Java tutorial
/* Copyright 2014 Twitter, Inc. Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. */ package com.twitter.algebra.matrix.multiply; import java.io.IOException; import java.util.Iterator; import java.util.List; import java.util.Map; import org.apache.hadoop.conf.Configuration; import org.apache.hadoop.fs.FileSystem; import org.apache.hadoop.fs.Path; import org.apache.hadoop.io.IntWritable; import org.apache.hadoop.mapreduce.Job; import org.apache.hadoop.mapreduce.Mapper; import org.apache.hadoop.mapreduce.Reducer; import org.apache.hadoop.mapreduce.lib.input.SequenceFileInputFormat; import org.apache.hadoop.mapreduce.lib.join.CompositeInputFormat; import org.apache.hadoop.mapreduce.lib.join.TupleWritable; import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat; import org.apache.mahout.common.AbstractJob; import org.apache.mahout.math.CardinalityException; import org.apache.mahout.math.RandomAccessSparseVector; import org.apache.mahout.math.SequentialAccessSparseVector; import org.apache.mahout.math.Vector; import org.apache.mahout.math.VectorWritable; import org.apache.mahout.math.function.Functions; import org.apache.mahout.math.hadoop.DistributedRowMatrix; import org.slf4j.Logger; import org.slf4j.LoggerFactory; import com.twitter.algebra.matrix.format.MatrixOutputFormat; /** * Perform A x B matrix multiplication * * Approach: Outer-join (borrowed from Mahout's {@link MatrixMultiplicationJob}) * * Number of jobs: 1 * * Assumption: (1) Transpose At is already available, (2) At and B have the * exactly the same partitioning (number of partitions and entries inside each * partition), (3) the entries inside each partition are sorted, (4) At and B * have different number of columns (to distinguish them in map side join). * * Design: Hadoop put the partitions with the same index together and iterates * over the entries of both partition at the same time (mapside join). Each mapper * perform Ati x Bi (row i of At and row i of B) multiplication and generates * partial matrix Ci. The reducers sum up partial Ci matrices to get C = A x B. */ public class AtBOuterStaticMapsideJoinJob extends AbstractJob { private static final Logger log = LoggerFactory.getLogger(AtBOuterStaticMapsideJoinJob.class); static final String OUT_CARD = "output.vector.cardinality"; @Override public int run(String[] strings) throws Exception { addOutputOption(); addOption("numColsB", "ncb", "Number of columns of the second input matrix", true); addOption("inputPathA", "ia", "Path to the first input matrix", true); addOption("inputPathB", "ib", "Path to the second input matrix", true); Map<String, List<String>> argMap = parseArguments(strings); if (argMap == null) { return -1; } run(getConf(), new Path(getOption("inputPathA")), new Path(getOption("inputPathB")), getOutputPath(), Integer.parseInt(getOption("numColsB"))); return 0; } public void run(Configuration conf, Path atPath, Path bPath, Path outPath, int outCardinality) throws IOException, InterruptedException, ClassNotFoundException { conf.setInt(OUT_CARD, outCardinality); @SuppressWarnings("deprecation") Job job = new Job(conf); job.setJobName(AtBOuterStaticMapsideJoinJob.class.getSimpleName()); job.setJarByClass(AtBOuterStaticMapsideJoinJob.class); FileSystem fs = FileSystem.get(atPath.toUri(), conf); atPath = fs.makeQualified(atPath); bPath = fs.makeQualified(bPath); job.setInputFormatClass(CompositeInputFormat.class); //mapside join expression job.getConfiguration().set(CompositeInputFormat.JOIN_EXPR, CompositeInputFormat.compose("inner", SequenceFileInputFormat.class, atPath, bPath)); job.setOutputFormatClass(MatrixOutputFormat.class); outPath = fs.makeQualified(outPath); FileOutputFormat.setOutputPath(job, outPath); job.setMapperClass(MyMapper.class); job.setMapOutputKeyClass(IntWritable.class); job.setMapOutputValueClass(VectorWritable.class); job.setCombinerClass(MyReducer.class); int numReducers = conf.getInt("algebra.reduceslots.multiply", 10); job.setNumReduceTasks(numReducers); job.setReducerClass(MyReducer.class); job.setOutputKeyClass(IntWritable.class); job.setOutputValueClass(VectorWritable.class); job.submit(); boolean res = job.waitForCompletion(true); if (!res) throw new IOException("Job failed"); } public static DistributedRowMatrix run(Configuration conf, DistributedRowMatrix A, DistributedRowMatrix B, String label) throws IOException, InterruptedException, ClassNotFoundException { log.info("running " + AtBOuterStaticMapsideJoinJob.class.getName()); if (A.numRows() != B.numRows()) { throw new CardinalityException(A.numRows(), B.numRows()); } Path outPath = new Path(A.getOutputTempPath(), label); FileSystem fs = FileSystem.get(outPath.toUri(), conf); AtBOuterStaticMapsideJoinJob job = new AtBOuterStaticMapsideJoinJob(); if (!fs.exists(outPath)) { job.run(conf, A.getRowPath(), B.getRowPath(), outPath, B.numCols()); } else { log.warn("----------- Skip already exists: " + outPath); } DistributedRowMatrix distRes = new DistributedRowMatrix(outPath, A.getOutputTempPath(), A.numCols(), B.numCols()); distRes.setConf(conf); return distRes; } public static class MyMapper extends Mapper<IntWritable, TupleWritable, IntWritable, VectorWritable> { private int outCardinality; private final IntWritable row = new IntWritable(); @Override public void setup(Context context) throws IOException { outCardinality = context.getConfiguration().getInt(OUT_CARD, Integer.MAX_VALUE); } @Override public void map(IntWritable index, TupleWritable v, Context context) throws IOException, InterruptedException { boolean firstIsOutFrag = ((VectorWritable) v.get(0)).get().size() == outCardinality; Vector outFrag = firstIsOutFrag ? ((VectorWritable) v.get(0)).get() : ((VectorWritable) v.get(1)).get(); Vector multiplier = firstIsOutFrag ? ((VectorWritable) v.get(1)).get() : ((VectorWritable) v.get(0)).get(); VectorWritable outVector = new VectorWritable(); Iterator<Vector.Element> it = multiplier.nonZeroes().iterator(); while (it.hasNext()) { Vector.Element e = it.next(); row.set(e.index()); outVector.set(outFrag.times(e.get())); context.write(row, outVector); } } } public static class MyReducer extends Reducer<IntWritable, VectorWritable, IntWritable, VectorWritable> { private VectorWritable outvw = new VectorWritable(); @Override public void reduce(IntWritable rowNum, Iterable<VectorWritable> values, Context context) throws IOException, InterruptedException { Iterator<VectorWritable> it = values.iterator(); if (!it.hasNext()) return; Vector accumulator = new RandomAccessSparseVector(it.next().get()); while (it.hasNext()) { Vector row = it.next().get(); accumulator.assign(row, Functions.PLUS); } outvw.set(new SequentialAccessSparseVector(accumulator)); context.write(rowNum, outvw); } } }