql.catalyst.expressions.aggregate._ +import org.apache.spark.sql.catalyst.plans.logical.{Aggregate, ColumnStat, LogicalPlan, Statistics} +import org.apache.spark.sql.execution.datasources.LogicalRelation +import org.apache.spark.sql.types._ + + +/** + * Analyzes the given columns of the given table in the current database to generate statistics, + * which will be used in query optimizations. + */ +case class AnalyzeColumnCommand( + tableIdent: TableIdentifier, + columnNames: Seq[String]) extends RunnableCommand { + + override def run(sparkSession: SparkSession): Seq[Row] = { + val sessionState = sparkSession.sessionState + val db = tableIdent.database.getOrElse(sessionState.catalog.getCurrentDatabase) + val tableIdentWithDB = TableIdentifier(tableIdent.table, Some(db)) + val relation = EliminateSubqueryAliases(sessionState.catalog.lookupRelation(tableIdentWithDB)) + + relation match { + case catalogRel: CatalogRelation => + updateStats(catalogRel.catalogTable, + AnalyzeTableCommand.calculateTotalSize(sessionState, catalogRel.catalogTable)) + + case logicalRel: LogicalRelation if logicalRel.catalogTable.isDefined => + updateStats(logicalRel.catalogTable.get, logicalRel.relation.sizeInBytes) + + case otherRelation => + throw new AnalysisException("ANALYZE TABLE is not supported for " + + s"${otherRelation.nodeName}.") + } + + def updateStats(catalogTable: CatalogTable, newTotalSize: Long): Unit = { + val (rowCount, columnStats) = computeColStats(sparkSession, relation) + val statistics = Statistics( + sizeInBytes = newTotalSize, + rowCount = Some(rowCount), + colStats = columnStats ++ catalogTable.stats.map(_.colStats).getOrElse(Map())) + sessionState.catalog.alterTable(catalogTable.copy(stats = Some(statistics))) + // Refresh the cached data source table in the catalog. + sessionState.catalog.refreshTable(tableIdentWithDB) + } + + Seq.empty[Row] + } + + def computeColStats( + sparkSession: SparkSession, + relation: LogicalPlan): (Long, Map[String, ColumnStat]) = { + + // check correctness of column names + val attributesToAnalyze = mutable.MutableList[Attribute]() + val caseSensitive = sparkSession.sessionState.conf.caseSensitiveAnalysis + columnNames.foreach { col => + val exprOption = relation.output.find { attr => + if (caseSensitive) attr.name == col else attr.name.equalsIgnoreCase(col) + } + val expr = exprOption.getOrElse(throw new AnalysisException(s"Invalid column name: $col.")) + // do deduplication + if (!attributesToAnalyze.contains(expr)) { + attributesToAnalyze += expr + } + } + + // Collect statistics per column. + // The first element in the result will be the overall row count, the following elements + // will be structs containing all column stats. + // The layout of each struct follows the layout of the ColumnStats. + val ndvMaxErr = sparkSession.sessionState.conf.ndvMaxError + val expressions = Count(Literal(1)).toAggregateExpression() +: + attributesToAnalyze.map(ColumnStatStruct(_, ndvMaxErr)) + val namedExpressions = expressions.map(e => Alias(e, e.toString)()) + val statsRow = Dataset.ofRows(sparkSession, Aggregate(Nil, namedExpressions, relation)) + .queryExecution.toRdd.collect().head + + // unwrap the result + val rowCount = statsRow.getLong(0) + val columnStats = attributesToAnalyze.zipWithIndex.map { case (expr, i) => + (expr.name, ColumnStatStruct.unwrapStruct(statsRow, i + 1, expr, ndvMaxErr, rowCount)) + }.toMap + (rowCount, columnStats) + } +} + +object ColumnStatStruct { + val zero = Literal(0, LongType) + val one = Literal(1, LongType) + + def numNulls(e: Expression): Expression = if (e.nullable) Sum(If(IsNull(e), one, zero)) else zero + def max(e: Expression): Expression = Max(e) + def min(e: Expression): Expression = Min(e) + def ndv(e: Expression, relativeSD: Double): Expression = HyperLogLogPlusPlus(e, relativeSD) + def avgLength(e: Expression): Expression = Average(Length(e)) + def maxLength(e: Expression): Expression = Max(Length(e)) + def numTrues(e: Expression): Expression = Sum(If(e, one, zero)) + def numFalses(e: Expression): Expression = Sum(If(Not(e), one, zero)) + + def getStruct(exprs: Seq[Expression]): CreateStruct = { + CreateStruct(exprs.map { + case af: AggregateFunction => af.toAggregateExpression() + case e: Expression => e + }) + } + + def numericColumnStat(e: Expression, relativeSD: Double): Seq[Expression] = { + Seq(numNulls(e), max(e), min(e), ndv(e, relativeSD)) + } + + def stringColumnStat(e: Expression, relativeSD: Double): Seq[Expression] = { + Seq(numNulls(e), avgLength(e), maxLength(e), ndv(e, relativeSD)) + } + + def binaryColumnStat(e: Expression): Seq[Expression] = { + Seq(numNulls(e), avgLength(e), maxLength(e)) + } + + def booleanColumnStat(e: Expression): Seq[Expression] = { + Seq(numNulls(e), numTrues(e), numFalses(e)) + } + + def apply(e: Attribute, relativeSD: Double): CreateStruct = e.dataType match { + // Use aggregate functions to compute statistics we need. + case _: NumericType | TimestampType | DateType => getStruct(numericColumnStat(e, relativeSD)) + case StringType => getStruct(stringColumnStat(e, relativeSD)) + case BinaryType => getStruct(binaryColumnStat(e)) + case BooleanType => getStruct(booleanColumnStat(e)) + case otherType => + throw new AnalysisException("Analyzing columns is not supported for column " + + s"${e.name} of data type: ${e.dataType}.") + } + + def unwrapStruct( + row: InternalRow, + offset: Int, + e: Expression, + relativeSD: Double, + rowCount: Long): ColumnStat = { + val numFields = e.dataType match { + case _: NumericType | TimestampType | DateType => numericColumnStat(e, relativeSD).length + case StringType => stringColumnStat(e, relativeSD).length + case BinaryType => binaryColumnStat(e).length + case BooleanType => booleanColumnStat(e).length + } + val struct = row.getStruct(offset, numFields) + // NumericType, TimestampType, DateType and StringType have ndv and its index is 3. + if (numFields >= 4 && !struct.isNullAt(3)) { + // ndv should not be larger than number of rows + if (struct.getLong(3) > rowCount) struct.asInstanceOf[UnsafeRow].setLong(3, rowCount) --- End diff -- ok, thanks!
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. --- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org