Search in sources :

Example 51 with Pair

use of org.apache.beam.vendor.calcite.v1_28_0.org.apache.calcite.util.Pair in project hive by apache.

the class RelOptHiveTable method generateKeys.

private Pair<List<ImmutableBitSet>, List<ImmutableBitSet>> generateKeys() {
    final PrimaryKeyInfo primaryKeyInfo = hiveTblMetadata.getPrimaryKeyInfo();
    final UniqueConstraint uniqueKeyInfo = hiveTblMetadata.getUniqueKeyInfo();
    ImmutableList.Builder<ImmutableBitSet> builder = ImmutableList.builder();
    ImmutableList.Builder<ImmutableBitSet> nonNullbuilder = ImmutableList.builder();
    // First PK
    if (primaryKeyInfo != null && !primaryKeyInfo.getColNames().isEmpty()) {
        ImmutableBitSet.Builder keys = ImmutableBitSet.builder();
        for (String pkColName : primaryKeyInfo.getColNames().values()) {
            int pkPos;
            for (pkPos = 0; pkPos < rowType.getFieldNames().size(); pkPos++) {
                String colName = rowType.getFieldNames().get(pkPos);
                if (pkColName.equals(colName)) {
                    break;
                }
            }
            if (pkPos == rowType.getFieldNames().size()) {
                LOG.error("Column for primary key definition " + pkColName + " not found");
            }
            keys.set(pkPos);
        }
        ImmutableBitSet key = keys.build();
        builder.add(key);
        nonNullbuilder.add(key);
    }
    // Then UKs
    if (uniqueKeyInfo != null && !uniqueKeyInfo.getUniqueConstraints().isEmpty()) {
        for (List<UniqueConstraintCol> ukCols : uniqueKeyInfo.getUniqueConstraints().values()) {
            ImmutableBitSet.Builder keys = ImmutableBitSet.builder();
            boolean isNonNullable = true;
            for (UniqueConstraintCol ukCol : ukCols) {
                int ukPos;
                for (ukPos = 0; ukPos < rowType.getFieldNames().size(); ukPos++) {
                    String colName = rowType.getFieldNames().get(ukPos);
                    if (ukCol.colName.equals(colName)) {
                        if (rowType.getFieldList().get(ukPos).getType().isNullable()) {
                            // they should all be nullable
                            isNonNullable = false;
                        }
                        break;
                    }
                }
                if (ukPos == rowType.getFieldNames().size()) {
                    LOG.error("Column for unique constraint definition " + ukCol.colName + " not found");
                }
                keys.set(ukPos);
            }
            ImmutableBitSet key = keys.build();
            builder.add(key);
            if (isNonNullable) {
                nonNullbuilder.add(key);
            }
        }
    }
    return new Pair<>(builder.build(), nonNullbuilder.build());
}
Also used : UniqueConstraintCol(org.apache.hadoop.hive.ql.metadata.UniqueConstraint.UniqueConstraintCol) PrimaryKeyInfo(org.apache.hadoop.hive.ql.metadata.PrimaryKeyInfo) ImmutableBitSet(org.apache.calcite.util.ImmutableBitSet) ImmutableList(com.google.common.collect.ImmutableList) UniqueConstraint(org.apache.hadoop.hive.ql.metadata.UniqueConstraint) UniqueConstraint(org.apache.hadoop.hive.ql.metadata.UniqueConstraint) RelReferentialConstraint(org.apache.calcite.rel.RelReferentialConstraint) IntPair(org.apache.calcite.util.mapping.IntPair) Pair(org.apache.calcite.util.Pair)

Example 52 with Pair

use of org.apache.beam.vendor.calcite.v1_28_0.org.apache.calcite.util.Pair in project hive by apache.

the class ParallelEdgeFixer method fixParallelEdges.

private void fixParallelEdges(OperatorGraph og) throws SemanticException {
    // Identify edge operators
    ListValuedMap<Pair<Cluster, Cluster>, Pair<Operator<?>, Operator<?>>> edgeOperators = new ArrayListValuedHashMap<>();
    for (Cluster c : og.getClusters()) {
        for (Operator<?> o : c.getMembers()) {
            for (Operator<? extends OperatorDesc> p : o.getParentOperators()) {
                Cluster parentCluster = og.clusterOf(p);
                if (parentCluster == c) {
                    continue;
                }
                edgeOperators.put(new Pair<>(parentCluster, c), new Pair<>(p, o));
            }
        }
    }
    // process all edges and fix parallel edges if there are any
    for (Pair<Cluster, Cluster> key : edgeOperators.keySet()) {
        List<Pair<Operator<?>, Operator<?>>> values = edgeOperators.get(key);
        if (values.size() <= 1) {
            continue;
        }
        // operator order must in stabile order - or we end up with falky plans causing flaky tests...
        values.sort(new OperatorPairComparator());
        // remove one optionally unsupported edge (it will be kept as is)
        removeOneEdge(values);
        Iterator<Pair<Operator<?>, Operator<?>>> it = values.iterator();
        while (it.hasNext()) {
            Pair<Operator<?>, Operator<?>> pair = it.next();
            fixParallelEdge(pair.left, pair.right);
        }
    }
}
Also used : ReduceSinkOperator(org.apache.hadoop.hive.ql.exec.ReduceSinkOperator) MapJoinOperator(org.apache.hadoop.hive.ql.exec.MapJoinOperator) TableScanOperator(org.apache.hadoop.hive.ql.exec.TableScanOperator) Operator(org.apache.hadoop.hive.ql.exec.Operator) Cluster(org.apache.hadoop.hive.ql.optimizer.graph.OperatorGraph.Cluster) ArrayListValuedHashMap(org.apache.commons.collections4.multimap.ArrayListValuedHashMap) Pair(org.apache.calcite.util.Pair)

Example 53 with Pair

use of org.apache.beam.vendor.calcite.v1_28_0.org.apache.calcite.util.Pair in project hive by apache.

the class HiveProjectVisitor method createColInfos.

private Pair<ArrayList<ColumnInfo>, Set<Integer>> createColInfos(List<RexNode> calciteExprs, List<ExprNodeDesc> hiveExprs, List<String> projNames, OpAttr inpOpAf) {
    if (hiveExprs.size() != projNames.size()) {
        throw new RuntimeException("Column expressions list doesn't match Column Names list");
    }
    RexNode rexN;
    ExprNodeDesc pe;
    ArrayList<ColumnInfo> colInfos = new ArrayList<ColumnInfo>();
    boolean vc;
    Set<Integer> newVColSet = new HashSet<Integer>();
    for (int i = 0; i < hiveExprs.size(); i++) {
        pe = hiveExprs.get(i);
        rexN = calciteExprs.get(i);
        vc = false;
        if (rexN instanceof RexInputRef) {
            if (inpOpAf.vcolsInCalcite.contains(((RexInputRef) rexN).getIndex())) {
                newVColSet.add(i);
                vc = true;
            }
        }
        colInfos.add(new ColumnInfo(projNames.get(i), pe.getTypeInfo(), inpOpAf.tabAlias, vc));
    }
    return new Pair<ArrayList<ColumnInfo>, Set<Integer>>(colInfos, newVColSet);
}
Also used : ArrayList(java.util.ArrayList) ColumnInfo(org.apache.hadoop.hive.ql.exec.ColumnInfo) RexInputRef(org.apache.calcite.rex.RexInputRef) ExprNodeDesc(org.apache.hadoop.hive.ql.plan.ExprNodeDesc) RexNode(org.apache.calcite.rex.RexNode) HashSet(java.util.HashSet) Pair(org.apache.calcite.util.Pair)

Example 54 with Pair

use of org.apache.beam.vendor.calcite.v1_28_0.org.apache.calcite.util.Pair in project hive by apache.

the class HiveRelMdRowCount method canHandleJoin.

/*
   * 1. Join condition must be an Equality Predicate.
   * 2. both sides must reference 1 column.
   * 3. If needed flip the columns.
   */
private static Pair<Integer, Integer> canHandleJoin(Join joinRel, List<RexNode> leftFilters, List<RexNode> rightFilters, List<RexNode> joinFilters) {
    /*
     * If after classifying filters there is more than 1 joining predicate, we
     * don't handle this. Return null.
     */
    if (joinFilters.size() != 1) {
        return null;
    }
    RexNode joinCond = joinFilters.get(0);
    int leftColIdx;
    int rightColIdx;
    if (!(joinCond instanceof RexCall)) {
        return null;
    }
    if (((RexCall) joinCond).getOperator() != SqlStdOperatorTable.EQUALS) {
        return null;
    }
    ImmutableBitSet leftCols = RelOptUtil.InputFinder.bits(((RexCall) joinCond).getOperands().get(0));
    ImmutableBitSet rightCols = RelOptUtil.InputFinder.bits(((RexCall) joinCond).getOperands().get(1));
    if (leftCols.cardinality() != 1 || rightCols.cardinality() != 1) {
        return null;
    }
    int nFieldsLeft = joinRel.getLeft().getRowType().getFieldList().size();
    int nFieldsRight = joinRel.getRight().getRowType().getFieldList().size();
    int nSysFields = joinRel.getSystemFieldList().size();
    ImmutableBitSet rightFieldsBitSet = ImmutableBitSet.range(nSysFields + nFieldsLeft, nSysFields + nFieldsLeft + nFieldsRight);
    /*
     * flip column references if join condition specified in reverse order to
     * join sources.
     */
    if (rightFieldsBitSet.contains(leftCols)) {
        ImmutableBitSet t = leftCols;
        leftCols = rightCols;
        rightCols = t;
    }
    leftColIdx = leftCols.nextSetBit(0) - nSysFields;
    rightColIdx = rightCols.nextSetBit(0) - (nSysFields + nFieldsLeft);
    return new Pair<Integer, Integer>(leftColIdx, rightColIdx);
}
Also used : RexCall(org.apache.calcite.rex.RexCall) ImmutableBitSet(org.apache.calcite.util.ImmutableBitSet) RexNode(org.apache.calcite.rex.RexNode) Pair(org.apache.calcite.util.Pair)

Example 55 with Pair

use of org.apache.beam.vendor.calcite.v1_28_0.org.apache.calcite.util.Pair in project hive by apache.

the class TestMaterializedViewsCache method testParallelism.

@Disabled("Testing parallelism only")
@Test
void testParallelism() {
    int ITERATIONS = 1000000;
    List<Pair<Table, HiveRelOptMaterialization>> testData = new ArrayList<>();
    for (int i = 0; i < 10; ++i) {
        Table table = new Table(new org.apache.hadoop.hive.metastore.api.Table());
        table.setDbName("default");
        table.setTableName("mat" + i);
        table.setViewOriginalText("select col0 from t" + i);
        HiveRelOptMaterialization materialization = createMaterialization(table);
        testData.add(new Pair<>(table, materialization));
    }
    for (int i = 0; i < 10; ++i) {
        Table table = new Table(new org.apache.hadoop.hive.metastore.api.Table());
        table.setDbName("db1");
        table.setTableName("mat" + i);
        table.setViewOriginalText("select col0 from t" + i);
        HiveRelOptMaterialization materialization = createMaterialization(table);
        testData.add(new Pair<>(table, materialization));
    }
    List<Callable<Void>> callableList = new ArrayList<>();
    callableList.add(() -> {
        for (Pair<Table, HiveRelOptMaterialization> entry : testData) {
            materializedViewsCache.refresh(entry.left, entry.left, entry.right);
        }
        return null;
    });
    callableList.add(() -> {
        for (int j = 0; j < ITERATIONS; ++j) {
            for (Pair<Table, HiveRelOptMaterialization> entry : testData) {
                materializedViewsCache.remove(entry.left);
                materializedViewsCache.putIfAbsent(entry.left, entry.right);
            }
        }
        return null;
    });
    for (Pair<Table, HiveRelOptMaterialization> entry : testData) {
        callableList.add(() -> {
            for (int j = 0; j < ITERATIONS; ++j) {
                materializedViewsCache.get(entry.left.getViewExpandedText());
            }
            return null;
        });
    }
    callableList.add(() -> {
        for (int j = 0; j < ITERATIONS; ++j) {
            List<HiveRelOptMaterialization> materializations = materializedViewsCache.values();
        }
        return null;
    });
    ExecutorService executor = Executors.newFixedThreadPool(12);
    try {
        executor.invokeAll(callableList);
    } catch (InterruptedException e) {
        e.printStackTrace();
    }
}
Also used : RelOptTable(org.apache.calcite.plan.RelOptTable) RelOptHiveTable(org.apache.hadoop.hive.ql.optimizer.calcite.RelOptHiveTable) ArrayList(java.util.ArrayList) Callable(java.util.concurrent.Callable) ExecutorService(java.util.concurrent.ExecutorService) Pair(org.apache.calcite.util.Pair) Test(org.junit.jupiter.api.Test) Disabled(org.junit.jupiter.api.Disabled)

Aggregations

Pair (org.apache.calcite.util.Pair)112 RexNode (org.apache.calcite.rex.RexNode)72 ArrayList (java.util.ArrayList)70 RelNode (org.apache.calcite.rel.RelNode)59 RelDataTypeField (org.apache.calcite.rel.type.RelDataTypeField)55 RexInputRef (org.apache.calcite.rex.RexInputRef)29 ImmutableBitSet (org.apache.calcite.util.ImmutableBitSet)29 HashMap (java.util.HashMap)26 RexBuilder (org.apache.calcite.rex.RexBuilder)23 Map (java.util.Map)21 AggregateCall (org.apache.calcite.rel.core.AggregateCall)20 List (java.util.List)19 RelDataType (org.apache.calcite.rel.type.RelDataType)19 ImmutableList (com.google.common.collect.ImmutableList)18 JoinRelType (org.apache.calcite.rel.core.JoinRelType)16 TreeMap (java.util.TreeMap)14 RelDataTypeFactory (org.apache.calcite.rel.type.RelDataTypeFactory)13 RelBuilder (org.apache.calcite.tools.RelBuilder)13 ImmutableMap (com.google.common.collect.ImmutableMap)12 ImmutableSortedMap (com.google.common.collect.ImmutableSortedMap)12