1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
|
/*
* 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.hive.execution
import scala.collection.JavaConverters._
import org.apache.hadoop.conf.Configuration
import org.apache.hadoop.hive.ql.metadata.{Partition => HivePartition}
import org.apache.hadoop.hive.ql.plan.TableDesc
import org.apache.hadoop.hive.serde.serdeConstants
import org.apache.hadoop.hive.serde2.objectinspector._
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspectorUtils.ObjectInspectorCopyOption
import org.apache.hadoop.hive.serde2.typeinfo.TypeInfoUtils
import org.apache.spark.rdd.RDD
import org.apache.spark.sql.SparkSession
import org.apache.spark.sql.catalyst.InternalRow
import org.apache.spark.sql.catalyst.catalog.CatalogRelation
import org.apache.spark.sql.catalyst.expressions._
import org.apache.spark.sql.catalyst.plans.QueryPlan
import org.apache.spark.sql.execution._
import org.apache.spark.sql.execution.metric.SQLMetrics
import org.apache.spark.sql.hive._
import org.apache.spark.sql.hive.client.HiveClientImpl
import org.apache.spark.sql.types.{BooleanType, DataType}
import org.apache.spark.util.Utils
/**
* The Hive table scan operator. Column and partition pruning are both handled.
*
* @param requestedAttributes Attributes to be fetched from the Hive table.
* @param relation The Hive table be scanned.
* @param partitionPruningPred An optional partition pruning predicate for partitioned table.
*/
private[hive]
case class HiveTableScanExec(
requestedAttributes: Seq[Attribute],
relation: CatalogRelation,
partitionPruningPred: Seq[Expression])(
@transient private val sparkSession: SparkSession)
extends LeafExecNode {
require(partitionPruningPred.isEmpty || relation.isPartitioned,
"Partition pruning predicates only supported for partitioned tables.")
override lazy val metrics = Map(
"numOutputRows" -> SQLMetrics.createMetric(sparkContext, "number of output rows"))
override def producedAttributes: AttributeSet = outputSet ++
AttributeSet(partitionPruningPred.flatMap(_.references))
private val originalAttributes = AttributeMap(relation.output.map(a => a -> a))
override val output: Seq[Attribute] = {
// Retrieve the original attributes based on expression ID so that capitalization matches.
requestedAttributes.map(originalAttributes)
}
// Bind all partition key attribute references in the partition pruning predicate for later
// evaluation.
private lazy val boundPruningPred = partitionPruningPred.reduceLeftOption(And).map { pred =>
require(
pred.dataType == BooleanType,
s"Data type of predicate $pred must be BooleanType rather than ${pred.dataType}.")
BindReferences.bindReference(pred, relation.partitionCols)
}
@transient private lazy val hiveQlTable = HiveClientImpl.toHiveTable(relation.tableMeta)
@transient private lazy val tableDesc = new TableDesc(
hiveQlTable.getInputFormatClass,
hiveQlTable.getOutputFormatClass,
hiveQlTable.getMetadata)
// Create a local copy of hadoopConf,so that scan specific modifications should not impact
// other queries
@transient private lazy val hadoopConf = {
val c = sparkSession.sessionState.newHadoopConf()
// append columns ids and names before broadcast
addColumnMetadataToConf(c)
c
}
@transient private lazy val hadoopReader = new HadoopTableReader(
output,
relation.partitionCols,
tableDesc,
sparkSession,
hadoopConf)
private def castFromString(value: String, dataType: DataType) = {
Cast(Literal(value), dataType).eval(null)
}
private def addColumnMetadataToConf(hiveConf: Configuration): Unit = {
// Specifies needed column IDs for those non-partitioning columns.
val columnOrdinals = AttributeMap(relation.dataCols.zipWithIndex)
val neededColumnIDs = output.flatMap(columnOrdinals.get).map(o => o: Integer)
HiveShim.appendReadColumns(hiveConf, neededColumnIDs, output.map(_.name))
val deserializer = tableDesc.getDeserializerClass.newInstance
deserializer.initialize(hiveConf, tableDesc.getProperties)
// Specifies types and object inspectors of columns to be scanned.
val structOI = ObjectInspectorUtils
.getStandardObjectInspector(
deserializer.getObjectInspector,
ObjectInspectorCopyOption.JAVA)
.asInstanceOf[StructObjectInspector]
val columnTypeNames = structOI
.getAllStructFieldRefs.asScala
.map(_.getFieldObjectInspector)
.map(TypeInfoUtils.getTypeInfoFromObjectInspector(_).getTypeName)
.mkString(",")
hiveConf.set(serdeConstants.LIST_COLUMN_TYPES, columnTypeNames)
hiveConf.set(serdeConstants.LIST_COLUMNS, relation.dataCols.map(_.name).mkString(","))
}
/**
* Prunes partitions not involve the query plan.
*
* @param partitions All partitions of the relation.
* @return Partitions that are involved in the query plan.
*/
private[hive] def prunePartitions(partitions: Seq[HivePartition]) = {
boundPruningPred match {
case None => partitions
case Some(shouldKeep) => partitions.filter { part =>
val dataTypes = relation.partitionCols.map(_.dataType)
val castedValues = part.getValues.asScala.zip(dataTypes)
.map { case (value, dataType) => castFromString(value, dataType) }
// Only partitioned values are needed here, since the predicate has already been bound to
// partition key attribute references.
val row = InternalRow.fromSeq(castedValues)
shouldKeep.eval(row).asInstanceOf[Boolean]
}
}
}
// exposed for tests
@transient lazy val rawPartitions = {
val prunedPartitions = if (sparkSession.sessionState.conf.metastorePartitionPruning) {
// Retrieve the original attributes based on expression ID so that capitalization matches.
val normalizedFilters = partitionPruningPred.map(_.transform {
case a: AttributeReference => originalAttributes(a)
})
sparkSession.sharedState.externalCatalog.listPartitionsByFilter(
relation.tableMeta.database,
relation.tableMeta.identifier.table,
normalizedFilters,
sparkSession.sessionState.conf.sessionLocalTimeZone)
} else {
sparkSession.sharedState.externalCatalog.listPartitions(
relation.tableMeta.database,
relation.tableMeta.identifier.table)
}
prunedPartitions.map(HiveClientImpl.toHivePartition(_, hiveQlTable))
}
protected override def doExecute(): RDD[InternalRow] = {
// Using dummyCallSite, as getCallSite can turn out to be expensive with
// with multiple partitions.
val rdd = if (!relation.isPartitioned) {
Utils.withDummyCallSite(sqlContext.sparkContext) {
hadoopReader.makeRDDForTable(hiveQlTable)
}
} else {
Utils.withDummyCallSite(sqlContext.sparkContext) {
hadoopReader.makeRDDForPartitionedTable(prunePartitions(rawPartitions))
}
}
val numOutputRows = longMetric("numOutputRows")
// Avoid to serialize MetastoreRelation because schema is lazy. (see SPARK-15649)
val outputSchema = schema
rdd.mapPartitionsWithIndexInternal { (index, iter) =>
val proj = UnsafeProjection.create(outputSchema)
proj.initialize(index)
iter.map { r =>
numOutputRows += 1
proj(r)
}
}
}
override lazy val canonicalized: HiveTableScanExec = {
val input: AttributeSeq = relation.output
HiveTableScanExec(
requestedAttributes.map(QueryPlan.normalizeExprId(_, input)),
relation.canonicalized.asInstanceOf[CatalogRelation],
partitionPruningPred.map(QueryPlan.normalizeExprId(_, input)))(sparkSession)
}
override def otherCopyArgs: Seq[AnyRef] = Seq(sparkSession)
}
|
