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

Example 36 with MapredWork

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

the class GenMapRedUtils method getMapRedWork.

/**
 * create a new plan and return.
 *
 * @return the new plan
 */
public static MapredWork getMapRedWork(ParseContext parseCtx) {
    MapredWork work = getMapRedWorkFromConf(parseCtx.getConf());
    work.getMapWork().setNameToSplitSample(parseCtx.getNameToSplitSample());
    return work;
}

Example 37 with MapredWork

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

the class GenMapRedUtils method createMRWorkForMergingFiles.

/**
 * Create a MapredWork based on input path, the top operator and the input
 * table descriptor.
 *
 * @param conf
 * @param topOp
 *          the table scan operator that is the root of the MapReduce task.
 * @param fsDesc
 *          the file sink descriptor that serves as the input to this merge task.
 * @param parentMR
 *          the parent MapReduce work
 * @param parentFS
 *          the last FileSinkOperator in the parent MapReduce work
 * @return the MapredWork
 */
private static MapWork createMRWorkForMergingFiles(HiveConf conf, TableScanOperator topOp, FileSinkDesc fsDesc) {
    ArrayList<String> aliases = new ArrayList<String>();
    Path inputDir = StringInternUtils.internUriStringsInPath(fsDesc.getMergeInputDirName());
    String inputDirStr = inputDir.toString().intern();
    TableDesc tblDesc = fsDesc.getTableInfo();
    // dummy alias: just use the input path
    aliases.add(inputDirStr);
    // constructing the default MapredWork
    MapredWork cMrPlan = GenMapRedUtils.getMapRedWorkFromConf(conf);
    MapWork cplan = cMrPlan.getMapWork();
    cplan.addPathToAlias(inputDir, aliases);
    cplan.addPathToPartitionInfo(inputDir, new PartitionDesc(tblDesc, null));
    cplan.getAliasToWork().put(inputDirStr, topOp);
    cplan.setMapperCannotSpanPartns(true);
    return cplan;
}

Example 38 with MapredWork

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

the class GenMapRedUtils method setUnionPlan.

private static void setUnionPlan(GenMRProcContext opProcCtx, boolean local, Task<? extends Serializable> currTask, GenMRUnionCtx uCtx, boolean mergeTask) throws SemanticException {
    TableScanOperator currTopOp = opProcCtx.getCurrTopOp();
    if (currTopOp != null) {
        String currAliasId = opProcCtx.getCurrAliasId();
        if (mergeTask || !opProcCtx.isSeenOp(currTask, currTopOp)) {
            setTaskPlan(currAliasId, currTopOp, currTask, local, opProcCtx);
        }
        currTopOp = null;
        opProcCtx.setCurrTopOp(currTopOp);
    } else {
        List<String> taskTmpDirLst = uCtx.getTaskTmpDir();
        if ((taskTmpDirLst != null) && !(taskTmpDirLst.isEmpty())) {
            List<TableDesc> tt_descLst = uCtx.getTTDesc();
            assert !taskTmpDirLst.isEmpty() && !tt_descLst.isEmpty();
            assert taskTmpDirLst.size() == tt_descLst.size();
            int size = taskTmpDirLst.size();
            assert local == false;
            List<TableScanOperator> topOperators = uCtx.getListTopOperators();
            MapredWork plan = (MapredWork) currTask.getWork();
            for (int pos = 0; pos < size; pos++) {
                String taskTmpDir = taskTmpDirLst.get(pos);
                Path taskTmpDirPath = new Path(taskTmpDir);
                MapWork mWork = plan.getMapWork();
                if (!mWork.getPathToAliases().containsKey(taskTmpDirPath)) {
                    taskTmpDir = taskTmpDir.intern();
                    StringInternUtils.internUriStringsInPath(taskTmpDirPath);
                    TableDesc tt_desc = tt_descLst.get(pos);
                    mWork.addPathToAlias(taskTmpDirPath, taskTmpDir);
                    mWork.addPathToPartitionInfo(taskTmpDirPath, new PartitionDesc(tt_desc, null));
                    mWork.getAliasToWork().put(taskTmpDir, topOperators.get(pos));
                }
            }
        }
    }
}

Example 39 with MapredWork

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

the class GenMapRedUtils method initUnionPlan.

/**
 * Initialize the current union plan.
 *
 * @param op
 *          the reduce sink operator encountered
 * @param opProcCtx
 *          processing context
 */
public static void initUnionPlan(ReduceSinkOperator op, UnionOperator currUnionOp, GenMRProcContext opProcCtx, Task<? extends Serializable> unionTask) throws SemanticException {
    Operator<? extends OperatorDesc> reducer = op.getChildOperators().get(0);
    MapredWork plan = (MapredWork) unionTask.getWork();
    HashMap<Operator<? extends OperatorDesc>, Task<? extends Serializable>> opTaskMap = opProcCtx.getOpTaskMap();
    opTaskMap.put(reducer, unionTask);
    plan.setReduceWork(new ReduceWork());
    plan.getReduceWork().setReducer(reducer);
    plan.getReduceWork().setReducer(reducer);
    ReduceSinkDesc desc = op.getConf();
    plan.getReduceWork().setNumReduceTasks(desc.getNumReducers());
    if (needsTagging(plan.getReduceWork())) {
        plan.getReduceWork().setNeedsTagging(true);
    }
    initUnionPlan(opProcCtx, currUnionOp, unionTask, false);
}

Example 40 with MapredWork

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

the class ExecDriver method main.

@SuppressWarnings("unchecked")
public static void main(String[] args) throws IOException, HiveException {
    String planFileName = null;
    String jobConfFileName = null;
    boolean noLog = false;
    String files = null;
    String libjars = null;
    boolean localtask = false;
    try {
        for (int i = 0; i < args.length; i++) {
            if (args[i].equals("-plan")) {
                planFileName = args[++i];
            } else if (args[i].equals("-jobconffile")) {
                jobConfFileName = args[++i];
            } else if (args[i].equals("-nolog")) {
                noLog = true;
            } else if (args[i].equals("-files")) {
                files = args[++i];
            } else if (args[i].equals("-libjars")) {
                libjars = args[++i];
            } else if (args[i].equals("-localtask")) {
                localtask = true;
            }
        }
    } catch (IndexOutOfBoundsException e) {
        System.err.println("Missing argument to option");
        printUsage();
    }
    JobConf conf;
    if (localtask) {
        conf = new JobConf(MapredLocalTask.class);
    } else {
        conf = new JobConf(ExecDriver.class);
    }
    if (jobConfFileName != null) {
        conf.addResource(new Path(jobConfFileName));
    }
    // Initialize the resources from command line
    if (files != null) {
        conf.set("tmpfiles", files);
    }
    if (libjars != null) {
        conf.set("tmpjars", libjars);
    }
    if (UserGroupInformation.isSecurityEnabled()) {
        String hadoopAuthToken = System.getenv(UserGroupInformation.HADOOP_TOKEN_FILE_LOCATION);
        if (hadoopAuthToken != null) {
            conf.set("mapreduce.job.credentials.binary", hadoopAuthToken);
        }
    }
    boolean isSilent = HiveConf.getBoolVar(conf, HiveConf.ConfVars.HIVESESSIONSILENT);
    String queryId = HiveConf.getVar(conf, HiveConf.ConfVars.HIVEQUERYID, "").trim();
    if (queryId.isEmpty()) {
        queryId = "unknown-" + System.currentTimeMillis();
        HiveConf.setVar(conf, HiveConf.ConfVars.HIVEQUERYID, queryId);
    }
    System.setProperty(HiveConf.ConfVars.HIVEQUERYID.toString(), queryId);
    LogUtils.registerLoggingContext(conf);
    if (noLog) {
        // If started from main(), and noLog is on, we should not output
        // any logs. To turn the log on, please set -Dtest.silent=false
        org.apache.logging.log4j.Logger logger = org.apache.logging.log4j.LogManager.getRootLogger();
        NullAppender appender = NullAppender.createNullAppender();
        appender.addToLogger(logger.getName(), Level.ERROR);
        appender.start();
    } else {
        setupChildLog4j(conf);
    }
    Logger LOG = LoggerFactory.getLogger(ExecDriver.class.getName());
    LogHelper console = new LogHelper(LOG, isSilent);
    if (planFileName == null) {
        console.printError("Must specify Plan File Name");
        printUsage();
    }
    // that it's easy to find reason for local mode execution failures
    for (Appender appender : ((org.apache.logging.log4j.core.Logger) LogManager.getRootLogger()).getAppenders().values()) {
        if (appender instanceof FileAppender) {
            console.printInfo("Execution log at: " + ((FileAppender) appender).getFileName());
        } else if (appender instanceof RollingFileAppender) {
            console.printInfo("Execution log at: " + ((RollingFileAppender) appender).getFileName());
        }
    }
    // the plan file should always be in local directory
    Path p = new Path(planFileName);
    FileSystem fs = FileSystem.getLocal(conf);
    InputStream pathData = fs.open(p);
    // child process. so we add it here explicitly
    try {
        // see also - code in CliDriver.java
        ClassLoader loader = conf.getClassLoader();
        if (StringUtils.isNotBlank(libjars)) {
            loader = Utilities.addToClassPath(loader, StringUtils.split(libjars, ","));
        }
        conf.setClassLoader(loader);
        // Also set this to the Thread ContextClassLoader, so new threads will
        // inherit
        // this class loader, and propagate into newly created Configurations by
        // those
        // new threads.
        Thread.currentThread().setContextClassLoader(loader);
    } catch (Exception e) {
        throw new HiveException(e.getMessage(), e);
    }
    int ret;
    if (localtask) {
        memoryMXBean = ManagementFactory.getMemoryMXBean();
        MapredLocalWork plan = SerializationUtilities.deserializePlan(pathData, MapredLocalWork.class);
        MapredLocalTask ed = new MapredLocalTask(plan, conf, isSilent);
        ret = ed.executeInProcess(new DriverContext());
    } else {
        MapredWork plan = SerializationUtilities.deserializePlan(pathData, MapredWork.class);
        ExecDriver ed = new ExecDriver(plan, conf, isSilent);
        ret = ed.execute(new DriverContext());
    }
    if (ret != 0) {
        System.exit(ret);
    }
}