Examples with HashMapWrapper org.apache.hadoop.hive.ql.exec.persistence.HashMapWrapper used on opensource projects

Example 1 with HashMapWrapper

use of org.apache.hadoop.hive.ql.exec.persistence.HashMapWrapper in project hive by apache.

the class MapJoinTestConfig method createMapJoin.

public static CreateMapJoinResult createMapJoin(MapJoinTestDescription testDesc, MapJoinTestData testData, MapJoinDesc mapJoinDesc, boolean isVectorMapJoin, boolean isOriginalMapJoin, MapJoinTableContainer shareMapJoinTableContainer) throws SerDeException, IOException, HiveException {
    final Byte bigTablePos = 0;
    MapJoinTableContainerSerDe mapJoinTableContainerSerDe = MapJoinTestConfig.createMapJoinTableContainerSerDe(mapJoinDesc);
    MapJoinObjectSerDeContext valCtx = mapJoinTableContainerSerDe.getValueContext();
    MapJoinTableContainer mapJoinTableContainer = (isOriginalMapJoin ? new HashMapWrapper(testDesc.hiveConf, -1) : new MapJoinBytesTableContainer(testDesc.hiveConf, valCtx, testData.smallTableKeyHashMap.size(), 0));
    mapJoinTableContainer.setSerde(mapJoinTableContainerSerDe.getKeyContext(), mapJoinTableContainerSerDe.getValueContext());
    loadTableContainerData(testDesc, testData, mapJoinTableContainer);
    MapJoinOperator operator;
    if (!isVectorMapJoin) {
        operator = new MapJoinOperator(new CompilationOpContext());
        operator.setConf(mapJoinDesc);
    } else {
        VectorizationContext vContext = new VectorizationContext("test", testDesc.bigTableColumnNameList);
        /*
      // UNDONE: Unclear this belonds in the input VectorizationContext...
      // Create scratch columns to hold small table results.
      for (int i = 0; i < testDesc.smallTableValueTypeInfos.length; i++) {
        vContext.allocateScratchColumn(testDesc.smallTableValueTypeInfos[i]);
      }
      */
        // This is what the Vectorizer class does.
        VectorMapJoinDesc vectorMapJoinDesc = new VectorMapJoinDesc();
        byte posBigTable = (byte) mapJoinDesc.getPosBigTable();
        VectorExpression[] allBigTableKeyExpressions = vContext.getVectorExpressions(mapJoinDesc.getKeys().get(posBigTable));
        vectorMapJoinDesc.setAllBigTableKeyExpressions(allBigTableKeyExpressions);
        Map<Byte, List<ExprNodeDesc>> exprs = mapJoinDesc.getExprs();
        VectorExpression[] allBigTableValueExpressions = vContext.getVectorExpressions(exprs.get(posBigTable));
        vectorMapJoinDesc.setAllBigTableValueExpressions(allBigTableValueExpressions);
        List<ExprNodeDesc> bigTableFilters = mapJoinDesc.getFilters().get(bigTablePos);
        boolean isOuterAndFiltered = (!mapJoinDesc.isNoOuterJoin() && bigTableFilters.size() > 0);
        if (!isOuterAndFiltered) {
            operator = new VectorMapJoinOperator(new CompilationOpContext(), mapJoinDesc, vContext, vectorMapJoinDesc);
        } else {
            operator = new VectorMapJoinOuterFilteredOperator(new CompilationOpContext(), mapJoinDesc, vContext, vectorMapJoinDesc);
        }
    }
    HiveConf.setBoolVar(testDesc.hiveConf, HiveConf.ConfVars.HIVE_MAPJOIN_TESTING_NO_HASH_TABLE_LOAD, true);
    return new CreateMapJoinResult(operator, mapJoinTableContainer, mapJoinTableContainerSerDe);
}

Example 2 with HashMapWrapper

use of org.apache.hadoop.hive.ql.exec.persistence.HashMapWrapper in project hive by apache.

the class MapJoinTestConfig method createMapJoin.

public static MapJoinOperator createMapJoin(MapJoinTestDescription testDesc, Operator<? extends OperatorDesc> collectorOperator, MapJoinTestData testData, MapJoinDesc mapJoinDesc, boolean isVectorMapJoin, boolean isOriginalMapJoin) throws SerDeException, IOException, HiveException {
    final Byte bigTablePos = 0;
    MapJoinTableContainerSerDe mapJoinTableContainerSerDe = MapJoinTestConfig.createMapJoinTableContainerSerDe(mapJoinDesc);
    MapJoinObjectSerDeContext valCtx = mapJoinTableContainerSerDe.getValueContext();
    MapJoinTableContainer mapJoinTableContainer = (isOriginalMapJoin ? new HashMapWrapper(testDesc.hiveConf, -1) : new MapJoinBytesTableContainer(testDesc.hiveConf, valCtx, testData.smallTableKeyHashMap.size(), 0));
    mapJoinTableContainer.setSerde(mapJoinTableContainerSerDe.getKeyContext(), mapJoinTableContainerSerDe.getValueContext());
    loadTableContainerData(testDesc, testData, mapJoinTableContainer);
    MapJoinOperator operator;
    if (!isVectorMapJoin) {
        operator = new MapJoinOperator(new CompilationOpContext());
        operator.setConf(mapJoinDesc);
    } else {
        VectorizationContext vContext = new VectorizationContext("test", testDesc.bigTableColumnNamesList);
        // Create scratch columns to hold small table results.
        for (int i = 0; i < testDesc.smallTableValueTypeInfos.length; i++) {
            vContext.allocateScratchColumn(testDesc.smallTableValueTypeInfos[i]);
        }
        // This is what the Vectorizer class does.
        VectorMapJoinDesc vectorMapJoinDesc = new VectorMapJoinDesc();
        byte posBigTable = (byte) mapJoinDesc.getPosBigTable();
        VectorExpression[] allBigTableKeyExpressions = vContext.getVectorExpressions(mapJoinDesc.getKeys().get(posBigTable));
        vectorMapJoinDesc.setAllBigTableKeyExpressions(allBigTableKeyExpressions);
        Map<Byte, List<ExprNodeDesc>> exprs = mapJoinDesc.getExprs();
        VectorExpression[] allBigTableValueExpressions = vContext.getVectorExpressions(exprs.get(posBigTable));
        vectorMapJoinDesc.setAllBigTableValueExpressions(allBigTableValueExpressions);
        List<ExprNodeDesc> bigTableFilters = mapJoinDesc.getFilters().get(bigTablePos);
        boolean isOuterAndFiltered = (!mapJoinDesc.isNoOuterJoin() && bigTableFilters.size() > 0);
        if (!isOuterAndFiltered) {
            operator = new VectorMapJoinOperator(new CompilationOpContext(), mapJoinDesc, vContext, vectorMapJoinDesc);
        } else {
            operator = new VectorMapJoinOuterFilteredOperator(new CompilationOpContext(), mapJoinDesc, vContext, vectorMapJoinDesc);
        }
    }
    MapJoinTestConfig.connectOperators(testDesc, operator, collectorOperator);
    operator.setTestMapJoinTableContainer(1, mapJoinTableContainer, mapJoinTableContainerSerDe);
    return operator;
}

Example 3 with HashMapWrapper

use of org.apache.hadoop.hive.ql.exec.persistence.HashMapWrapper in project hive by apache.

the class HashTableSinkOperator method initializeOp.

@Override
@SuppressWarnings("unchecked")
protected void initializeOp(Configuration hconf) throws HiveException {
    super.initializeOp(hconf);
    boolean isSilent = HiveConf.getBoolVar(hconf, HiveConf.ConfVars.HIVESESSIONSILENT);
    console = new LogHelper(LOG, isSilent);
    memoryExhaustionChecker = MemoryExhaustionCheckerFactory.getChecker(console, hconf, conf);
    emptyRowContainer.addRow(emptyObjectArray);
    // for small tables only; so get the big table position first
    posBigTableAlias = conf.getPosBigTable();
    order = conf.getTagOrder();
    // initialize some variables, which used to be initialized in CommonJoinOperator
    this.hconf = hconf;
    filterMaps = conf.getFilterMap();
    int tagLen = conf.getTagLength();
    // process join keys
    joinKeys = new List[tagLen];
    JoinUtil.populateJoinKeyValue(joinKeys, conf.getKeys(), posBigTableAlias, hconf);
    joinKeysObjectInspectors = JoinUtil.getObjectInspectorsFromEvaluators(joinKeys, inputObjInspectors, posBigTableAlias, tagLen);
    // process join values
    joinValues = new List[tagLen];
    JoinUtil.populateJoinKeyValue(joinValues, conf.getExprs(), posBigTableAlias, hconf);
    joinValuesObjectInspectors = JoinUtil.getObjectInspectorsFromEvaluators(joinValues, inputObjInspectors, posBigTableAlias, tagLen);
    // process join filters
    joinFilters = new List[tagLen];
    JoinUtil.populateJoinKeyValue(joinFilters, conf.getFilters(), posBigTableAlias, hconf);
    joinFilterObjectInspectors = JoinUtil.getObjectInspectorsFromEvaluators(joinFilters, inputObjInspectors, posBigTableAlias, tagLen);
    if (!conf.isNoOuterJoin()) {
        for (Byte alias : order) {
            if (alias == posBigTableAlias || joinValues[alias] == null) {
                continue;
            }
            List<ObjectInspector> rcOIs = joinValuesObjectInspectors[alias];
            if (filterMaps != null && filterMaps[alias] != null) {
                // for each alias, add object inspector for filter tag as the last element
                rcOIs = new ArrayList<ObjectInspector>(rcOIs);
                rcOIs.add(PrimitiveObjectInspectorFactory.writableShortObjectInspector);
            }
        }
    }
    mapJoinTables = new MapJoinPersistableTableContainer[tagLen];
    mapJoinTableSerdes = new MapJoinTableContainerSerDe[tagLen];
    hashTableScale = HiveConf.getLongVar(hconf, HiveConf.ConfVars.HIVEHASHTABLESCALE);
    if (hashTableScale <= 0) {
        hashTableScale = 1;
    }
    try {
        TableDesc keyTableDesc = conf.getKeyTblDesc();
        AbstractSerDe keySerDe = (AbstractSerDe) ReflectionUtils.newInstance(keyTableDesc.getSerDeClass(), null);
        keySerDe.initialize(null, keyTableDesc.getProperties(), null);
        MapJoinObjectSerDeContext keyContext = new MapJoinObjectSerDeContext(keySerDe, false);
        for (Byte pos : order) {
            if (pos == posBigTableAlias) {
                continue;
            }
            mapJoinTables[pos] = new HashMapWrapper(hconf, -1);
            TableDesc valueTableDesc = conf.getValueTblFilteredDescs().get(pos);
            AbstractSerDe valueSerDe = (AbstractSerDe) ReflectionUtils.newInstance(valueTableDesc.getSerDeClass(), null);
            valueSerDe.initialize(null, valueTableDesc.getProperties(), null);
            mapJoinTableSerdes[pos] = new MapJoinTableContainerSerDe(keyContext, new MapJoinObjectSerDeContext(valueSerDe, hasFilter(pos)));
        }
    } catch (SerDeException e) {
        throw new HiveException(e);
    }
}

Example 4 with HashMapWrapper

use of org.apache.hadoop.hive.ql.exec.persistence.HashMapWrapper in project hive by apache.

the class HashTableLoader method load.

@Override
public void load(MapJoinTableContainer[] mapJoinTables, MapJoinTableContainerSerDe[] mapJoinTableSerdes) throws HiveException {
    Map<Integer, String> parentToInput = desc.getParentToInput();
    Map<Integer, Long> parentKeyCounts = desc.getParentKeyCounts();
    boolean isCrossProduct = false;
    List<ExprNodeDesc> joinExprs = desc.getKeys().values().iterator().next();
    if (joinExprs.size() == 0) {
        isCrossProduct = true;
    }
    boolean useOptimizedTables = HiveConf.getBoolVar(hconf, HiveConf.ConfVars.HIVEMAPJOINUSEOPTIMIZEDTABLE);
    boolean useHybridGraceHashJoin = desc.isHybridHashJoin();
    boolean isFirstKey = true;
    // Get the total available memory from memory manager
    long totalMapJoinMemory = desc.getMemoryNeeded();
    LOG.info("Memory manager allocates " + totalMapJoinMemory + " bytes for the loading hashtable.");
    if (totalMapJoinMemory <= 0) {
        totalMapJoinMemory = HiveConf.getLongVar(hconf, HiveConf.ConfVars.HIVECONVERTJOINNOCONDITIONALTASKTHRESHOLD);
    }
    long processMaxMemory = ManagementFactory.getMemoryMXBean().getHeapMemoryUsage().getMax();
    if (totalMapJoinMemory > processMaxMemory) {
        float hashtableMemoryUsage = HiveConf.getFloatVar(hconf, HiveConf.ConfVars.HIVEHASHTABLEFOLLOWBYGBYMAXMEMORYUSAGE);
        LOG.warn("totalMapJoinMemory value of " + totalMapJoinMemory + " is greater than the max memory size of " + processMaxMemory);
        // Don't want to attempt to grab more memory than we have available .. percentage is a bit arbitrary
        totalMapJoinMemory = (long) (processMaxMemory * hashtableMemoryUsage);
    }
    // Only applicable to n-way Hybrid Grace Hash Join
    HybridHashTableConf nwayConf = null;
    long totalSize = 0;
    // position of the biggest small table
    int biggest = 0;
    Map<Integer, Long> tableMemorySizes = null;
    if (useHybridGraceHashJoin && mapJoinTables.length > 2) {
        // Create a Conf for n-way HybridHashTableContainers
        nwayConf = new HybridHashTableConf();
        LOG.info("N-way join: " + (mapJoinTables.length - 1) + " small tables.");
        // Find the biggest small table; also calculate total data size of all small tables
        // the size of the biggest small table
        long maxSize = Long.MIN_VALUE;
        for (int pos = 0; pos < mapJoinTables.length; pos++) {
            if (pos == desc.getPosBigTable()) {
                continue;
            }
            long smallTableSize = desc.getParentDataSizes().get(pos);
            totalSize += smallTableSize;
            if (maxSize < smallTableSize) {
                maxSize = smallTableSize;
                biggest = pos;
            }
        }
        tableMemorySizes = divideHybridHashTableMemory(mapJoinTables, desc, totalSize, totalMapJoinMemory);
        // Using biggest small table, calculate number of partitions to create for each small table
        long memory = tableMemorySizes.get(biggest);
        int numPartitions = 0;
        try {
            numPartitions = HybridHashTableContainer.calcNumPartitions(memory, maxSize, HiveConf.getIntVar(hconf, HiveConf.ConfVars.HIVEHYBRIDGRACEHASHJOINMINNUMPARTITIONS), HiveConf.getIntVar(hconf, HiveConf.ConfVars.HIVEHYBRIDGRACEHASHJOINMINWBSIZE));
        } catch (IOException e) {
            throw new HiveException(e);
        }
        nwayConf.setNumberOfPartitions(numPartitions);
    }
    MemoryMonitorInfo memoryMonitorInfo = desc.getMemoryMonitorInfo();
    boolean doMemCheck = false;
    long effectiveThreshold = 0;
    if (memoryMonitorInfo != null) {
        effectiveThreshold = memoryMonitorInfo.getEffectiveThreshold(desc.getMaxMemoryAvailable());
        // Flip the flag at runtime in case if we are running outside of LLAP
        if (!LlapDaemonInfo.INSTANCE.isLlap()) {
            memoryMonitorInfo.setLlap(false);
        }
        if (memoryMonitorInfo.doMemoryMonitoring()) {
            doMemCheck = true;
            LOG.info("Memory monitoring for hash table loader enabled. {}", memoryMonitorInfo);
        }
    }
    if (!doMemCheck) {
        LOG.info("Not doing hash table memory monitoring. {}", memoryMonitorInfo);
    }
    for (int pos = 0; pos < mapJoinTables.length; pos++) {
        if (pos == desc.getPosBigTable()) {
            continue;
        }
        long numEntries = 0;
        String inputName = parentToInput.get(pos);
        LogicalInput input = tezContext.getInput(inputName);
        try {
            input.start();
            tezContext.getTezProcessorContext().waitForAnyInputReady(Collections.<Input>singletonList(input));
        } catch (Exception e) {
            throw new HiveException(e);
        }
        try {
            KeyValueReader kvReader = (KeyValueReader) input.getReader();
            MapJoinObjectSerDeContext keyCtx = mapJoinTableSerdes[pos].getKeyContext(), valCtx = mapJoinTableSerdes[pos].getValueContext();
            if (useOptimizedTables) {
                ObjectInspector keyOi = keyCtx.getSerDe().getObjectInspector();
                if (!MapJoinBytesTableContainer.isSupportedKey(keyOi)) {
                    if (isFirstKey) {
                        useOptimizedTables = false;
                        LOG.info(describeOi("Not using optimized hash table. " + "Only a subset of mapjoin keys is supported. Unsupported key: ", keyOi));
                    } else {
                        throw new HiveException(describeOi("Only a subset of mapjoin keys is supported. Unsupported key: ", keyOi));
                    }
                }
            }
            isFirstKey = false;
            Long keyCountObj = parentKeyCounts.get(pos);
            long estKeyCount = (keyCountObj == null) ? -1 : keyCountObj;
            long inputRecords = -1;
            try {
                // TODO : Need to use class instead of string.
                // https://issues.apache.org/jira/browse/HIVE-23981
                inputRecords = ((AbstractLogicalInput) input).getContext().getCounters().findCounter("org.apache.tez.common.counters.TaskCounter", "APPROXIMATE_INPUT_RECORDS").getValue();
            } catch (Exception e) {
                LOG.debug("Failed to get value for counter APPROXIMATE_INPUT_RECORDS", e);
            }
            long keyCount = Math.max(estKeyCount, inputRecords);
            long memory = 0;
            if (useHybridGraceHashJoin) {
                if (mapJoinTables.length > 2) {
                    memory = tableMemorySizes.get(pos);
                } else {
                    // binary join
                    memory = totalMapJoinMemory;
                }
            }
            MapJoinTableContainer tableContainer;
            if (useOptimizedTables) {
                if (!useHybridGraceHashJoin || isCrossProduct) {
                    tableContainer = new MapJoinBytesTableContainer(hconf, valCtx, keyCount, 0);
                } else {
                    tableContainer = new HybridHashTableContainer(hconf, keyCount, memory, desc.getParentDataSizes().get(pos), nwayConf);
                }
            } else {
                tableContainer = new HashMapWrapper(hconf, keyCount);
            }
            LOG.info("Loading hash table for input: {} cacheKey: {} tableContainer: {} smallTablePos: {} " + "estKeyCount : {} keyCount : {}", inputName, cacheKey, tableContainer.getClass().getSimpleName(), pos, estKeyCount, keyCount);
            tableContainer.setSerde(keyCtx, valCtx);
            long startTime = System.currentTimeMillis();
            while (kvReader.next()) {
                tableContainer.putRow((Writable) kvReader.getCurrentKey(), (Writable) kvReader.getCurrentValue());
                numEntries++;
                if (doMemCheck && (numEntries % memoryMonitorInfo.getMemoryCheckInterval() == 0)) {
                    final long estMemUsage = tableContainer.getEstimatedMemorySize();
                    if (estMemUsage > effectiveThreshold) {
                        String msg = "Hash table loading exceeded memory limits for input: " + inputName + " numEntries: " + numEntries + " estimatedMemoryUsage: " + estMemUsage + " effectiveThreshold: " + effectiveThreshold + " memoryMonitorInfo: " + memoryMonitorInfo;
                        LOG.error(msg);
                        throw new MapJoinMemoryExhaustionError(msg);
                    } else {
                        LOG.info("Checking hash table loader memory usage for input: {} numEntries: {} " + "estimatedMemoryUsage: {} effectiveThreshold: {}", inputName, numEntries, estMemUsage, effectiveThreshold);
                    }
                }
            }
            long delta = System.currentTimeMillis() - startTime;
            htLoadCounter.increment(delta);
            tableContainer.seal();
            mapJoinTables[pos] = tableContainer;
            if (doMemCheck) {
                LOG.info("Finished loading hash table for input: {} cacheKey: {} numEntries: {} " + "estimatedMemoryUsage: {} Load Time : {} ", inputName, cacheKey, numEntries, tableContainer.getEstimatedMemorySize(), delta);
            } else {
                LOG.info("Finished loading hash table for input: {} cacheKey: {} numEntries: {} Load Time : {} ", inputName, cacheKey, numEntries, delta);
            }
        } catch (Exception e) {
            throw new HiveException(e);
        }
    }
}