Search in sources :

Example 26 with TableScan

use of org.apache.beam.vendor.calcite.v1_28_0.org.apache.calcite.rel.core.TableScan in project drill by axbaretto.

the class HivePushPartitionFilterIntoScan method getFilterOnProject.

public static final StoragePluginOptimizerRule getFilterOnProject(OptimizerRulesContext optimizerRulesContext, final String defaultPartitionValue) {
    return new PruneScanRule(RelOptHelper.some(DrillFilterRel.class, RelOptHelper.some(DrillProjectRel.class, RelOptHelper.any(DrillScanRel.class))), "HivePushPartitionFilterIntoScan:Filter_On_Project_Hive", optimizerRulesContext) {

        @Override
        public PartitionDescriptor getPartitionDescriptor(PlannerSettings settings, TableScan scanRel) {
            return new HivePartitionDescriptor(settings, (DrillScanRel) scanRel, getOptimizerRulesContext().getManagedBuffer(), defaultPartitionValue);
        }

        @Override
        public boolean matches(RelOptRuleCall call) {
            final DrillScanRel scan = (DrillScanRel) call.rel(2);
            GroupScan groupScan = scan.getGroupScan();
            // this rule is applicable only for Hive based partition pruning
            if (PrelUtil.getPlannerSettings(scan.getCluster().getPlanner()).isHepPartitionPruningEnabled()) {
                return groupScan instanceof HiveScan && groupScan.supportsPartitionFilterPushdown() && !scan.partitionFilterPushdown();
            } else {
                return groupScan instanceof HiveScan && groupScan.supportsPartitionFilterPushdown();
            }
        }

        @Override
        public void onMatch(RelOptRuleCall call) {
            final DrillFilterRel filterRel = call.rel(0);
            final DrillProjectRel projectRel = call.rel(1);
            final DrillScanRel scanRel = call.rel(2);
            doOnMatch(call, filterRel, projectRel, scanRel);
        }
    };
}
Also used : GroupScan(org.apache.drill.exec.physical.base.GroupScan) TableScan(org.apache.calcite.rel.core.TableScan) DrillScanRel(org.apache.drill.exec.planner.logical.DrillScanRel) PruneScanRule(org.apache.drill.exec.planner.logical.partition.PruneScanRule) PlannerSettings(org.apache.drill.exec.planner.physical.PlannerSettings) DrillProjectRel(org.apache.drill.exec.planner.logical.DrillProjectRel) HiveScan(org.apache.drill.exec.store.hive.HiveScan) DrillFilterRel(org.apache.drill.exec.planner.logical.DrillFilterRel) RelOptRuleCall(org.apache.calcite.plan.RelOptRuleCall) HivePartitionDescriptor(org.apache.drill.exec.planner.sql.HivePartitionDescriptor)

Example 27 with TableScan

use of org.apache.beam.vendor.calcite.v1_28_0.org.apache.calcite.rel.core.TableScan in project drill by axbaretto.

the class HivePushPartitionFilterIntoScan method getFilterOnScan.

public static final StoragePluginOptimizerRule getFilterOnScan(OptimizerRulesContext optimizerRulesContext, final String defaultPartitionValue) {
    return new PruneScanRule(RelOptHelper.some(DrillFilterRel.class, RelOptHelper.any(DrillScanRel.class)), "HivePushPartitionFilterIntoScan:Filter_On_Scan_Hive", optimizerRulesContext) {

        @Override
        public PartitionDescriptor getPartitionDescriptor(PlannerSettings settings, TableScan scanRel) {
            return new HivePartitionDescriptor(settings, (DrillScanRel) scanRel, getOptimizerRulesContext().getManagedBuffer(), defaultPartitionValue);
        }

        @Override
        public boolean matches(RelOptRuleCall call) {
            final DrillScanRel scan = (DrillScanRel) call.rel(1);
            GroupScan groupScan = scan.getGroupScan();
            // this rule is applicable only for Hive based partition pruning
            if (PrelUtil.getPlannerSettings(scan.getCluster().getPlanner()).isHepPartitionPruningEnabled()) {
                return groupScan instanceof HiveScan && groupScan.supportsPartitionFilterPushdown() && !scan.partitionFilterPushdown();
            } else {
                return groupScan instanceof HiveScan && groupScan.supportsPartitionFilterPushdown();
            }
        }

        @Override
        public void onMatch(RelOptRuleCall call) {
            final DrillFilterRel filterRel = call.rel(0);
            final DrillScanRel scanRel = call.rel(1);
            doOnMatch(call, filterRel, null, scanRel);
        }
    };
}
Also used : GroupScan(org.apache.drill.exec.physical.base.GroupScan) TableScan(org.apache.calcite.rel.core.TableScan) DrillScanRel(org.apache.drill.exec.planner.logical.DrillScanRel) PruneScanRule(org.apache.drill.exec.planner.logical.partition.PruneScanRule) PlannerSettings(org.apache.drill.exec.planner.physical.PlannerSettings) HiveScan(org.apache.drill.exec.store.hive.HiveScan) DrillFilterRel(org.apache.drill.exec.planner.logical.DrillFilterRel) RelOptRuleCall(org.apache.calcite.plan.RelOptRuleCall) HivePartitionDescriptor(org.apache.drill.exec.planner.sql.HivePartitionDescriptor)

Example 28 with TableScan

use of org.apache.beam.vendor.calcite.v1_28_0.org.apache.calcite.rel.core.TableScan in project drill by axbaretto.

the class ParquetPruneScanRule method getFilterOnProjectParquet.

public static final RelOptRule getFilterOnProjectParquet(OptimizerRulesContext optimizerRulesContext) {
    return new PruneScanRule(RelOptHelper.some(DrillFilterRel.class, RelOptHelper.some(DrillProjectRel.class, RelOptHelper.any(DrillScanRel.class))), "PruneScanRule:Filter_On_Project_Parquet", optimizerRulesContext) {

        @Override
        public PartitionDescriptor getPartitionDescriptor(PlannerSettings settings, TableScan scanRel) {
            return new ParquetPartitionDescriptor(settings, (DrillScanRel) scanRel);
        }

        @Override
        public boolean matches(RelOptRuleCall call) {
            final DrillScanRel scan = call.rel(2);
            GroupScan groupScan = scan.getGroupScan();
            // this rule is applicable only for parquet based partition pruning
            if (PrelUtil.getPlannerSettings(scan.getCluster().getPlanner()).isHepPartitionPruningEnabled()) {
                return groupScan instanceof ParquetGroupScan && groupScan.supportsPartitionFilterPushdown() && !scan.partitionFilterPushdown();
            } else {
                return groupScan instanceof ParquetGroupScan && groupScan.supportsPartitionFilterPushdown();
            }
        }

        @Override
        public void onMatch(RelOptRuleCall call) {
            final DrillFilterRel filterRel = call.rel(0);
            final DrillProjectRel projectRel = call.rel(1);
            final DrillScanRel scanRel = call.rel(2);
            doOnMatch(call, filterRel, projectRel, scanRel);
        }
    };
}
Also used : FileGroupScan(org.apache.drill.exec.physical.base.FileGroupScan) GroupScan(org.apache.drill.exec.physical.base.GroupScan) ParquetGroupScan(org.apache.drill.exec.store.parquet.ParquetGroupScan) TableScan(org.apache.calcite.rel.core.TableScan) DrillScanRel(org.apache.drill.exec.planner.logical.DrillScanRel) PlannerSettings(org.apache.drill.exec.planner.physical.PlannerSettings) DrillProjectRel(org.apache.drill.exec.planner.logical.DrillProjectRel) DrillFilterRel(org.apache.drill.exec.planner.logical.DrillFilterRel) ParquetGroupScan(org.apache.drill.exec.store.parquet.ParquetGroupScan) RelOptRuleCall(org.apache.calcite.plan.RelOptRuleCall) ParquetPartitionDescriptor(org.apache.drill.exec.planner.ParquetPartitionDescriptor)

Example 29 with TableScan

use of org.apache.beam.vendor.calcite.v1_28_0.org.apache.calcite.rel.core.TableScan in project drill by axbaretto.

the class DrillPushProjIntoScan method onMatch.

@Override
public void onMatch(RelOptRuleCall call) {
    final Project proj = call.rel(0);
    final TableScan scan = call.rel(1);
    try {
        ProjectPushInfo columnInfo = PrelUtil.getColumns(scan.getRowType(), proj.getProjects());
        // get DrillTable, either wrapped in RelOptTable, or DrillTranslatableTable.
        DrillTable table = scan.getTable().unwrap(DrillTable.class);
        if (table == null) {
            table = scan.getTable().unwrap(DrillTranslatableTable.class).getDrillTable();
        }
        if (// 
        columnInfo == null || columnInfo.isStarQuery() || !// 
        table.getGroupScan().canPushdownProjects(columnInfo.columns)) {
            return;
        }
        final DrillScanRel newScan = new DrillScanRel(scan.getCluster(), scan.getTraitSet().plus(DrillRel.DRILL_LOGICAL), scan.getTable(), columnInfo.createNewRowType(proj.getInput().getCluster().getTypeFactory()), columnInfo.columns);
        List<RexNode> newProjects = Lists.newArrayList();
        for (RexNode n : proj.getChildExps()) {
            newProjects.add(n.accept(columnInfo.getInputRewriter()));
        }
        final DrillProjectRel newProj = new DrillProjectRel(proj.getCluster(), proj.getTraitSet().plus(DrillRel.DRILL_LOGICAL), newScan, newProjects, proj.getRowType());
        if (ProjectRemoveRule.isTrivial(newProj)) {
            call.transformTo(newScan);
        } else {
            call.transformTo(newProj);
        }
    } catch (IOException e) {
        throw new DrillRuntimeException(e);
    }
}
Also used : Project(org.apache.calcite.rel.core.Project) LogicalProject(org.apache.calcite.rel.logical.LogicalProject) TableScan(org.apache.calcite.rel.core.TableScan) EnumerableTableScan(org.apache.calcite.adapter.enumerable.EnumerableTableScan) ProjectPushInfo(org.apache.drill.exec.planner.physical.PrelUtil.ProjectPushInfo) IOException(java.io.IOException) DrillRuntimeException(org.apache.drill.common.exceptions.DrillRuntimeException) RexNode(org.apache.calcite.rex.RexNode)

Example 30 with TableScan

use of org.apache.beam.vendor.calcite.v1_28_0.org.apache.calcite.rel.core.TableScan in project hive by apache.

the class HiveAugmentMaterializationRule method onMatch.

@Override
public void onMatch(RelOptRuleCall call) {
    final TableScan tableScan = call.rel(0);
    if (!visited.add(tableScan)) {
        // Already visited
        return;
    }
    final String tableQName = ((RelOptHiveTable) tableScan.getTable()).getHiveTableMD().getFullyQualifiedName();
    final ValidWriteIdList tableCurrentTxnList = currentTxnList.getTableValidWriteIdList(tableQName);
    final ValidWriteIdList tableMaterializationTxnList = materializationTxnList.getTableValidWriteIdList(tableQName);
    if (TxnIdUtils.checkEquivalentWriteIds(tableCurrentTxnList, tableMaterializationTxnList)) {
        // nothing to do
        return;
    }
    // ROW__ID: struct<transactionid:bigint,bucketid:int,rowid:bigint>
    int rowIDPos = tableScan.getTable().getRowType().getField(VirtualColumn.ROWID.getName(), false, false).getIndex();
    RexNode rowIDFieldAccess = rexBuilder.makeFieldAccess(rexBuilder.makeInputRef(tableScan.getTable().getRowType().getFieldList().get(rowIDPos).getType(), rowIDPos), 0);
    // Now we create the filter with the transactions information.
    // In particular, each table in the materialization will only have contents such that:
    // ROW_ID.writeid <= high_watermark and ROW_ID.writeid not in (open/invalid_ids)
    // Hence, we add that condition on top of the source table.
    // The rewriting will then have the possibility to create partial rewritings that read
    // the materialization and the source tables, and hence, produce an incremental
    // rebuild that is more efficient than the full rebuild.
    final RelBuilder relBuilder = call.builder();
    relBuilder.push(tableScan);
    List<RexNode> conds = new ArrayList<>();
    RelDataType bigIntType = relBuilder.getTypeFactory().createSqlType(SqlTypeName.BIGINT);
    final RexNode literalHighWatermark = rexBuilder.makeLiteral(tableMaterializationTxnList.getHighWatermark(), bigIntType, false);
    conds.add(rexBuilder.makeCall(SqlStdOperatorTable.LESS_THAN_OR_EQUAL, ImmutableList.of(rowIDFieldAccess, literalHighWatermark)));
    for (long invalidTxn : tableMaterializationTxnList.getInvalidWriteIds()) {
        final RexNode literalInvalidTxn = rexBuilder.makeLiteral(invalidTxn, bigIntType, false);
        conds.add(rexBuilder.makeCall(SqlStdOperatorTable.NOT_EQUALS, ImmutableList.of(rowIDFieldAccess, literalInvalidTxn)));
    }
    relBuilder.filter(conds);
    call.transformTo(relBuilder.build());
}
Also used : TableScan(org.apache.calcite.rel.core.TableScan) RelBuilder(org.apache.calcite.tools.RelBuilder) ValidWriteIdList(org.apache.hadoop.hive.common.ValidWriteIdList) ArrayList(java.util.ArrayList) RelDataType(org.apache.calcite.rel.type.RelDataType) RexNode(org.apache.calcite.rex.RexNode)

Aggregations

TableScan (org.apache.calcite.rel.core.TableScan)51 RelNode (org.apache.calcite.rel.RelNode)19 ArrayList (java.util.ArrayList)14 PlannerSettings (org.apache.drill.exec.planner.physical.PlannerSettings)13 Project (org.apache.calcite.rel.core.Project)12 DrillScanRel (org.apache.drill.exec.planner.logical.DrillScanRel)11 Filter (org.apache.calcite.rel.core.Filter)10 IOException (java.io.IOException)9 RexNode (org.apache.calcite.rex.RexNode)9 GroupScan (org.apache.drill.exec.physical.base.GroupScan)9 RelOptRuleCall (org.apache.calcite.plan.RelOptRuleCall)8 DrillFilterRel (org.apache.drill.exec.planner.logical.DrillFilterRel)8 Aggregate (org.apache.calcite.rel.core.Aggregate)7 LogicalProject (org.apache.calcite.rel.logical.LogicalProject)7 RexBuilder (org.apache.calcite.rex.RexBuilder)6 RelOptCluster (org.apache.calcite.plan.RelOptCluster)5 RelShuttleImpl (org.apache.calcite.rel.RelShuttleImpl)5 LogicalJoin (org.apache.calcite.rel.logical.LogicalJoin)5 RelDataType (org.apache.calcite.rel.type.RelDataType)5 SchemaPath (org.apache.drill.common.expression.SchemaPath)5