List of usage examples for org.apache.hadoop.mapred JobConf getBoolean
public boolean getBoolean(String name, boolean defaultValue)
name property as a boolean. From source file:edu.ucsb.cs.partitioning.cosine.Partitioner.java
License:Apache License
/** * //from w ww.j a va 2 s . co m * @param job * @param inputDir * @param interDir * @param maxDir * @param nPartitions * @param norm_weight_all * @return number of partitions actaully produced */ public static int produceStaticParitions(JobConf job, String inputDir, String interDir, String maxDir, int nPartitions, int norm_weight_all) { SequenceFile.Writer partOut = null; float maxn = 0, maxw = 0, pChoice = job.getFloat(NormSortMain.P_NORM_PROPERTY, NormSortMain.P_NORM_VALUE); int maxs = 0; LongWritable prevK = null, key = new LongWritable(); FeatureWeightArrayWritable prevV = null, value = new FeatureWeightArrayWritable(); try { Path inputPath = new Path(inputDir); FileSystem hdfs = inputPath.getFileSystem(new Configuration()); Path interDirectory = new Path(interDir); Path maxPath = new Path(maxDir); clearPath(hdfs, maxPath); clearPath(hdfs, interDirectory); long nDocuments = Collector.countDirVectors(hdfs, inputPath, job); if (nDocuments == 0) return 0; double partitionSize; uniformPartitions = job.getBoolean(Config.UNIFORM_PARTITIONING_PROPERTY, Config.UNIFORM_PARTITIONING_VALUE); if (uniformPartitions) partitionSize = Math.ceil(nDocuments / (double) nPartitions); else partitionSize = Math.ceil(nDocuments / (double) (GijComparisons.choose(nPartitions + 1, 2))); if (partitionSize == 1) System.err.println("WARN: Number of partitions = number of documents!!"); FileStatus[] files = setFiles(hdfs, inputPath); FSDataOutputStream maxOut = hdfs.create(maxPath); int documentNo = 0, partitionNo = 1; // partition naming start at 1 for (int i = 0; i < files.length; i++) { inputPath = files[i].getPath(); if ((hdfs.isDirectory(inputPath) || inputPath.getName().startsWith("_"))) continue; Reader in = new SequenceFile.Reader(hdfs, inputPath, job); while (in.next(key, value)) { // id,vector documentNo++; prevK = key; prevV = value; if (isFirstDocument(partOut)) { maxn = value.getPNorm(pChoice); maxw = value.getMaxWeight(); maxs = value.vectorSize; partOut = openFile(hdfs, job, interDirectory, partitionNo); } partOut.append(key, value); maxw = (value.getMaxWeight() > maxw) ? value.getMaxWeight() : maxw; maxs = (value.vectorSize > maxs) ? value.vectorSize : maxs; maxn = (value.getPNorm(pChoice) > maxn) ? value.getPNorm(pChoice) : maxn; if (isLastDocument(documentNo, partitionNo, partitionSize, uniformPartitions)) { partOut = writeMax(norm_weight_all, partOut, maxOut, maxn, maxw, maxs); documentNo = 0; partitionNo++; } prevK = key; prevV = value; } in.close(); } if (partOut != null) partOut = writeMax(norm_weight_all, partOut, maxOut, maxn, maxw, maxs); nPartitions = partitionNo - 1; maxOut.close(); } catch (Exception e) { e.printStackTrace(); } return (nPartitions); }
From source file:edu.umd.cloud9.webgraph.ClueExtractLinks.java
License:Apache License
public int runTool() throws Exception { JobConf conf = new JobConf(getConf(), ClueExtractLinks.class); FileSystem fs = FileSystem.get(conf); int numMappers = conf.getInt("Cloud9.Mappers", 1); int numReducers = conf.getInt("Cloud9.Reducers", 200); String inputPath = conf.get("Cloud9.InputPath"); String outputPath = conf.get("Cloud9.OutputPath"); String mappingFile = conf.get("Cloud9.DocnoMappingFile"); if (!fs.exists(new Path(mappingFile))) { throw new RuntimeException("Error: Docno mapping data file " + mappingFile + " doesn't exist!"); }/*from ww w .java 2 s .c om*/ DistributedCache.addCacheFile(new URI(mappingFile), conf); conf.setJobName("ClueExtractLinks"); conf.set("mapred.child.java.opts", "-Xmx2048m"); conf.setInt("mapred.task.timeout", 60000000); conf.set("mapreduce.map.memory.mb", "2048"); conf.set("mapreduce.map.java.opts", "-Xmx2048m"); conf.set("mapreduce.reduce.memory.mb", "2048"); conf.set("mapreduce.reduce.java.opts", "-Xmx2048m"); conf.set("mapreduce.task.timeout", "60000000"); conf.setNumMapTasks(numMappers); conf.setNumReduceTasks(numReducers); // TODO: to read!! conf.setMapperClass(Map.class); conf.setCombinerClass(Reduce.class); conf.setReducerClass(Reduce.class); conf.setOutputKeyClass(Text.class); conf.setOutputValueClass(ArrayListWritable.class); conf.setInputFormat(SequenceFileInputFormat.class); conf.setOutputFormat(SequenceFileOutputFormat.class); SequenceFileOutputFormat.setCompressOutput(conf, true); SequenceFileOutputFormat.setOutputCompressionType(conf, SequenceFile.CompressionType.BLOCK); SequenceFileInputFormat.setInputPaths(conf, inputPath); FileOutputFormat.setOutputPath(conf, new Path(outputPath)); LOG.info("ClueExtractLinks"); LOG.info(" - input path: " + inputPath); LOG.info(" - output path: " + outputPath); LOG.info(" - mapping file: " + mappingFile); LOG.info(" - include internal links? " + conf.getBoolean("Cloud9.IncludeInternalLinks", false)); if (!fs.exists(new Path(outputPath))) { JobClient.runJob(conf); } else { LOG.info(outputPath + " already exists! Skipping this step..."); } return 0; }
From source file:edu.umd.cloud9.webgraph.ExtractLinks.java
License:Apache License
public int runTool() throws Exception { JobConf conf = new JobConf(getConf(), ExtractLinks.class); FileSystem fs = FileSystem.get(conf); int numMappers = conf.getInt("Cloud9.Mappers", 1); int numReducers = conf.getInt("Cloud9.Reducers", 200); String inputPath = conf.get("Cloud9.InputPath"); String outputPath = conf.get("Cloud9.OutputPath"); String mappingFile = conf.get("Cloud9.DocnoMappingFile"); if (!fs.exists(new Path(mappingFile))) throw new RuntimeException("Error: Docno mapping data file " + mappingFile + " doesn't exist!"); DistributedCache.addCacheFile(new URI(mappingFile), conf); conf.setJobName("ExtractLinks"); conf.set("mapred.child.java.opts", "-Xmx2048m"); conf.setInt("mapred.task.timeout", 60000000); conf.setNumMapTasks(numMappers);//from w w w.java 2 s . c o m conf.setNumReduceTasks(numReducers); conf.setMapperClass(Map.class); conf.setCombinerClass(Reduce.class); conf.setReducerClass(Reduce.class); conf.setOutputKeyClass(Text.class); conf.setOutputValueClass(ArrayListWritable.class); conf.setInputFormat(SequenceFileInputFormat.class); conf.setOutputFormat(SequenceFileOutputFormat.class); SequenceFileOutputFormat.setCompressOutput(conf, true); SequenceFileOutputFormat.setOutputCompressionType(conf, SequenceFile.CompressionType.BLOCK); SequenceFileInputFormat.setInputPaths(conf, inputPath); FileOutputFormat.setOutputPath(conf, new Path(outputPath)); LOG.info("ExtractLinks"); LOG.info(" - input path: " + inputPath); LOG.info(" - output path: " + outputPath); LOG.info(" - mapping file: " + mappingFile); LOG.info(" - include internal links? " + conf.getBoolean("Cloud9.IncludeInternalLinks", false)); if (!fs.exists(new Path(outputPath))) { JobClient.runJob(conf); } else { LOG.info(outputPath + " already exists! Skipping this step..."); } return 0; }
From source file:edu.umn.cs.spatialHadoop.nasa.DistributedAggregateSpatioTemporalIndexer.java
License:Open Source License
/** * Build a bunch of AggregateQuadTrees using a Map-Reduce job * /*from w ww. jav a 2s .c o m*/ * @param inputPathsDictionaryPath * @param params * @throws IOException */ public static void aggregateQuadTreeMapReduce(Path inputPathsDictionaryPath, OperationsParams params) throws IOException { // configure a map-reduce job JobConf job = new JobConf(params, DistributedAggregateSpatioTemporalIndexer.class); Path outputPath; String outputPathPrefix = "aggQuadTree_"; FileSystem outFs = FileSystem.get(job); do { outputPath = new Path(outputPathPrefix + (int) (Math.random() * 1000000)); } while (outFs.exists(outputPath)); job.setJobName("AggregateQuadTree"); job.setMapOutputKeyClass(Text.class); job.setMapOutputValueClass(Text.class); job.setMapperClass(AggregateQuadTreeMaper.class); job.set(HDFSIndexPath, hdfsIndexPath.toString()); ClusterStatus clusterStatus = new JobClient(job).getClusterStatus(); job.setNumMapTasks(clusterStatus.getMaxMapTasks() * 5); job.setInputFormat(TextInputFormat.class); job.setOutputFormat(TextOutputFormat.class); TextInputFormat.setInputPaths(job, inputPathsDictionaryPath); TextOutputFormat.setOutputPath(job, outputPath); if (job.getBoolean("local", false)) { // Enforce local execution if explicitly set by user or for small // files job.set("mapred.job.tracker", "local"); // Use multithreading too job.setInt(LocalJobRunner.LOCAL_MAX_MAPS, 16); } job.setNumReduceTasks(0); // Submit the job JobClient.runJob(job); outFs.delete(outputPath, true); }
From source file:edu.umn.cs.spatialHadoop.operations.Contains.java
License:Open Source License
public static <S extends Shape> long contains(Path[] inFiles, Path userOutputPath, OperationsParams params) throws IOException, InterruptedException { JobConf job = new JobConf(params, Contains.class); LOG.info("Contains journey starts ...."); FileSystem inFs = inFiles[0].getFileSystem(job); Path outputPath = userOutputPath; if (outputPath == null) { FileSystem outFs = FileSystem.get(job); do {/*w ww .j a va2 s .c om*/ outputPath = new Path(inFiles[0].getName() + ".sjmr_" + (int) (Math.random() * 1000000)); } while (outFs.exists(outputPath)); } FileSystem outFs = outputPath.getFileSystem(job); ClusterStatus clusterStatus = new JobClient(job).getClusterStatus(); job.setJobName("Within"); job.setMapperClass(ContainsMap.class); job.setMapOutputKeyClass(IntWritable.class); job.setMapOutputValueClass(IndexedText.class); job.setNumMapTasks(5 * Math.max(1, clusterStatus.getMaxMapTasks())); job.setLong("mapred.min.split.size", Math.max(inFs.getFileStatus(inFiles[0]).getBlockSize(), inFs.getFileStatus(inFiles[1]).getBlockSize())); job.setReducerClass(ContainsReduce.class); job.setNumReduceTasks(Math.max(1, clusterStatus.getMaxReduceTasks())); job.setInputFormat(ShapeLineInputFormat.class); if (job.getBoolean("output", true)) job.setOutputFormat(TextOutputFormat.class); else job.setOutputFormat(NullOutputFormat.class); ShapeLineInputFormat.setInputPaths(job, inFiles); // Calculate and set the dimensions of the grid to use in the map phase long total_size = 0; Rectangle mbr = new Rectangle(Double.MAX_VALUE, Double.MAX_VALUE, -Double.MAX_VALUE, -Double.MAX_VALUE); for (Path file : inFiles) { FileSystem fs = file.getFileSystem(params); Rectangle file_mbr = FileMBR.fileMBR(file, params); mbr.expand(file_mbr); total_size += FileUtil.getPathSize(fs, file); } // If the largest file is globally indexed, use its partitions total_size += total_size * job.getFloat(SpatialSite.INDEXING_OVERHEAD, 0.2f); int sjmrPartitioningGridFactor = params.getInt(PartitioiningFactor, 20); int num_cells = (int) Math.max(1, total_size * sjmrPartitioningGridFactor / outFs.getDefaultBlockSize(outputPath)); LOG.info("Number of cells is configured to be " + num_cells); OperationsParams.setInactiveModeFlag(job, InactiveMode, isReduceInactive); OperationsParams.setJoiningThresholdPerOnce(job, JoiningThresholdPerOnce, joiningThresholdPerOnce); OperationsParams.setFilterOnlyModeFlag(job, isFilterOnlyMode, isFilterOnly); GridInfo gridInfo = new GridInfo(mbr.x1, mbr.y1, mbr.x2, mbr.y2); gridInfo.calculateCellDimensions(num_cells); OperationsParams.setShape(job, PartitionGrid, gridInfo); TextOutputFormat.setOutputPath(job, outputPath); if (OperationsParams.isLocal(job, inFiles)) { // Enforce local execution if explicitly set by user or for small files job.set("mapred.job.tracker", "local"); } // Start the job RunningJob runningJob = JobClient.runJob(job); Counters counters = runningJob.getCounters(); Counter outputRecordCounter = counters.findCounter(Task.Counter.REDUCE_OUTPUT_RECORDS); final long resultCount = outputRecordCounter.getValue(); return resultCount; }
From source file:edu.umn.cs.spatialHadoop.operations.Crosses.java
License:Open Source License
public static <S extends Shape> long crosses(Path[] inFiles, Path userOutputPath, OperationsParams params) throws IOException, InterruptedException { JobConf job = new JobConf(params, Crosses.class); LOG.info("Crosses journey starts ...."); FileSystem inFs = inFiles[0].getFileSystem(job); Path outputPath = userOutputPath; if (outputPath == null) { FileSystem outFs = FileSystem.get(job); do {//from ww w. ja v a 2s . co m outputPath = new Path(inFiles[0].getName() + ".sjmr_" + (int) (Math.random() * 1000000)); } while (outFs.exists(outputPath)); } FileSystem outFs = outputPath.getFileSystem(job); ClusterStatus clusterStatus = new JobClient(job).getClusterStatus(); job.setJobName("Crosses"); job.setMapperClass(CrossesMap.class); job.setMapOutputKeyClass(IntWritable.class); job.setMapOutputValueClass(IndexedText.class); job.setNumMapTasks(5 * Math.max(1, clusterStatus.getMaxMapTasks())); job.setLong("mapred.min.split.size", Math.max(inFs.getFileStatus(inFiles[0]).getBlockSize(), inFs.getFileStatus(inFiles[1]).getBlockSize())); job.setReducerClass(CrossesReduce.class); job.setNumReduceTasks(Math.max(1, clusterStatus.getMaxReduceTasks())); job.setInputFormat(ShapeLineInputFormat.class); if (job.getBoolean("output", true)) job.setOutputFormat(TextOutputFormat.class); else job.setOutputFormat(NullOutputFormat.class); ShapeLineInputFormat.setInputPaths(job, inFiles); // Calculate and set the dimensions of the grid to use in the map phase long total_size = 0; Rectangle mbr = new Rectangle(Double.MAX_VALUE, Double.MAX_VALUE, -Double.MAX_VALUE, -Double.MAX_VALUE); for (Path file : inFiles) { FileSystem fs = file.getFileSystem(params); Rectangle file_mbr = FileMBR.fileMBR(file, params); mbr.expand(file_mbr); total_size += FileUtil.getPathSize(fs, file); } // If the largest file is globally indexed, use its partitions total_size += total_size * job.getFloat(SpatialSite.INDEXING_OVERHEAD, 0.2f); int sjmrPartitioningGridFactor = params.getInt(PartitioiningFactor, 20); int num_cells = (int) Math.max(1, total_size * sjmrPartitioningGridFactor / outFs.getDefaultBlockSize(outputPath)); LOG.info("Number of cells is configured to be " + num_cells); OperationsParams.setInactiveModeFlag(job, InactiveMode, isReduceInactive); OperationsParams.setJoiningThresholdPerOnce(job, JoiningThresholdPerOnce, joiningThresholdPerOnce); OperationsParams.setFilterOnlyModeFlag(job, isFilterOnlyMode, isFilterOnly); GridInfo gridInfo = new GridInfo(mbr.x1, mbr.y1, mbr.x2, mbr.y2); gridInfo.calculateCellDimensions(num_cells); OperationsParams.setShape(job, PartitionGrid, gridInfo); TextOutputFormat.setOutputPath(job, outputPath); if (OperationsParams.isLocal(job, inFiles)) { // Enforce local execution if explicitly set by user or for small files job.set("mapred.job.tracker", "local"); } // Start the job RunningJob runningJob = JobClient.runJob(job); Counters counters = runningJob.getCounters(); Counter outputRecordCounter = counters.findCounter(Task.Counter.REDUCE_OUTPUT_RECORDS); final long resultCount = outputRecordCounter.getValue(); return resultCount; }
From source file:edu.umn.cs.spatialHadoop.operations.Disjoint.java
License:Open Source License
public static <S extends Shape> long disjoint(Path[] inFiles, Path userOutputPath, OperationsParams params) throws IOException, InterruptedException { JobConf job = new JobConf(params, Disjoint.class); LOG.info("Touches journey starts ...."); FileSystem inFs = inFiles[0].getFileSystem(job); Path outputPath = userOutputPath; if (outputPath == null) { FileSystem outFs = FileSystem.get(job); do {/*from w w w .j a va2 s. com*/ outputPath = new Path(inFiles[0].getName() + ".sjmr_" + (int) (Math.random() * 1000000)); } while (outFs.exists(outputPath)); } FileSystem outFs = outputPath.getFileSystem(job); ClusterStatus clusterStatus = new JobClient(job).getClusterStatus(); job.setJobName("Disjoint"); job.setMapperClass(DisjointMap.class); job.setMapOutputKeyClass(IntWritable.class); job.setMapOutputValueClass(IndexedText.class); job.setNumMapTasks(5 * Math.max(1, clusterStatus.getMaxMapTasks())); job.setLong("mapred.min.split.size", Math.max(inFs.getFileStatus(inFiles[0]).getBlockSize(), inFs.getFileStatus(inFiles[1]).getBlockSize())); job.setReducerClass(DisjointReduce.class); job.setNumReduceTasks(Math.max(1, clusterStatus.getMaxReduceTasks())); job.setInputFormat(ShapeLineInputFormat.class); if (job.getBoolean("output", true)) job.setOutputFormat(TextOutputFormat.class); else job.setOutputFormat(NullOutputFormat.class); ShapeLineInputFormat.setInputPaths(job, inFiles); // Calculate and set the dimensions of the grid to use in the map phase long total_size = 0; Rectangle mbr = new Rectangle(Double.MAX_VALUE, Double.MAX_VALUE, -Double.MAX_VALUE, -Double.MAX_VALUE); for (Path file : inFiles) { FileSystem fs = file.getFileSystem(params); Rectangle file_mbr = FileMBR.fileMBR(file, params); mbr.expand(file_mbr); total_size += FileUtil.getPathSize(fs, file); } // If the largest file is globally indexed, use its partitions total_size += total_size * job.getFloat(SpatialSite.INDEXING_OVERHEAD, 0.2f); int sjmrPartitioningGridFactor = params.getInt(PartitioiningFactor, 20); int num_cells = (int) Math.max(1, total_size * sjmrPartitioningGridFactor / outFs.getDefaultBlockSize(outputPath)); LOG.info("Number of cells is configured to be " + num_cells); OperationsParams.setInactiveModeFlag(job, InactiveMode, isReduceInactive); OperationsParams.setJoiningThresholdPerOnce(job, JoiningThresholdPerOnce, joiningThresholdPerOnce); OperationsParams.setFilterOnlyModeFlag(job, isFilterOnlyMode, isFilterOnly); GridInfo gridInfo = new GridInfo(mbr.x1, mbr.y1, mbr.x2, mbr.y2); gridInfo.calculateCellDimensions(num_cells); OperationsParams.setShape(job, PartitionGrid, gridInfo); TextOutputFormat.setOutputPath(job, outputPath); if (OperationsParams.isLocal(job, inFiles)) { // Enforce local execution if explicitly set by user or for small files job.set("mapred.job.tracker", "local"); } // Start the job RunningJob runningJob = JobClient.runJob(job); Counters counters = runningJob.getCounters(); Counter outputRecordCounter = counters.findCounter(Task.Counter.REDUCE_OUTPUT_RECORDS); final long resultCount = outputRecordCounter.getValue(); return resultCount; }
From source file:edu.umn.cs.spatialHadoop.operations.DistributedJoin.java
License:Open Source License
/** * Performs a redistribute join between the given files using the * redistribute join algorithm. Currently, we only support a pair of files. * @param inFiles/*from w ww. j av a 2 s . c o m*/ * @param userOutputPath * @param params * @return * @throws IOException */ public static <S extends Shape> long joinStep(Path[] inFiles, Path userOutputPath, OperationsParams params) throws IOException { long t1 = System.currentTimeMillis(); JobConf job = new JobConf(params, DistributedJoin.class); FileSystem fs[] = new FileSystem[inFiles.length]; for (int i_file = 0; i_file < inFiles.length; i_file++) fs[i_file] = inFiles[i_file].getFileSystem(job); Path outputPath = userOutputPath; if (outputPath == null) { do { outputPath = new Path(inFiles[0].getName() + ".dj_" + (int) (Math.random() * 1000000)); } while (fs[0].exists(outputPath)); } job.setJobName("DistributedJoin"); ClusterStatus clusterStatus = new JobClient(job).getClusterStatus(); GlobalIndex<Partition> gindex1 = SpatialSite.getGlobalIndex(fs[0], inFiles[0]); GlobalIndex<Partition> gindex2 = SpatialSite.getGlobalIndex(fs[1], inFiles[1]); OperationsParams.setFilterOnlyModeFlag(job, isFilterOnlyMode, isFilterOnly); LOG.info("Joining " + inFiles[0] + " X " + inFiles[1]); if (SpatialSite.isRTree(fs[0], inFiles[0]) && SpatialSite.isRTree(fs[1], inFiles[1])) { job.setInputFormat(DJInputFormatRTree.class); } else { if (isOneShotReadMode) { // Ensure all objects are read in one shot job.setInt(SpatialSite.MaxBytesInOneRead, -1); job.setInt(SpatialSite.MaxShapesInOneRead, -1); } else { job.setInt(SpatialSite.MaxBytesInOneRead, maxBytesInOneRead); job.setInt(SpatialSite.MaxShapesInOneRead, maxShapesInOneRead); } job.setInputFormat(DJInputFormatArray.class); } // Set input paths and map function if (inFiles[0].equals(inFiles[1])) { // Self join job.setInputFormat(ShapeArrayInputFormat.class); // Remove the spatial filter to ensure all partitions are loaded FileInputFormat.setInputPaths(job, inFiles[0]); if (gindex1 != null && gindex1.isReplicated()) job.setMapperClass(RedistributeJoinMap.class); else job.setMapperClass(RedistributeJoinMapNoDupAvoidance.class); } else { // Binary version of spatial join (two different input files) job.setClass(SpatialSite.FilterClass, SpatialJoinFilter.class, BlockFilter.class); FileInputFormat.setInputPaths(job, inFiles); if ((gindex1 != null && gindex1.isReplicated()) || (gindex2 != null && gindex2.isReplicated())) { // Need the map function with duplicate avoidance step. job.setMapperClass(RedistributeJoinMap.class); } else { // No replication in both indexes, use map function with no dup // avoidance job.setMapperClass(RedistributeJoinMapNoDupAvoidance.class); } } Shape shape = params.getShape("shape"); job.setMapOutputKeyClass(shape.getClass()); job.setMapOutputValueClass(shape.getClass()); job.setNumMapTasks(10 * Math.max(1, clusterStatus.getMaxMapTasks())); job.setNumReduceTasks(0); // No reduce needed for this task if (job.getBoolean("output", true)) job.setOutputFormat(TextOutputFormat.class); else job.setOutputFormat(NullOutputFormat.class); TextOutputFormat.setOutputPath(job, outputPath); if (!params.getBoolean("background", false)) { LOG.info("Submit job in sync mode"); RunningJob runningJob = JobClient.runJob(job); Counters counters = runningJob.getCounters(); Counter outputRecordCounter = counters.findCounter(Task.Counter.MAP_OUTPUT_RECORDS); final long resultCount = outputRecordCounter.getValue(); // Output number of running map tasks Counter mapTaskCountCounter = counters.findCounter(JobInProgress.Counter.TOTAL_LAUNCHED_MAPS); System.out.println("Number of map tasks " + mapTaskCountCounter.getValue()); // Delete output directory if not explicitly set by user if (userOutputPath == null) fs[0].delete(outputPath, true); long t2 = System.currentTimeMillis(); System.out.println("Join time " + (t2 - t1) + " millis"); return resultCount; } else { JobClient jc = new JobClient(job); LOG.info("Submit job in async mode"); lastRunningJob = jc.submitJob(job); LOG.info("Job " + lastRunningJob + " submitted successfully"); return -1; } }
From source file:edu.umn.cs.spatialHadoop.operations.Equals.java
License:Open Source License
public static <S extends Shape> long equals(Path[] inFiles, Path userOutputPath, OperationsParams params) throws IOException, InterruptedException { JobConf job = new JobConf(params, Equals.class); LOG.info("Equals journey starts ...."); FileSystem inFs = inFiles[0].getFileSystem(job); Path outputPath = userOutputPath; if (outputPath == null) { FileSystem outFs = FileSystem.get(job); do {/* ww w . ja va2s .c o m*/ outputPath = new Path(inFiles[0].getName() + ".sjmr_" + (int) (Math.random() * 1000000)); } while (outFs.exists(outputPath)); } FileSystem outFs = outputPath.getFileSystem(job); ClusterStatus clusterStatus = new JobClient(job).getClusterStatus(); job.setJobName("Equals"); job.setMapperClass(EqualsMap.class); job.setMapOutputKeyClass(IntWritable.class); job.setMapOutputValueClass(IndexedText.class); job.setNumMapTasks(5 * Math.max(1, clusterStatus.getMaxMapTasks())); job.setLong("mapred.min.split.size", Math.max(inFs.getFileStatus(inFiles[0]).getBlockSize(), inFs.getFileStatus(inFiles[1]).getBlockSize())); job.setReducerClass(EqualsReduce.class); job.setNumReduceTasks(Math.max(1, clusterStatus.getMaxReduceTasks())); job.setInputFormat(ShapeLineInputFormat.class); if (job.getBoolean("output", true)) job.setOutputFormat(TextOutputFormat.class); else job.setOutputFormat(NullOutputFormat.class); ShapeLineInputFormat.setInputPaths(job, inFiles); // Calculate and set the dimensions of the grid to use in the map phase long total_size = 0; Rectangle mbr = new Rectangle(Double.MAX_VALUE, Double.MAX_VALUE, -Double.MAX_VALUE, -Double.MAX_VALUE); for (Path file : inFiles) { FileSystem fs = file.getFileSystem(params); Rectangle file_mbr = FileMBR.fileMBR(file, params); mbr.expand(file_mbr); total_size += FileUtil.getPathSize(fs, file); } // If the largest file is globally indexed, use its partitions total_size += total_size * job.getFloat(SpatialSite.INDEXING_OVERHEAD, 0.2f); int sjmrPartitioningGridFactor = params.getInt(PartitioiningFactor, 20); int num_cells = (int) Math.max(1, total_size * sjmrPartitioningGridFactor / outFs.getDefaultBlockSize(outputPath)); LOG.info("Number of cells is configured to be " + num_cells); OperationsParams.setInactiveModeFlag(job, InactiveMode, isReduceInactive); OperationsParams.setJoiningThresholdPerOnce(job, JoiningThresholdPerOnce, joiningThresholdPerOnce); OperationsParams.setFilterOnlyModeFlag(job, isFilterOnlyMode, isFilterOnly); GridInfo gridInfo = new GridInfo(mbr.x1, mbr.y1, mbr.x2, mbr.y2); gridInfo.calculateCellDimensions(num_cells); OperationsParams.setShape(job, PartitionGrid, gridInfo); TextOutputFormat.setOutputPath(job, outputPath); if (OperationsParams.isLocal(job, inFiles)) { // Enforce local execution if explicitly set by user or for small files job.set("mapred.job.tracker", "local"); } // Start the job RunningJob runningJob = JobClient.runJob(job); Counters counters = runningJob.getCounters(); Counter outputRecordCounter = counters.findCounter(Task.Counter.REDUCE_OUTPUT_RECORDS); final long resultCount = outputRecordCounter.getValue(); return resultCount; }
From source file:edu.umn.cs.spatialHadoop.operations.Indexer.java
License:Open Source License
private static void indexLocal(Path inPath, Path outPath, OperationsParams params) throws IOException { JobConf job = new JobConf(params); String sindex = params.get("sindex"); Partitioner partitioner = createPartitioner(inPath, outPath, job, sindex); // Start reading input file Vector<InputSplit> splits = new Vector<InputSplit>(); final ShapeIterInputFormat inputFormat = new ShapeIterInputFormat(); FileSystem inFs = inPath.getFileSystem(params); FileStatus inFStatus = inFs.getFileStatus(inPath); if (inFStatus != null && !inFStatus.isDir()) { // One file, retrieve it immediately. // This is useful if the input is a hidden file which is automatically // skipped by FileInputFormat. We need to plot a hidden file for the case // of plotting partition boundaries of a spatial index splits.add(new FileSplit(inPath, 0, inFStatus.getLen(), new String[0])); } else {//from w w w .j a v a 2 s . c om ShapeIterInputFormat.addInputPath(job, inPath); for (InputSplit s : inputFormat.getSplits(job, 1)) splits.add(s); } // Copy splits to a final array to be used in parallel final FileSplit[] fsplits = splits.toArray(new FileSplit[splits.size()]); boolean replicate = job.getBoolean("replicate", false); final IndexRecordWriter<Shape> recordWriter = new IndexRecordWriter<Shape>(partitioner, replicate, sindex, outPath, params); for (FileSplit fsplit : fsplits) { RecordReader<Rectangle, Iterable<? extends Shape>> reader = inputFormat.getRecordReader(fsplit, job, null); Rectangle partitionMBR = reader.createKey(); Iterable<? extends Shape> shapes = reader.createValue(); final IntWritable partitionID = new IntWritable(); while (reader.next(partitionMBR, shapes)) { if (replicate) { // Replicate each shape to all overlapping partitions for (final Shape s : shapes) { partitioner.overlapPartitions(s, new ResultCollector<Integer>() { @Override public void collect(Integer id) { partitionID.set(id); try { recordWriter.write(partitionID, s); } catch (IOException e) { throw new RuntimeException(e); } } }); } } else { for (Shape s : shapes) { partitionID.set(partitioner.overlapPartition(s)); recordWriter.write(partitionID, s); } } } reader.close(); } recordWriter.close(null); }