private static void loadPairs(
HMapIV<ArrayListOfIntsWritable> pwsimMapping, int langID, JobConf job, Reporter reporter) {
try {
Path[] localFiles = DistributedCache.getLocalCacheFiles(job);
String pwsimFile = job.get("PwsimPairs");
for (Path localFile : localFiles) {
if (localFile.toString().contains(getFilename(pwsimFile))) {
SequenceFile.Reader reader =
new SequenceFile.Reader(FileSystem.getLocal(job), localFile, job);
PairOfInts key = (PairOfInts) reader.getKeyClass().newInstance();
IntWritable value = (IntWritable) reader.getValueClass().newInstance();
int cnt = 0;
while (reader.next(key, value)) {
int fDocno = key.getRightElement();
int eDocno = key.getLeftElement();
if ((eDocno == 6127 && fDocno == 1000000074)
|| (eDocno == 6127 && fDocno == 1000000071)) {
sLogger.info(key);
}
if (langID == CLIRUtils.E) {
if (!pwsimMapping.containsKey(eDocno)) {
pwsimMapping.put(eDocno, new ArrayListOfIntsWritable());
}
pwsimMapping
.get(eDocno)
.add(
fDocno); // we add 1000000000 to foreign docnos to distinguish them during
// pwsim algo
} else {
if (!pwsimMapping.containsKey(fDocno)) {
pwsimMapping.put(fDocno, new ArrayListOfIntsWritable());
}
pwsimMapping
.get(fDocno)
.add(
eDocno); // we add 1000000000 to foreign docnos to distinguish them during
// pwsim algo
}
cnt++;
key = (PairOfInts) reader.getKeyClass().newInstance();
value = (IntWritable) reader.getValueClass().newInstance();
}
reader.close();
sLogger.info(pwsimMapping.size() + "," + cnt + " pairs loaded from " + localFile);
}
}
} catch (Exception e) {
throw new RuntimeException(e);
}
}
public void map(
IntWritable docnoKey,
WikipediaPage p,
OutputCollector<PairOfInts, PairOfIntString> output,
Reporter reporter)
throws IOException {
int docno = docnoKey.get();
String title = p.getTitle();
String lang = p.getLanguage();
int langID = lang.equals(srcLang) ? CLIRUtils.F : CLIRUtils.E;
if (langID == CLIRUtils.F) {
docno += 1000000000;
if (samplesMap != null && !samplesMap.containsKey(docno)) {
return;
}
}
// we only load the mapping once, during the first map() call of a mapper.
// this works b/c all input kv pairs of a given mapper will have same lang id (reason
// explained above)
if (pwsimMapping.isEmpty()) {
loadPairs(pwsimMapping, langID, mJob, reporter);
sLogger.info("Mapping loaded: " + pwsimMapping.size());
}
// if no similar docs for docno, return
if (pwsimMapping.containsKey(docno)) {
similarDocnos = pwsimMapping.get(docno);
} else {
return;
}
for (int similarDocno : similarDocnos) {
if (langID == CLIRUtils.E) {
if (samplesMap != null && !samplesMap.containsKey(similarDocno)) {
continue;
}
keyOut.set(similarDocno, docno);
} else {
keyOut.set(docno, similarDocno);
}
valOut.set(langID, title);
output.collect(keyOut, valOut);
}
}
public void map(
ArrayListOfLongsWritable key,
ArrayListWritable<IntWritable> sentenceList,
OutputCollector<PairOfInts, IntWritable> output,
Reporter reporter)
throws IOException {
IntWritable s1;
IntWritable s2;
int s1line;
int s2line;
for (int i = 0; i < sentenceList.size() && i < MAXSIZE; i++) {
s1 = sentenceList.get(i);
s1line = s1.get() / nSamples;
for (int j = i + 1; j < sentenceList.size() && j < MAXSIZE; j++) {
s2 = sentenceList.get(j);
s2line = s2.get() / nSamples;
pairOut = new PairOfInts();
if (s1line == s2line) {
continue;
}
/*
if(((s1.get()/nSamples)%2 == 0 && (s2.get()/nSamples)%2 == 0) ||
((s1.get()/nSamples)%2 != 0 && (s2.get()/nSamples)%2 != 0)){
continue;
}
*/
if (s1.get() < s2.get()) {
// pairOut.set(s1.get(), s2.get());
pairOut.set(s1line, s2line);
} else {
// pairOut.set(s2.get(), s1.get());
pairOut.set(s2line, s1line);
}
output.collect(pairOut, ONE);
}
}
}
/** Runs this tool. */
@SuppressWarnings({"static-access"})
public int run(String[] args) throws Exception {
Options options = new Options();
options.addOption(
OptionBuilder.withArgName("path").hasArg().withDescription("input path").create(INDEX));
options.addOption(
OptionBuilder.withArgName("path")
.hasArg()
.withDescription("output path")
.create(COLLECTION));
CommandLine cmdline = null;
CommandLineParser parser = new GnuParser();
try {
cmdline = parser.parse(options, args);
} catch (ParseException exp) {
System.err.println("Error parsing command line: " + exp.getMessage());
System.exit(-1);
}
if (!cmdline.hasOption(INDEX) || !cmdline.hasOption(COLLECTION)) {
System.out.println("args: " + Arrays.toString(args));
HelpFormatter formatter = new HelpFormatter();
formatter.setWidth(120);
formatter.printHelp(LookupPostingsCompressed.class.getName(), options);
ToolRunner.printGenericCommandUsage(System.out);
System.exit(-1);
}
String indexPath = cmdline.getOptionValue(INDEX);
String collectionPath = cmdline.getOptionValue(COLLECTION);
if (collectionPath.endsWith(".gz")) {
System.out.println("gzipped collection is not seekable: use compressed version!");
System.exit(-1);
}
Configuration config = new Configuration();
FileSystem fs = FileSystem.get(config);
MapFile.Reader reader = new MapFile.Reader(new Path(indexPath + "/part-r-00000"), config);
FSDataInputStream collection = fs.open(new Path(collectionPath));
BufferedReader d = new BufferedReader(new InputStreamReader(collection));
Text key = new Text();
ArrayListWritable<PairOfInts> postings;
BytesWritable bytesValue = new BytesWritable();
System.out.println("Looking up postings for the term \"starcross'd\"");
key.set("starcross'd");
reader.get(key, bytesValue);
postings = deserializePosting(bytesValue);
// ArrayListWritable<PairOfVInts> postings = value;
for (PairOfInts pair : postings) {
System.out.println(pair);
collection.seek(pair.getLeftElement());
System.out.println(d.readLine());
}
bytesValue = new BytesWritable();
key.set("gold");
reader.get(key, bytesValue);
postings = deserializePosting(bytesValue);
System.out.println(
"Complete postings list for 'gold': (" + postings.size() + ", " + postings + ")");
Int2IntFrequencyDistribution goldHist = new Int2IntFrequencyDistributionEntry();
// postings = value;
for (PairOfInts pair : postings) {
goldHist.increment(pair.getRightElement());
}
System.out.println("histogram of tf values for gold");
for (PairOfInts pair : goldHist) {
System.out.println(pair.getLeftElement() + "\t" + pair.getRightElement());
}
bytesValue = new BytesWritable();
key.set("silver");
reader.get(key, bytesValue);
postings = deserializePosting(bytesValue);
System.out.println(
"Complete postings list for 'silver': (" + postings.size() + ", " + postings + ")");
Int2IntFrequencyDistribution silverHist = new Int2IntFrequencyDistributionEntry();
// postings = value;
for (PairOfInts pair : postings) {
silverHist.increment(pair.getRightElement());
}
System.out.println("histogram of tf values for silver");
for (PairOfInts pair : silverHist) {
System.out.println(pair.getLeftElement() + "\t" + pair.getRightElement());
}
bytesValue = new BytesWritable();
key.set("bronze");
Writable w = reader.get(key, bytesValue);
if (w == null) {
System.out.println("the term bronze does not appear in the collection");
}
collection.close();
reader.close();
return 0;
}
@Test
public void testIterable() {
Int2IntFrequencyDistribution fd = new Int2IntFrequencyDistributionOpen();
fd.set(1, 1);
fd.set(4, 3);
fd.set(2, 4);
fd.set(5, 7);
fd.set(6, 9);
fd.set(3, 2);
assertEquals(6, fd.getNumberOfEvents());
assertEquals(26, fd.getSumOfCounts());
SortedSet<PairOfInts> list = new TreeSet<PairOfInts>();
for (PairOfInts pair : fd) {
list.add(pair.clone());
}
assertEquals(6, list.size());
Iterator<PairOfInts> iter = list.iterator();
PairOfInts e = iter.next();
assertEquals(1, e.getLeftElement());
assertEquals(1, e.getRightElement());
e = iter.next();
assertEquals(2, e.getLeftElement());
assertEquals(4, e.getRightElement());
e = iter.next();
assertEquals(3, e.getLeftElement());
assertEquals(2, e.getRightElement());
e = iter.next();
assertEquals(4, e.getLeftElement());
assertEquals(3, e.getRightElement());
e = iter.next();
assertEquals(5, e.getLeftElement());
assertEquals(7, e.getRightElement());
e = iter.next();
assertEquals(6, e.getLeftElement());
assertEquals(9, e.getRightElement());
}
@SuppressWarnings("static-access")
@Override
public int run(String[] args) throws Exception {
Options options = new Options();
options.addOption(
OptionBuilder.withArgName("path")
.hasArg()
.withDescription("output path")
.create(matchOutput));
options.addOption(
OptionBuilder.withArgName("path")
.hasArg()
.withDescription("output path")
.create(nomatchOutput));
options.addOption(
OptionBuilder.withArgName("integer")
.hasArg()
.withDescription("number of samples")
.create(nSamplesOption));
CommandLine cmdline;
CommandLineParser parser = new GnuParser();
try {
cmdline = parser.parse(options, args);
} catch (ParseException exp) {
System.err.println("Error parsing command line: " + exp.getMessage());
return -1;
}
if (!cmdline.hasOption(matchOutput)
|| !cmdline.hasOption(nomatchOutput)
|| !cmdline.hasOption(nSamplesOption)) {
HelpFormatter formatter = new HelpFormatter();
formatter.setWidth(120);
formatter.printHelp(this.getClass().getName(), options);
ToolRunner.printGenericCommandUsage(System.out);
return -1;
}
String matchOutputPath = cmdline.getOptionValue(matchOutput);
String nomatchOutputPath = cmdline.getOptionValue(nomatchOutput);
String nSamplesIn = cmdline.getOptionValue(nSamplesOption);
LOG.info("Tool name: " + this.getClass().getName());
// LOG.info(" - input file: " + inputPath);
// LOG.info(" - output file: " + outputPath);
JobConf conf = new JobConf(getConf(), JaccardCompare.class);
conf.setJobName(String.format("JaccardCompare"));
// FileInputFormat.setInputPaths(conf, new Path(inputPath));
// FileOutputFormat.setOutputPath(conf, new Path(outputPath));
int nSentences = 1000;
int nSamples = Integer.parseInt(nSamplesIn);
try {
File matchFile = new File(matchOutputPath);
File nomatchFile = new File(nomatchOutputPath);
FileOutputStream fosM = null, fosNM = null;
BufferedWriter dosM = null, dosNM = null;
fosM = new FileOutputStream(matchFile);
fosNM = new FileOutputStream(nomatchFile);
dosM = new BufferedWriter(new OutputStreamWriter(fosM));
dosNM = new BufferedWriter(new OutputStreamWriter(fosNM));
MapFile.Reader id2sentenceReader =
new MapFile.Reader(new Path("id2sentence.map/part-00000"), conf);
HashMap<Integer, ArrayListWritable<Text>> id2sentence =
new HashMap<Integer, ArrayListWritable<Text>>();
IntWritable key = new IntWritable();
ArrayListWritable<Text> val = new ArrayListWritable<Text>();
while (id2sentenceReader.next(key, val)) {
id2sentence.put(key.get(), val);
val = new ArrayListWritable<Text>();
}
MapFile.Reader sentence2translationReader =
new MapFile.Reader(new Path("sentence2translation.map/part-00000"), conf);
HashMap<Integer, ArrayListOfIntsWritable> sentence2translation =
new HashMap<Integer, ArrayListOfIntsWritable>();
IntWritable key2 = new IntWritable();
ArrayListOfIntsWritable val2 = new ArrayListOfIntsWritable();
while (sentence2translationReader.next(key2, val2)) {
sentence2translation.put(key2.get(), val2);
val2 = new ArrayListOfIntsWritable();
}
MapFile.Reader sentencematchReader =
new MapFile.Reader(new Path("sentencematchpairs.map/part-00000"), conf);
HashSet<PairOfInts> sentencematchpairs = new HashSet<PairOfInts>();
PairOfInts key3 = new PairOfInts();
IntWritable val3 = new IntWritable();
while (sentencematchReader.next(key3, val3)) {
sentencematchpairs.add(key3);
key3 = new PairOfInts();
}
System.out.println("Done reading");
PairOfInts p = new PairOfInts();
IntWritable match;
IntWritable eLineNum = new IntWritable();
IntWritable eLineId = new IntWritable();
ArrayListWritable<Text> eSentence = new ArrayListWritable<Text>();
for (int i = 0; i < nSentences; i++) {
if (i % 100 == 0) System.out.println("eLine " + i);
// eLineNum.set(2*i);
ArrayListOfIntsWritable transIdList = sentence2translation.get(2 * i);
// ArrayListOfIntsWritable transIdList = new ArrayListOfIntsWritable();
// sentence2translationReader.get(eLineNum, transIdList);
// System.out.println("transIdList " + transIdList);
for (int j = 0; j < nSentences; j++) {
// System.out.println("fLine " + j);
ArrayListWritable<Text> fSentence = id2sentence.get((2 * j + 1) * nSamples);
// ArrayListWritable<Text> fSentence = new ArrayListWritable<Text>();
// IntWritable fLineId = new IntWritable();
// fLineId.set((2*j+1)*nSamples);
// id2sentenceReader.get(fLineId, fSentence);
// System.out.println("fLineId " + (2*j+1)*nSamples + " FSentence " + fSentence);
float jsimMax = -1.0f;
float jsimAvg = 0.0f;
for (int id : transIdList) {
eSentence = id2sentence.get(id);
// eLineId.set(id);
// id2sentenceReader.get(eLineId, eSentence);
float jsim = JaccardSim.jaccardSim(eSentence, fSentence);
// System.out.println("\teSentence " + eSentence + " " + jsim);
jsimAvg += jsim;
if (jsim > jsimMax) {
jsimMax = jsim;
}
}
jsimAvg = jsimAvg / transIdList.size();
if (2 * i < 2 * j + 1) {
p.set(2 * i, 2 * j + 1);
} else {
p.set(2 * j + 1, 2 * i);
}
// match = new IntWritable();
// match = (IntWritable) sentencematchReader.get(p, match);
// if(match != null){
if (sentencematchpairs.contains(p)) {
if (jsimMax < .5) {
System.out.println("Low match: ");
System.out.println("\teSentence: " + i + " " + eSentence);
System.out.println("\tfSentence: " + j + " " + fSentence);
}
// System.out.println("match");
dosM.write(Float.toString(jsimMax));
// dosM.write(Float.toString(jsimAvg));
dosM.write("\n");
} else {
// System.out.println("no match");
dosNM.write(Float.toString(jsimMax));
// dosNM.write(Float.toString(jsimAvg));
dosNM.write("\n");
}
}
}
sentencematchReader.close();
sentence2translationReader.close();
id2sentenceReader.close();
dosM.close();
dosNM.close();
} catch (IOException e2) {
// TODO Auto-generated catch block
e2.printStackTrace();
}
// Delete the output directory if it exists already.
// Path outputDir = new Path(outputPath);
// FileSystem.get(conf).delete(outputDir, true);
// JobClient.runJob(conf);
return 0;
}