Examples with LoadFileDesc org.apache.hadoop.hive.ql.plan.LoadFileDesc used on opensource projects

Example 1 with LoadFileDesc

use of org.apache.hadoop.hive.ql.plan.LoadFileDesc in project hive by apache.

the class TaskCompiler method compile.

@SuppressWarnings({ "nls", "unchecked" })
public void compile(final ParseContext pCtx, final List<Task<? extends Serializable>> rootTasks, final HashSet<ReadEntity> inputs, final HashSet<WriteEntity> outputs) throws SemanticException {
    Context ctx = pCtx.getContext();
    GlobalLimitCtx globalLimitCtx = pCtx.getGlobalLimitCtx();
    List<Task<MoveWork>> mvTask = new ArrayList<Task<MoveWork>>();
    List<LoadTableDesc> loadTableWork = pCtx.getLoadTableWork();
    List<LoadFileDesc> loadFileWork = pCtx.getLoadFileWork();
    boolean isCStats = pCtx.getQueryProperties().isAnalyzeRewrite();
    int outerQueryLimit = pCtx.getQueryProperties().getOuterQueryLimit();
    if (pCtx.getFetchTask() != null) {
        if (pCtx.getFetchTask().getTblDesc() == null) {
            return;
        }
        pCtx.getFetchTask().getWork().setHiveServerQuery(SessionState.get().isHiveServerQuery());
        TableDesc resultTab = pCtx.getFetchTask().getTblDesc();
        // then either the ThriftFormatter or the DefaultFetchFormatter should be used.
        if (!resultTab.getSerdeClassName().equalsIgnoreCase(ThriftJDBCBinarySerDe.class.getName())) {
            if (SessionState.get().isHiveServerQuery()) {
                conf.set(SerDeUtils.LIST_SINK_OUTPUT_FORMATTER, ThriftFormatter.class.getName());
            } else {
                String formatterName = conf.get(SerDeUtils.LIST_SINK_OUTPUT_FORMATTER);
                if (formatterName == null || formatterName.isEmpty()) {
                    conf.set(SerDeUtils.LIST_SINK_OUTPUT_FORMATTER, DefaultFetchFormatter.class.getName());
                }
            }
        }
        return;
    }
    optimizeOperatorPlan(pCtx, inputs, outputs);
    /*
     * In case of a select, use a fetch task instead of a move task.
     * If the select is from analyze table column rewrite, don't create a fetch task. Instead create
     * a column stats task later.
     */
    if (pCtx.getQueryProperties().isQuery() && !isCStats) {
        if ((!loadTableWork.isEmpty()) || (loadFileWork.size() != 1)) {
            throw new SemanticException(ErrorMsg.INVALID_LOAD_TABLE_FILE_WORK.getMsg());
        }
        LoadFileDesc loadFileDesc = loadFileWork.get(0);
        String cols = loadFileDesc.getColumns();
        String colTypes = loadFileDesc.getColumnTypes();
        String resFileFormat;
        TableDesc resultTab = pCtx.getFetchTableDesc();
        if (resultTab == null) {
            resFileFormat = HiveConf.getVar(conf, HiveConf.ConfVars.HIVEQUERYRESULTFILEFORMAT);
            if (SessionState.get().getIsUsingThriftJDBCBinarySerDe() && (resFileFormat.equalsIgnoreCase("SequenceFile"))) {
                resultTab = PlanUtils.getDefaultQueryOutputTableDesc(cols, colTypes, resFileFormat, ThriftJDBCBinarySerDe.class);
                // Set the fetch formatter to be a no-op for the ListSinkOperator, since we'll
                // read formatted thrift objects from the output SequenceFile written by Tasks.
                conf.set(SerDeUtils.LIST_SINK_OUTPUT_FORMATTER, NoOpFetchFormatter.class.getName());
            } else {
                resultTab = PlanUtils.getDefaultQueryOutputTableDesc(cols, colTypes, resFileFormat, LazySimpleSerDe.class);
            }
        } else {
            if (resultTab.getProperties().getProperty(serdeConstants.SERIALIZATION_LIB).equalsIgnoreCase(ThriftJDBCBinarySerDe.class.getName())) {
                // Set the fetch formatter to be a no-op for the ListSinkOperator, since we'll
                // read formatted thrift objects from the output SequenceFile written by Tasks.
                conf.set(SerDeUtils.LIST_SINK_OUTPUT_FORMATTER, NoOpFetchFormatter.class.getName());
            }
        }
        FetchWork fetch = new FetchWork(loadFileDesc.getSourcePath(), resultTab, outerQueryLimit);
        boolean isHiveServerQuery = SessionState.get().isHiveServerQuery();
        fetch.setHiveServerQuery(isHiveServerQuery);
        fetch.setSource(pCtx.getFetchSource());
        fetch.setSink(pCtx.getFetchSink());
        if (isHiveServerQuery && null != resultTab && resultTab.getSerdeClassName().equalsIgnoreCase(ThriftJDBCBinarySerDe.class.getName()) && HiveConf.getBoolVar(conf, HiveConf.ConfVars.HIVE_SERVER2_THRIFT_RESULTSET_SERIALIZE_IN_TASKS)) {
            fetch.setIsUsingThriftJDBCBinarySerDe(true);
        } else {
            fetch.setIsUsingThriftJDBCBinarySerDe(false);
        }
        pCtx.setFetchTask((FetchTask) TaskFactory.get(fetch, conf));
        // For the FetchTask, the limit optimization requires we fetch all the rows
        // in memory and count how many rows we get. It's not practical if the
        // limit factor is too big
        int fetchLimit = HiveConf.getIntVar(conf, HiveConf.ConfVars.HIVELIMITOPTMAXFETCH);
        if (globalLimitCtx.isEnable() && globalLimitCtx.getGlobalLimit() > fetchLimit) {
            LOG.info("For FetchTask, LIMIT " + globalLimitCtx.getGlobalLimit() + " > " + fetchLimit + ". Doesn't qualify limit optimization.");
            globalLimitCtx.disableOpt();
        }
        if (outerQueryLimit == 0) {
            // Believe it or not, some tools do generate queries with limit 0 and than expect
            // query to run quickly. Lets meet their requirement.
            LOG.info("Limit 0. No query execution needed.");
            return;
        }
    } else if (!isCStats) {
        for (LoadTableDesc ltd : loadTableWork) {
            Task<MoveWork> tsk = TaskFactory.get(new MoveWork(null, null, ltd, null, false), conf);
            mvTask.add(tsk);
            // Check to see if we are stale'ing any indexes and auto-update them if we want
            if (HiveConf.getBoolVar(conf, HiveConf.ConfVars.HIVEINDEXAUTOUPDATE)) {
                IndexUpdater indexUpdater = new IndexUpdater(loadTableWork, inputs, conf);
                try {
                    List<Task<? extends Serializable>> indexUpdateTasks = indexUpdater.generateUpdateTasks();
                    for (Task<? extends Serializable> updateTask : indexUpdateTasks) {
                        tsk.addDependentTask(updateTask);
                    }
                } catch (HiveException e) {
                    console.printInfo("WARNING: could not auto-update stale indexes, which are not in sync");
                }
            }
        }
        boolean oneLoadFile = true;
        for (LoadFileDesc lfd : loadFileWork) {
            if (pCtx.getQueryProperties().isCTAS() || pCtx.getQueryProperties().isMaterializedView()) {
                // should not have more than 1 load file for
                assert (oneLoadFile);
                // CTAS
                // make the movetask's destination directory the table's destination.
                Path location;
                String loc = pCtx.getQueryProperties().isCTAS() ? pCtx.getCreateTable().getLocation() : pCtx.getCreateViewDesc().getLocation();
                if (loc == null) {
                    // get the default location
                    Path targetPath;
                    try {
                        String protoName = null;
                        if (pCtx.getQueryProperties().isCTAS()) {
                            protoName = pCtx.getCreateTable().getTableName();
                        } else if (pCtx.getQueryProperties().isMaterializedView()) {
                            protoName = pCtx.getCreateViewDesc().getViewName();
                        }
                        String[] names = Utilities.getDbTableName(protoName);
                        if (!db.databaseExists(names[0])) {
                            throw new SemanticException("ERROR: The database " + names[0] + " does not exist.");
                        }
                        Warehouse wh = new Warehouse(conf);
                        targetPath = wh.getTablePath(db.getDatabase(names[0]), names[1]);
                    } catch (HiveException e) {
                        throw new SemanticException(e);
                    } catch (MetaException e) {
                        throw new SemanticException(e);
                    }
                    location = targetPath;
                } else {
                    location = new Path(loc);
                }
                lfd.setTargetDir(location);
                oneLoadFile = false;
            }
            mvTask.add(TaskFactory.get(new MoveWork(null, null, null, lfd, false), conf));
        }
    }
    generateTaskTree(rootTasks, pCtx, mvTask, inputs, outputs);
    // For each task, set the key descriptor for the reducer
    for (Task<? extends Serializable> rootTask : rootTasks) {
        GenMapRedUtils.setKeyAndValueDescForTaskTree(rootTask);
    }
    // to be used, please do so
    for (Task<? extends Serializable> rootTask : rootTasks) {
        setInputFormat(rootTask);
    }
    optimizeTaskPlan(rootTasks, pCtx, ctx);
    /*
     * If the query was the result of analyze table column compute statistics rewrite, create
     * a column stats task instead of a fetch task to persist stats to the metastore.
     */
    if (isCStats || !pCtx.getColumnStatsAutoGatherContexts().isEmpty()) {
        Set<Task<? extends Serializable>> leafTasks = new LinkedHashSet<Task<? extends Serializable>>();
        getLeafTasks(rootTasks, leafTasks);
        if (isCStats) {
            genColumnStatsTask(pCtx.getAnalyzeRewrite(), loadFileWork, leafTasks, outerQueryLimit, 0);
        } else {
            for (ColumnStatsAutoGatherContext columnStatsAutoGatherContext : pCtx.getColumnStatsAutoGatherContexts()) {
                if (!columnStatsAutoGatherContext.isInsertInto()) {
                    genColumnStatsTask(columnStatsAutoGatherContext.getAnalyzeRewrite(), columnStatsAutoGatherContext.getLoadFileWork(), leafTasks, outerQueryLimit, 0);
                } else {
                    int numBitVector;
                    try {
                        numBitVector = HiveStatsUtils.getNumBitVectorsForNDVEstimation(conf);
                    } catch (Exception e) {
                        throw new SemanticException(e.getMessage());
                    }
                    genColumnStatsTask(columnStatsAutoGatherContext.getAnalyzeRewrite(), columnStatsAutoGatherContext.getLoadFileWork(), leafTasks, outerQueryLimit, numBitVector);
                }
            }
        }
    }
    decideExecMode(rootTasks, ctx, globalLimitCtx);
    if (pCtx.getQueryProperties().isCTAS() && !pCtx.getCreateTable().isMaterialization()) {
        // generate a DDL task and make it a dependent task of the leaf
        CreateTableDesc crtTblDesc = pCtx.getCreateTable();
        crtTblDesc.validate(conf);
        Task<? extends Serializable> crtTblTask = TaskFactory.get(new DDLWork(inputs, outputs, crtTblDesc), conf);
        patchUpAfterCTASorMaterializedView(rootTasks, outputs, crtTblTask);
    } else if (pCtx.getQueryProperties().isMaterializedView()) {
        // generate a DDL task and make it a dependent task of the leaf
        CreateViewDesc viewDesc = pCtx.getCreateViewDesc();
        Task<? extends Serializable> crtViewTask = TaskFactory.get(new DDLWork(inputs, outputs, viewDesc), conf);
        patchUpAfterCTASorMaterializedView(rootTasks, outputs, crtViewTask);
    }
    if (globalLimitCtx.isEnable() && pCtx.getFetchTask() != null) {
        LOG.info("set least row check for FetchTask: " + globalLimitCtx.getGlobalLimit());
        pCtx.getFetchTask().getWork().setLeastNumRows(globalLimitCtx.getGlobalLimit());
    }
    if (globalLimitCtx.isEnable() && globalLimitCtx.getLastReduceLimitDesc() != null) {
        LOG.info("set least row check for LimitDesc: " + globalLimitCtx.getGlobalLimit());
        globalLimitCtx.getLastReduceLimitDesc().setLeastRows(globalLimitCtx.getGlobalLimit());
        List<ExecDriver> mrTasks = Utilities.getMRTasks(rootTasks);
        for (ExecDriver tsk : mrTasks) {
            tsk.setRetryCmdWhenFail(true);
        }
        List<SparkTask> sparkTasks = Utilities.getSparkTasks(rootTasks);
        for (SparkTask sparkTask : sparkTasks) {
            sparkTask.setRetryCmdWhenFail(true);
        }
    }
    Interner<TableDesc> interner = Interners.newStrongInterner();
    for (Task<? extends Serializable> rootTask : rootTasks) {
        GenMapRedUtils.internTableDesc(rootTask, interner);
        GenMapRedUtils.deriveFinalExplainAttributes(rootTask, pCtx.getConf());
    }
}

Example 2 with LoadFileDesc

use of org.apache.hadoop.hive.ql.plan.LoadFileDesc in project hive by apache.

the class MoveTask method isLocal.

/*
   * Does the move task involve moving to a local file system
   */
public boolean isLocal() {
    LoadTableDesc tbd = work.getLoadTableWork();
    if (tbd != null) {
        return false;
    }
    LoadFileDesc lfd = work.getLoadFileWork();
    if (lfd != null) {
        if (lfd.getIsDfsDir()) {
            return false;
        } else {
            return true;
        }
    }
    return false;
}

Example 3 with LoadFileDesc

use of org.apache.hadoop.hive.ql.plan.LoadFileDesc in project hive by apache.

the class GenMapRedUtils method createMRWorkForMergingFiles.

/**
   * @param fsInput The FileSink operator.
   * @param ctx The MR processing context.
   * @param finalName the final destination path the merge job should output.
   * @param dependencyTask
   * @param mvTasks
   * @param conf
   * @param currTask
   * @throws SemanticException

   * create a Map-only merge job using CombineHiveInputFormat for all partitions with
   * following operators:
   *          MR job J0:
   *          ...
   *          |
   *          v
   *          FileSinkOperator_1 (fsInput)
   *          |
   *          v
   *          Merge job J1:
   *          |
   *          v
   *          TableScan (using CombineHiveInputFormat) (tsMerge)
   *          |
   *          v
   *          FileSinkOperator (fsMerge)
   *
   *          Here the pathToPartitionInfo & pathToAlias will remain the same, which means the paths
   *          do
   *          not contain the dynamic partitions (their parent). So after the dynamic partitions are
   *          created (after the first job finished before the moveTask or ConditionalTask start),
   *          we need to change the pathToPartitionInfo & pathToAlias to include the dynamic
   *          partition
   *          directories.
   *
   */
public static void createMRWorkForMergingFiles(FileSinkOperator fsInput, Path finalName, DependencyCollectionTask dependencyTask, List<Task<MoveWork>> mvTasks, HiveConf conf, Task<? extends Serializable> currTask) throws SemanticException {
    //
    // 1. create the operator tree
    //
    FileSinkDesc fsInputDesc = fsInput.getConf();
    // Create a TableScan operator
    RowSchema inputRS = fsInput.getSchema();
    TableScanOperator tsMerge = GenMapRedUtils.createTemporaryTableScanOperator(fsInput.getCompilationOpContext(), inputRS);
    // Create a FileSink operator
    TableDesc ts = (TableDesc) fsInputDesc.getTableInfo().clone();
    FileSinkDesc fsOutputDesc = new FileSinkDesc(finalName, ts, conf.getBoolVar(ConfVars.COMPRESSRESULT));
    FileSinkOperator fsOutput = (FileSinkOperator) OperatorFactory.getAndMakeChild(fsOutputDesc, inputRS, tsMerge);
    // If the input FileSinkOperator is a dynamic partition enabled, the tsMerge input schema
    // needs to include the partition column, and the fsOutput should have
    // a DynamicPartitionCtx to indicate that it needs to dynamically partitioned.
    DynamicPartitionCtx dpCtx = fsInputDesc.getDynPartCtx();
    if (dpCtx != null && dpCtx.getNumDPCols() > 0) {
        // adding DP ColumnInfo to the RowSchema signature
        ArrayList<ColumnInfo> signature = inputRS.getSignature();
        String tblAlias = fsInputDesc.getTableInfo().getTableName();
        for (String dpCol : dpCtx.getDPColNames()) {
            ColumnInfo colInfo = new ColumnInfo(dpCol, // all partition column type should be string
            TypeInfoFactory.stringTypeInfo, tblAlias, // partition column is virtual column
            true);
            signature.add(colInfo);
        }
        inputRS.setSignature(signature);
        // create another DynamicPartitionCtx, which has a different input-to-DP column mapping
        DynamicPartitionCtx dpCtx2 = new DynamicPartitionCtx(dpCtx);
        fsOutputDesc.setDynPartCtx(dpCtx2);
        // update the FileSinkOperator to include partition columns
        usePartitionColumns(fsInputDesc.getTableInfo().getProperties(), dpCtx.getDPColNames());
    } else {
        // non-partitioned table
        fsInputDesc.getTableInfo().getProperties().remove(org.apache.hadoop.hive.metastore.api.hive_metastoreConstants.META_TABLE_PARTITION_COLUMNS);
    }
    //
    // 2. Constructing a conditional task consisting of a move task and a map reduce task
    //
    MoveWork dummyMv = new MoveWork(null, null, null, new LoadFileDesc(fsInputDesc.getFinalDirName(), finalName, true, null, null), false);
    MapWork cplan;
    Serializable work;
    if ((conf.getBoolVar(ConfVars.HIVEMERGERCFILEBLOCKLEVEL) && fsInputDesc.getTableInfo().getInputFileFormatClass().equals(RCFileInputFormat.class)) || (conf.getBoolVar(ConfVars.HIVEMERGEORCFILESTRIPELEVEL) && fsInputDesc.getTableInfo().getInputFileFormatClass().equals(OrcInputFormat.class))) {
        cplan = GenMapRedUtils.createMergeTask(fsInputDesc, finalName, dpCtx != null && dpCtx.getNumDPCols() > 0, fsInput.getCompilationOpContext());
        if (conf.getVar(ConfVars.HIVE_EXECUTION_ENGINE).equals("tez")) {
            work = new TezWork(conf.getVar(HiveConf.ConfVars.HIVEQUERYID), conf);
            cplan.setName("File Merge");
            ((TezWork) work).add(cplan);
        } else if (conf.getVar(ConfVars.HIVE_EXECUTION_ENGINE).equals("spark")) {
            work = new SparkWork(conf.getVar(HiveConf.ConfVars.HIVEQUERYID));
            cplan.setName("Spark Merge File Work");
            ((SparkWork) work).add(cplan);
        } else {
            work = cplan;
        }
    } else {
        cplan = createMRWorkForMergingFiles(conf, tsMerge, fsInputDesc);
        if (conf.getVar(ConfVars.HIVE_EXECUTION_ENGINE).equals("tez")) {
            work = new TezWork(conf.getVar(HiveConf.ConfVars.HIVEQUERYID), conf);
            cplan.setName("File Merge");
            ((TezWork) work).add(cplan);
        } else if (conf.getVar(ConfVars.HIVE_EXECUTION_ENGINE).equals("spark")) {
            work = new SparkWork(conf.getVar(HiveConf.ConfVars.HIVEQUERYID));
            cplan.setName("Spark Merge File Work");
            ((SparkWork) work).add(cplan);
        } else {
            work = new MapredWork();
            ((MapredWork) work).setMapWork(cplan);
        }
    }
    // use CombineHiveInputFormat for map-only merging
    cplan.setInputformat("org.apache.hadoop.hive.ql.io.CombineHiveInputFormat");
    // NOTE: we should gather stats in MR1 rather than MR2 at merge job since we don't
    // know if merge MR2 will be triggered at execution time
    Task<MoveWork> mvTask = GenMapRedUtils.findMoveTask(mvTasks, fsOutput);
    ConditionalTask cndTsk = GenMapRedUtils.createCondTask(conf, currTask, dummyMv, work, fsInputDesc.getFinalDirName(), finalName, mvTask, dependencyTask);
    // keep the dynamic partition context in conditional task resolver context
    ConditionalResolverMergeFilesCtx mrCtx = (ConditionalResolverMergeFilesCtx) cndTsk.getResolverCtx();
    mrCtx.setDPCtx(fsInputDesc.getDynPartCtx());
    mrCtx.setLbCtx(fsInputDesc.getLbCtx());
}

Example 4 with LoadFileDesc

use of org.apache.hadoop.hive.ql.plan.LoadFileDesc in project hive by apache.

the class SemanticAnalyzer method genFileSinkPlan.

@SuppressWarnings("nls")
protected Operator genFileSinkPlan(String dest, QB qb, Operator input) throws SemanticException {
    RowResolver inputRR = opParseCtx.get(input).getRowResolver();
    QBMetaData qbm = qb.getMetaData();
    Integer dest_type = qbm.getDestTypeForAlias(dest);
    // destination table if any
    Table dest_tab = null;
    // should the destination table be written to using ACID
    boolean destTableIsAcid = false;
    boolean destTableIsTemporary = false;
    boolean destTableIsMaterialization = false;
    // destination partition if any
    Partition dest_part = null;
    // the intermediate destination directory
    Path queryTmpdir = null;
    // the final destination directory
    Path dest_path = null;
    TableDesc table_desc = null;
    int currentTableId = 0;
    boolean isLocal = false;
    SortBucketRSCtx rsCtx = new SortBucketRSCtx();
    DynamicPartitionCtx dpCtx = null;
    LoadTableDesc ltd = null;
    ListBucketingCtx lbCtx = null;
    Map<String, String> partSpec = null;
    switch(dest_type.intValue()) {
        case QBMetaData.DEST_TABLE:
            {
                dest_tab = qbm.getDestTableForAlias(dest);
                destTableIsAcid = AcidUtils.isAcidTable(dest_tab);
                destTableIsTemporary = dest_tab.isTemporary();
                // Is the user trying to insert into a external tables
                if ((!conf.getBoolVar(HiveConf.ConfVars.HIVE_INSERT_INTO_EXTERNAL_TABLES)) && (dest_tab.getTableType().equals(TableType.EXTERNAL_TABLE))) {
                    throw new SemanticException(ErrorMsg.INSERT_EXTERNAL_TABLE.getMsg(dest_tab.getTableName()));
                }
                partSpec = qbm.getPartSpecForAlias(dest);
                dest_path = dest_tab.getPath();
                // verify that our destination is empty before proceeding
                if (dest_tab.isImmutable() && qb.getParseInfo().isInsertIntoTable(dest_tab.getDbName(), dest_tab.getTableName())) {
                    try {
                        FileSystem fs = dest_path.getFileSystem(conf);
                        if (!MetaStoreUtils.isDirEmpty(fs, dest_path)) {
                            LOG.warn("Attempted write into an immutable table : " + dest_tab.getTableName() + " : " + dest_path);
                            throw new SemanticException(ErrorMsg.INSERT_INTO_IMMUTABLE_TABLE.getMsg(dest_tab.getTableName()));
                        }
                    } catch (IOException ioe) {
                        LOG.warn("Error while trying to determine if immutable table has any data : " + dest_tab.getTableName() + " : " + dest_path);
                        throw new SemanticException(ErrorMsg.INSERT_INTO_IMMUTABLE_TABLE.getMsg(ioe.getMessage()));
                    }
                }
                // check for partition
                List<FieldSchema> parts = dest_tab.getPartitionKeys();
                if (parts != null && parts.size() > 0) {
                    // table is partitioned
                    if (partSpec == null || partSpec.size() == 0) {
                        // user did NOT specify partition
                        throw new SemanticException(generateErrorMessage(qb.getParseInfo().getDestForClause(dest), ErrorMsg.NEED_PARTITION_ERROR.getMsg()));
                    }
                    dpCtx = qbm.getDPCtx(dest);
                    if (dpCtx == null) {
                        dest_tab.validatePartColumnNames(partSpec, false);
                        dpCtx = new DynamicPartitionCtx(dest_tab, partSpec, conf.getVar(HiveConf.ConfVars.DEFAULTPARTITIONNAME), conf.getIntVar(HiveConf.ConfVars.DYNAMICPARTITIONMAXPARTSPERNODE));
                        qbm.setDPCtx(dest, dpCtx);
                    }
                    if (!HiveConf.getBoolVar(conf, HiveConf.ConfVars.DYNAMICPARTITIONING)) {
                        // allow DP
                        throw new SemanticException(generateErrorMessage(qb.getParseInfo().getDestForClause(dest), ErrorMsg.DYNAMIC_PARTITION_DISABLED.getMsg()));
                    }
                    if (dpCtx.getSPPath() != null) {
                        dest_path = new Path(dest_tab.getPath(), dpCtx.getSPPath());
                    }
                    if ((dest_tab.getNumBuckets() > 0)) {
                        dpCtx.setNumBuckets(dest_tab.getNumBuckets());
                    }
                }
                boolean isNonNativeTable = dest_tab.isNonNative();
                if (isNonNativeTable) {
                    queryTmpdir = dest_path;
                } else {
                    queryTmpdir = ctx.getTempDirForPath(dest_path, true);
                }
                if (dpCtx != null) {
                    // set the root of the temporary path where dynamic partition columns will populate
                    dpCtx.setRootPath(queryTmpdir);
                }
                // this table_desc does not contain the partitioning columns
                table_desc = Utilities.getTableDesc(dest_tab);
                // Add sorting/bucketing if needed
                input = genBucketingSortingDest(dest, input, qb, table_desc, dest_tab, rsCtx);
                idToTableNameMap.put(String.valueOf(destTableId), dest_tab.getTableName());
                currentTableId = destTableId;
                destTableId++;
                lbCtx = constructListBucketingCtx(dest_tab.getSkewedColNames(), dest_tab.getSkewedColValues(), dest_tab.getSkewedColValueLocationMaps(), dest_tab.isStoredAsSubDirectories(), conf);
                // NOTE: specify Dynamic partitions in dest_tab for WriteEntity
                if (!isNonNativeTable) {
                    AcidUtils.Operation acidOp = AcidUtils.Operation.NOT_ACID;
                    if (destTableIsAcid) {
                        acidOp = getAcidType(table_desc.getOutputFileFormatClass(), dest);
                        checkAcidConstraints(qb, table_desc, dest_tab);
                    }
                    ltd = new LoadTableDesc(queryTmpdir, table_desc, dpCtx, acidOp);
                    ltd.setReplace(!qb.getParseInfo().isInsertIntoTable(dest_tab.getDbName(), dest_tab.getTableName()));
                    ltd.setLbCtx(lbCtx);
                    loadTableWork.add(ltd);
                } else {
                    // This is a non-native table.
                    // We need to set stats as inaccurate.
                    setStatsForNonNativeTable(dest_tab);
                    // true if it is insert overwrite.
                    boolean overwrite = !qb.getParseInfo().isInsertIntoTable(String.format("%s.%s", dest_tab.getDbName(), dest_tab.getTableName()));
                    createInsertDesc(dest_tab, overwrite);
                }
                WriteEntity output = null;
                // list of dynamically created partitions are known.
                if ((dpCtx == null || dpCtx.getNumDPCols() == 0)) {
                    output = new WriteEntity(dest_tab, determineWriteType(ltd, isNonNativeTable, dest));
                    if (!outputs.add(output)) {
                        throw new SemanticException(ErrorMsg.OUTPUT_SPECIFIED_MULTIPLE_TIMES.getMsg(dest_tab.getTableName()));
                    }
                }
                if ((dpCtx != null) && (dpCtx.getNumDPCols() >= 0)) {
                    // No static partition specified
                    if (dpCtx.getNumSPCols() == 0) {
                        output = new WriteEntity(dest_tab, determineWriteType(ltd, isNonNativeTable, dest), false);
                        outputs.add(output);
                        output.setDynamicPartitionWrite(true);
                    } else // part of the partition specified
                    // Create a DummyPartition in this case. Since, the metastore does not store partial
                    // partitions currently, we need to store dummy partitions
                    {
                        try {
                            String ppath = dpCtx.getSPPath();
                            ppath = ppath.substring(0, ppath.length() - 1);
                            DummyPartition p = new DummyPartition(dest_tab, dest_tab.getDbName() + "@" + dest_tab.getTableName() + "@" + ppath, partSpec);
                            output = new WriteEntity(p, getWriteType(dest), false);
                            output.setDynamicPartitionWrite(true);
                            outputs.add(output);
                        } catch (HiveException e) {
                            throw new SemanticException(e.getMessage(), e);
                        }
                    }
                }
                ctx.getLoadTableOutputMap().put(ltd, output);
                break;
            }
        case QBMetaData.DEST_PARTITION:
            {
                dest_part = qbm.getDestPartitionForAlias(dest);
                dest_tab = dest_part.getTable();
                destTableIsAcid = AcidUtils.isAcidTable(dest_tab);
                if ((!conf.getBoolVar(HiveConf.ConfVars.HIVE_INSERT_INTO_EXTERNAL_TABLES)) && dest_tab.getTableType().equals(TableType.EXTERNAL_TABLE)) {
                    throw new SemanticException(ErrorMsg.INSERT_EXTERNAL_TABLE.getMsg(dest_tab.getTableName()));
                }
                Path tabPath = dest_tab.getPath();
                Path partPath = dest_part.getDataLocation();
                // verify that our destination is empty before proceeding
                if (dest_tab.isImmutable() && qb.getParseInfo().isInsertIntoTable(dest_tab.getDbName(), dest_tab.getTableName())) {
                    try {
                        FileSystem fs = partPath.getFileSystem(conf);
                        if (!MetaStoreUtils.isDirEmpty(fs, partPath)) {
                            LOG.warn("Attempted write into an immutable table partition : " + dest_tab.getTableName() + " : " + partPath);
                            throw new SemanticException(ErrorMsg.INSERT_INTO_IMMUTABLE_TABLE.getMsg(dest_tab.getTableName()));
                        }
                    } catch (IOException ioe) {
                        LOG.warn("Error while trying to determine if immutable table partition has any data : " + dest_tab.getTableName() + " : " + partPath);
                        throw new SemanticException(ErrorMsg.INSERT_INTO_IMMUTABLE_TABLE.getMsg(ioe.getMessage()));
                    }
                }
                // if the table is in a different dfs than the partition,
                // replace the partition's dfs with the table's dfs.
                dest_path = new Path(tabPath.toUri().getScheme(), tabPath.toUri().getAuthority(), partPath.toUri().getPath());
                queryTmpdir = ctx.getTempDirForPath(dest_path, true);
                table_desc = Utilities.getTableDesc(dest_tab);
                // Add sorting/bucketing if needed
                input = genBucketingSortingDest(dest, input, qb, table_desc, dest_tab, rsCtx);
                idToTableNameMap.put(String.valueOf(destTableId), dest_tab.getTableName());
                currentTableId = destTableId;
                destTableId++;
                lbCtx = constructListBucketingCtx(dest_part.getSkewedColNames(), dest_part.getSkewedColValues(), dest_part.getSkewedColValueLocationMaps(), dest_part.isStoredAsSubDirectories(), conf);
                AcidUtils.Operation acidOp = AcidUtils.Operation.NOT_ACID;
                if (destTableIsAcid) {
                    acidOp = getAcidType(table_desc.getOutputFileFormatClass(), dest);
                    checkAcidConstraints(qb, table_desc, dest_tab);
                }
                ltd = new LoadTableDesc(queryTmpdir, table_desc, dest_part.getSpec(), acidOp);
                ltd.setReplace(!qb.getParseInfo().isInsertIntoTable(dest_tab.getDbName(), dest_tab.getTableName()));
                ltd.setLbCtx(lbCtx);
                loadTableWork.add(ltd);
                if (!outputs.add(new WriteEntity(dest_part, determineWriteType(ltd, dest_tab.isNonNative(), dest)))) {
                    throw new SemanticException(ErrorMsg.OUTPUT_SPECIFIED_MULTIPLE_TIMES.getMsg(dest_tab.getTableName() + "@" + dest_part.getName()));
                }
                break;
            }
        case QBMetaData.DEST_LOCAL_FILE:
            isLocal = true;
        // fall through
        case QBMetaData.DEST_DFS_FILE:
            {
                dest_path = new Path(qbm.getDestFileForAlias(dest));
                if (isLocal) {
                    // for local directory - we always write to map-red intermediate
                    // store and then copy to local fs
                    queryTmpdir = ctx.getMRTmpPath();
                } else {
                    try {
                        Path qPath = FileUtils.makeQualified(dest_path, conf);
                        queryTmpdir = ctx.getTempDirForPath(qPath, true);
                    } catch (Exception e) {
                        throw new SemanticException("Error creating temporary folder on: " + dest_path, e);
                    }
                }
                String cols = "";
                String colTypes = "";
                ArrayList<ColumnInfo> colInfos = inputRR.getColumnInfos();
                // CTAS case: the file output format and serde are defined by the create
                // table command rather than taking the default value
                List<FieldSchema> field_schemas = null;
                CreateTableDesc tblDesc = qb.getTableDesc();
                CreateViewDesc viewDesc = qb.getViewDesc();
                if (tblDesc != null) {
                    field_schemas = new ArrayList<FieldSchema>();
                    destTableIsTemporary = tblDesc.isTemporary();
                    destTableIsMaterialization = tblDesc.isMaterialization();
                } else if (viewDesc != null) {
                    field_schemas = new ArrayList<FieldSchema>();
                    destTableIsTemporary = false;
                }
                boolean first = true;
                for (ColumnInfo colInfo : colInfos) {
                    String[] nm = inputRR.reverseLookup(colInfo.getInternalName());
                    if (nm[1] != null) {
                        // non-null column alias
                        colInfo.setAlias(nm[1]);
                    }
                    //default column name
                    String colName = colInfo.getInternalName();
                    if (field_schemas != null) {
                        FieldSchema col = new FieldSchema();
                        if (!("".equals(nm[0])) && nm[1] != null) {
                            // remove ``
                            colName = unescapeIdentifier(colInfo.getAlias()).toLowerCase();
                        }
                        colName = fixCtasColumnName(colName);
                        col.setName(colName);
                        String typeName = colInfo.getType().getTypeName();
                        // CTAS should NOT create a VOID type
                        if (typeName.equals(serdeConstants.VOID_TYPE_NAME)) {
                            throw new SemanticException(ErrorMsg.CTAS_CREATES_VOID_TYPE.getMsg(colName));
                        }
                        col.setType(typeName);
                        field_schemas.add(col);
                    }
                    if (!first) {
                        cols = cols.concat(",");
                        colTypes = colTypes.concat(":");
                    }
                    first = false;
                    cols = cols.concat(colName);
                    // Replace VOID type with string when the output is a temp table or
                    // local files.
                    // A VOID type can be generated under the query:
                    //
                    // select NULL from tt;
                    // or
                    // insert overwrite local directory "abc" select NULL from tt;
                    //
                    // where there is no column type to which the NULL value should be
                    // converted.
                    //
                    String tName = colInfo.getType().getTypeName();
                    if (tName.equals(serdeConstants.VOID_TYPE_NAME)) {
                        colTypes = colTypes.concat(serdeConstants.STRING_TYPE_NAME);
                    } else {
                        colTypes = colTypes.concat(tName);
                    }
                }
                // update the create table descriptor with the resulting schema.
                if (tblDesc != null) {
                    tblDesc.setCols(new ArrayList<FieldSchema>(field_schemas));
                } else if (viewDesc != null) {
                    viewDesc.setSchema(new ArrayList<FieldSchema>(field_schemas));
                }
                boolean isDestTempFile = true;
                if (!ctx.isMRTmpFileURI(dest_path.toUri().toString())) {
                    idToTableNameMap.put(String.valueOf(destTableId), dest_path.toUri().toString());
                    currentTableId = destTableId;
                    destTableId++;
                    isDestTempFile = false;
                }
                boolean isDfsDir = (dest_type.intValue() == QBMetaData.DEST_DFS_FILE);
                loadFileWork.add(new LoadFileDesc(tblDesc, viewDesc, queryTmpdir, dest_path, isDfsDir, cols, colTypes));
                if (tblDesc == null) {
                    if (viewDesc != null) {
                        table_desc = PlanUtils.getTableDesc(viewDesc, cols, colTypes);
                    } else if (qb.getIsQuery()) {
                        String fileFormat;
                        if (SessionState.get().getIsUsingThriftJDBCBinarySerDe()) {
                            fileFormat = "SequenceFile";
                            HiveConf.setVar(conf, HiveConf.ConfVars.HIVEQUERYRESULTFILEFORMAT, fileFormat);
                            table_desc = PlanUtils.getDefaultQueryOutputTableDesc(cols, colTypes, fileFormat, ThriftJDBCBinarySerDe.class);
                            // Set the fetch formatter to be a no-op for the ListSinkOperator, since we'll
                            // write out formatted thrift objects to SequenceFile
                            conf.set(SerDeUtils.LIST_SINK_OUTPUT_FORMATTER, NoOpFetchFormatter.class.getName());
                        } else {
                            fileFormat = HiveConf.getVar(conf, HiveConf.ConfVars.HIVEQUERYRESULTFILEFORMAT);
                            table_desc = PlanUtils.getDefaultQueryOutputTableDesc(cols, colTypes, fileFormat, LazySimpleSerDe.class);
                        }
                    } else {
                        table_desc = PlanUtils.getDefaultTableDesc(qb.getDirectoryDesc(), cols, colTypes);
                    }
                } else {
                    table_desc = PlanUtils.getTableDesc(tblDesc, cols, colTypes);
                }
                if (!outputs.add(new WriteEntity(dest_path, !isDfsDir, isDestTempFile))) {
                    throw new SemanticException(ErrorMsg.OUTPUT_SPECIFIED_MULTIPLE_TIMES.getMsg(dest_path.toUri().toString()));
                }
                break;
            }
        default:
            throw new SemanticException("Unknown destination type: " + dest_type);
    }
    input = genConversionSelectOperator(dest, qb, input, table_desc, dpCtx);
    inputRR = opParseCtx.get(input).getRowResolver();
    ArrayList<ColumnInfo> vecCol = new ArrayList<ColumnInfo>();
    if (updating(dest) || deleting(dest)) {
        vecCol.add(new ColumnInfo(VirtualColumn.ROWID.getName(), VirtualColumn.ROWID.getTypeInfo(), "", true));
    } else {
        try {
            StructObjectInspector rowObjectInspector = (StructObjectInspector) table_desc.getDeserializer(conf).getObjectInspector();
            List<? extends StructField> fields = rowObjectInspector.getAllStructFieldRefs();
            for (int i = 0; i < fields.size(); i++) {
                vecCol.add(new ColumnInfo(fields.get(i).getFieldName(), TypeInfoUtils.getTypeInfoFromObjectInspector(fields.get(i).getFieldObjectInspector()), "", false));
            }
        } catch (Exception e) {
            throw new SemanticException(e.getMessage(), e);
        }
    }
    RowSchema fsRS = new RowSchema(vecCol);
    // The output files of a FileSink can be merged if they are either not being written to a table
    // or are being written to a table which is not bucketed
    // and table the table is not sorted
    boolean canBeMerged = (dest_tab == null || !((dest_tab.getNumBuckets() > 0) || (dest_tab.getSortCols() != null && dest_tab.getSortCols().size() > 0)));
    // If this table is working with ACID semantics, turn off merging
    canBeMerged &= !destTableIsAcid;
    // Generate the partition columns from the parent input
    if (dest_type.intValue() == QBMetaData.DEST_TABLE || dest_type.intValue() == QBMetaData.DEST_PARTITION) {
        genPartnCols(dest, input, qb, table_desc, dest_tab, rsCtx);
    }
    FileSinkDesc fileSinkDesc = new FileSinkDesc(queryTmpdir, table_desc, conf.getBoolVar(HiveConf.ConfVars.COMPRESSRESULT), currentTableId, rsCtx.isMultiFileSpray(), canBeMerged, rsCtx.getNumFiles(), rsCtx.getTotalFiles(), rsCtx.getPartnCols(), dpCtx, dest_path);
    boolean isHiveServerQuery = SessionState.get().isHiveServerQuery();
    fileSinkDesc.setHiveServerQuery(isHiveServerQuery);
    // FileSinkOperator knows how to properly write to it.
    if (destTableIsAcid) {
        AcidUtils.Operation wt = updating(dest) ? AcidUtils.Operation.UPDATE : (deleting(dest) ? AcidUtils.Operation.DELETE : AcidUtils.Operation.INSERT);
        fileSinkDesc.setWriteType(wt);
        acidFileSinks.add(fileSinkDesc);
    }
    fileSinkDesc.setTemporary(destTableIsTemporary);
    fileSinkDesc.setMaterialization(destTableIsMaterialization);
    /* Set List Bucketing context. */
    if (lbCtx != null) {
        lbCtx.processRowSkewedIndex(fsRS);
        lbCtx.calculateSkewedValueSubDirList();
    }
    fileSinkDesc.setLbCtx(lbCtx);
    // set the stats publishing/aggregating key prefix
    // the same as directory name. The directory name
    // can be changed in the optimizer but the key should not be changed
    // it should be the same as the MoveWork's sourceDir.
    fileSinkDesc.setStatsAggPrefix(fileSinkDesc.getDirName().toString());
    if (!destTableIsMaterialization && HiveConf.getVar(conf, HIVESTATSDBCLASS).equalsIgnoreCase(StatDB.fs.name())) {
        String statsTmpLoc = ctx.getTempDirForPath(dest_path).toString();
        fileSinkDesc.setStatsTmpDir(statsTmpLoc);
        LOG.debug("Set stats collection dir : " + statsTmpLoc);
    }
    if (dest_part != null) {
        try {
            String staticSpec = Warehouse.makePartPath(dest_part.getSpec());
            fileSinkDesc.setStaticSpec(staticSpec);
        } catch (MetaException e) {
            throw new SemanticException(e);
        }
    } else if (dpCtx != null) {
        fileSinkDesc.setStaticSpec(dpCtx.getSPPath());
    }
    if (isHiveServerQuery && null != table_desc && table_desc.getSerdeClassName().equalsIgnoreCase(ThriftJDBCBinarySerDe.class.getName()) && HiveConf.getBoolVar(conf, HiveConf.ConfVars.HIVE_SERVER2_THRIFT_RESULTSET_SERIALIZE_IN_TASKS)) {
        fileSinkDesc.setIsUsingThriftJDBCBinarySerDe(true);
    } else {
        fileSinkDesc.setIsUsingThriftJDBCBinarySerDe(false);
    }
    Operator output = putOpInsertMap(OperatorFactory.getAndMakeChild(fileSinkDesc, fsRS, input), inputRR);
    if (ltd != null && SessionState.get() != null) {
        SessionState.get().getLineageState().mapDirToOp(ltd.getSourcePath(), (FileSinkOperator) output);
    } else if (queryState.getCommandType().equals(HiveOperation.CREATETABLE_AS_SELECT.getOperationName())) {
        Path tlocation = null;
        String tName = Utilities.getDbTableName(tableDesc.getTableName())[1];
        try {
            Warehouse wh = new Warehouse(conf);
            tlocation = wh.getTablePath(db.getDatabase(tableDesc.getDatabaseName()), tName);
        } catch (MetaException | HiveException e) {
            throw new SemanticException(e);
        }
        SessionState.get().getLineageState().mapDirToOp(tlocation, (FileSinkOperator) output);
    }
    if (LOG.isDebugEnabled()) {
        LOG.debug("Created FileSink Plan for clause: " + dest + "dest_path: " + dest_path + " row schema: " + inputRR.toString());
    }
    FileSinkOperator fso = (FileSinkOperator) output;
    fso.getConf().setTable(dest_tab);
    fsopToTable.put(fso, dest_tab);
    // and it is an insert overwrite or insert into table
    if (dest_tab != null && conf.getBoolVar(ConfVars.HIVESTATSAUTOGATHER) && conf.getBoolVar(ConfVars.HIVESTATSCOLAUTOGATHER) && ColumnStatsAutoGatherContext.canRunAutogatherStats(fso)) {
        if (dest_type.intValue() == QBMetaData.DEST_TABLE) {
            genAutoColumnStatsGatheringPipeline(qb, table_desc, partSpec, input, qb.getParseInfo().isInsertIntoTable(dest_tab.getDbName(), dest_tab.getTableName()));
        } else if (dest_type.intValue() == QBMetaData.DEST_PARTITION) {
            genAutoColumnStatsGatheringPipeline(qb, table_desc, dest_part.getSpec(), input, qb.getParseInfo().isInsertIntoTable(dest_tab.getDbName(), dest_tab.getTableName()));
        }
    }
    return output;
}

Example 5 with LoadFileDesc

use of org.apache.hadoop.hive.ql.plan.LoadFileDesc in project hive by apache.

the class GenMapRedUtils method mergeMovePaths.

/**
   * Merges the given Conditional input path and the linked MoveWork into one only MoveWork.
   * This is an optimization for BlobStore systems to avoid doing two renames or copies that are not necessary.
   *
   * @param condInputPath A path that the ConditionalTask uses as input for its sub-tasks.
   * @param linkedMoveWork A MoveWork that the ConditionalTask uses to link to its sub-tasks.
   * @return A new MoveWork that has the Conditional input path as source and the linkedMoveWork as target.
   */
@VisibleForTesting
protected static MoveWork mergeMovePaths(Path condInputPath, MoveWork linkedMoveWork) {
    MoveWork newWork = new MoveWork(linkedMoveWork);
    LoadFileDesc fileDesc = null;
    LoadTableDesc tableDesc = null;
    if (linkedMoveWork.getLoadFileWork() != null) {
        fileDesc = new LoadFileDesc(linkedMoveWork.getLoadFileWork());
        fileDesc.setSourcePath(condInputPath);
    } else if (linkedMoveWork.getLoadTableWork() != null) {
        tableDesc = new LoadTableDesc(linkedMoveWork.getLoadTableWork());
        tableDesc.setSourcePath(condInputPath);
    } else {
        throw new IllegalArgumentException("Merging a path with a MoveWork with multi-files work is not allowed.");
    }
    newWork.setLoadFileWork(fileDesc);
    newWork.setLoadTableWork(tableDesc);
    return newWork;
}