@SuppressWarnings("unchecked")
@Override
public List<String> getClusterByCarrot2(String query) {
// TODO Auto-generated method stub
List<String> strs = new ArrayList<String>();
final Controller controller = ControllerFactory.createCachingPooling(IDocumentSource.class);
final List<org.carrot2.core.Document> documents = Lists.newArrayList();
try {
q = getParser().parse(QueryParserUtil.escape(query));
docs = getIndexSearcher().search(q, Integer.MAX_VALUE);
hits = docs.scoreDocs;
for (int i = 0; i < hits.length; i++) {
Document doc = getIndexSearcher().doc(hits[i].doc);
documents.add(
new org.carrot2.core.Document(
doc.get(CONTENTS_FIELD), doc.get(TITLE_FIELD), doc.get(USER_FIELD)));
}
final ProcessingResult byTopicClusters =
controller.process(documents, query, LingoClusteringAlgorithm.class);
final List<Cluster> clustersByTopic = byTopicClusters.getClusters();
final ProcessingResult byDomainClusters =
controller.process(documents, query, ByUrlClusteringAlgorithm.class);
final List<Cluster> clustersByDomain = byDomainClusters.getClusters();
for (Cluster c : clustersByDomain) {
strs.add(c.getLabel());
}
for (Cluster c : clustersByTopic) {
strs.add(c.getLabel());
}
} catch (Exception ex) {
}
return strs;
}
@SuppressWarnings("unchecked")
@Override
public List<String> getClusterByCarrotVersion2(String query) {
// TODO Auto-generated method stub
List<String> strs = new ArrayList<String>();
final Controller controller = ControllerFactory.createPooling();
final Map<String, Object> luceneGlobalAttributes = new HashMap<String, Object>();
LuceneDocumentSourceDescriptor.attributeBuilder(luceneGlobalAttributes).directory(directory);
SimpleFieldMapperDescriptor.attributeBuilder(luceneGlobalAttributes)
.titleField(TITLE_FIELD)
.contentField(CONTENTS_FIELD)
.searchFields(Arrays.asList(new String[] {TITLE_FIELD, CONTENTS_FIELD}));
controller.init(
new HashMap<String, Object>(),
new ProcessingComponentConfiguration(
LuceneDocumentSource.class, "lucene", luceneGlobalAttributes));
final Map<String, Object> processingAttributes = Maps.newHashMap();
CommonAttributesDescriptor.attributeBuilder(processingAttributes).query(query);
ProcessingResult process =
controller.process(
processingAttributes, "lucene", LingoClusteringAlgorithm.class.getName());
for (Cluster c : process.getClusters()) {
strs.add(c.getLabel() + " >>>> " + c.getAllDocuments().size());
}
return strs;
}
@UsesExternalServices
@Test
public void testRequestIndependence() {
@SuppressWarnings("unchecked")
final Controller controller =
ControllerFactory.createCachingPooling(org.carrot2.core.IDocumentSource.class);
closeAfterTest(controller);
final Map<String, Object> attrs = Maps.newHashMap();
CommonAttributesDescriptor.attributeBuilder(attrs).results(50).query("data mining");
controller.process(
attrs, org.carrot2.webapp.source.WebDocumentSource.class, LingoClusteringAlgorithm.class);
attrs.clear();
CommonAttributesDescriptor.attributeBuilder(attrs)
.results(50)
.query(WebDocumentSource.QUERY_FAILURE);
try {
controller.process(
attrs, org.carrot2.webapp.source.WebDocumentSource.class, LingoClusteringAlgorithm.class);
fail();
} catch (ProcessingException e) {
assertThat(e.getCause().getMessage()).contains("Synthetic failure");
}
}
public static void main(String[] args) throws IOException {
/*
* We will use the CachingController for this example. Running
* LuceneDocumentSource within the CachingController will let us open the index
* once per component initialization and not once per query, which would be the
* case with SimpleController. We will also use this opportunity to show how
* component-specific attribute values can be passed during CachingComponent
* initialization.
*/
/*
* Create a caching controller that will reuse processing component instances, but
* will not perform any caching of results produced by components. We will leave
* caching of documents from Lucene index to Lucene and the operating system
* caches.
*/
final Controller controller = ControllerFactory.createPooling();
/*
* Prepare a map with component-specific attributes. Here, this map will contain
* the index location and names of fields to be used to fetch document title and
* summary.
*/
final Map<String, Object> luceneGlobalAttributes = new HashMap<String, Object>();
String indexPath = "put your index path here or pass as the first argument";
if (args.length == 1) {
indexPath = args[0];
}
LuceneDocumentSourceDescriptor.attributeBuilder(luceneGlobalAttributes)
.directory(FSDirectory.open(new File(indexPath)));
/*
* Specify fields providing data inside your Lucene index.
*/
SimpleFieldMapperDescriptor.attributeBuilder(luceneGlobalAttributes)
.titleField("title")
.contentField("snippet")
.searchFields(Arrays.asList(new String[] {"titleField", "fullContent"}));
/*
* Initialize the controller passing the above attributes as component-specific
* for Lucene. The global attributes map will be empty. Note that we've provided
* an identifier for our specially-configured Lucene component, we'll need to use
* this identifier when performing processing.
*/
controller.init(
new HashMap<String, Object>(),
new ProcessingComponentConfiguration(
LuceneDocumentSource.class, "lucene", luceneGlobalAttributes));
/*
* Perform processing.
*/
String query = "mining";
final Map<String, Object> processingAttributes = Maps.newHashMap();
CommonAttributesDescriptor.attributeBuilder(processingAttributes).query(query);
/*
* We need to refer to the Lucene component by its identifier we set during
* initialization. As we've not assigned any identifier to the
* LingoClusteringAlgorithm we want to use, we can its fully qualified class name.
*/
ProcessingResult process =
controller.process(
processingAttributes, "lucene", LingoClusteringAlgorithm.class.getName());
ConsoleFormatter.displayResults(process);
}
/** Entry point. */
public static void main(String[] args) throws IOException {
/*
* We will use the CachingController for this example. Running
* LuceneDocumentSource within the CachingController will let us open the index
* once per component initialization and not once per query, which would be the
* case with SimpleController. We will also use this opportunity to show how
* component-specific attribute values can be passed during CachingComponent
* initialization.
*/
/*
* Create a caching controller that will reuse processing component instances, but
* will not perform any caching of results produced by components. We will leave
* caching of documents from Lucene index to Lucene and the operating system
* caches.
*/
final Controller controller = ControllerFactory.createPooling();
/*
* Prepare a map with component-specific attributes. Here, this map will contain
* the index location and names of fields to be used to fetch document title and
* summary.
*/
final Map<String, Object> luceneGlobalAttributes = new HashMap<String, Object>();
String indexPath = "put your index path here or pass as the first argument";
if (args.length == 1) {
indexPath = args[0];
}
// Sanity check.
if (!new File(indexPath).isDirectory()) {
System.err.println("Index directory does not exist: " + indexPath);
return;
}
LuceneDocumentSourceDescriptor.attributeBuilder(luceneGlobalAttributes)
.directory(FSDirectory.open(new File(indexPath)));
/*
* In ClusteringDataFromLucene we used a simple configuration of
* LuceneDocumentSource whereby we only provided the names of Lucene fields to be
* used for titles and summaries. If more advanced mapping of Lucene documents is
* required, you can implement your own version of IFieldMapper as below.
*
* Note that we could also provide here an instance of the mapper rather than
* its class. The differences are summarized below:
*
* > Class: Class has to have a no-parameter constructor. Instances of the
* class will not be shared between processing threads, which means the
* implementation does not have to be thread-safe. Recommended in most
* situations unless the instances are expensive to create.
*
* > Instance: The provided instance will be shared across processing threads,
* which means the implementation MUST be thread-safe.
*/
LuceneDocumentSourceDescriptor.attributeBuilder(luceneGlobalAttributes)
.fieldMapper(new CustomFieldMapper());
/*
* The Analyzer used by Lucene while searching can also be provided via factory
* because it does not have a parameterless constructor.
*/
LuceneDocumentSourceDescriptor.attributeBuilder(luceneGlobalAttributes)
.analyzer(StandardAnalyzerFactory.class);
/*
* Initialize the controller passing the above attributes as component-specific
* for Lucene. The global attributes map will be empty. Note that we've provided
* an identifier for our specially-configured Lucene component, we'll need to use
* this identifier when performing processing.
*/
controller.init(
new HashMap<String, Object>(),
new ProcessingComponentConfiguration(
LuceneDocumentSource.class, "lucene", luceneGlobalAttributes));
/*
* Perform processing.
*/
final String query = "mining";
final Map<String, Object> processingAttributes = Maps.newHashMap();
CommonAttributesDescriptor.attributeBuilder(processingAttributes).query(query);
/*
* We need to refer to the Lucene component by its identifier we set during
* initialization. As we've not assigned any identifier to the
* LingoClusteringAlgorithm we want to use, we can its fully qualified class name.
*/
ProcessingResult process =
controller.process(
processingAttributes, "lucene", LingoClusteringAlgorithm.class.getName());
ConsoleFormatter.displayResults(process);
}
@Test
public void checkBasicAuthAccess() throws Throwable {
final Server server = new Server();
final SelectChannelConnector connector = new SelectChannelConnector();
connector.setPort(/* any */ 0);
connector.setReuseAddress(false);
connector.setSoLingerTime(0);
server.addConnector(connector);
HashLoginService loginService = new HashLoginService();
loginService.putUser("username", new Password("userpass"), new String[] {"role1", "role2"});
final CountDownLatch latch = new CountDownLatch(1);
WebAppContext wac = new WebAppContext();
wac.getSecurityHandler().setLoginService(loginService);
wac.setContextPath("/");
connector.addLifeCycleListener(
new ListenerAdapter() {
public void lifeCycleStarted(LifeCycle lc) {
System.out.println("Started on port: " + connector.getLocalPort());
latch.countDown();
}
public void lifeCycleFailure(LifeCycle lc, Throwable t) {
System.out.println("Failure: " + t);
latch.countDown();
}
});
wac.setParentLoaderPriority(true);
URL resource = getClass().getResource("/auth/basic/kaczynski.xml");
assertThat(resource.toURI().getScheme()).isEqualTo("file");
File webapp = new File(resource.toURI());
webapp = webapp.getParentFile(); // /auth/basic
webapp = webapp.getParentFile(); // /auth
wac.setWar(webapp.getAbsolutePath());
wac.setClassLoader(Thread.currentThread().getContextClassLoader());
server.setHandler(wac);
server.setStopAtShutdown(true);
try {
server.start();
latch.await();
System.setProperty(HttpAuthHub.USERNAME_PROPERTY, "username");
System.setProperty(HttpAuthHub.PASSWORD_PROPERTY, "userpass");
Controller c = ControllerFactory.createSimple();
try {
Map<String, Object> attrs = new HashMap<String, Object>();
XmlDocumentSourceDescriptor.attributeBuilder(attrs)
.xml(
new URLResourceWithParams(
new URL(
"http://localhost:" + connector.getLocalPort() + "/basic/kaczynski.xml")));
ProcessingResult r = c.process(attrs, XmlDocumentSource.class);
assertThat(r.getDocuments()).hasSize(50);
} finally {
c.dispose();
}
} finally {
server.stop();
}
}
@SuppressForbidden(reason = "C2 integration (File API)")
@Override
protected void doStart() throws ElasticsearchException {
try {
Settings.Builder builder = Settings.builder();
Path pluginConfigPath = environment.configFile().resolve(ClusteringPlugin.PLUGIN_NAME);
if (!Files.isDirectory(pluginConfigPath)) {
Path srcConfig = Paths.get("src/main/config");
if (Files.isDirectory(srcConfig)) {
// Allow running from within the IDE.
pluginConfigPath = srcConfig;
} else {
throw new ElasticsearchException("Config folder missing: " + pluginConfigPath);
}
} else {
logger.info("Configuration files at: {}", pluginConfigPath.toAbsolutePath());
}
for (String configName :
new String[] {"config.yml", "config.yaml", "config.json", "config.properties"}) {
try {
Path resolved = pluginConfigPath.resolve(configName);
if (resolved != null && Files.exists(resolved)) {
builder.loadFromPath(resolved);
}
} catch (NoClassDefFoundError e) {
logger.warn("Could not parse: {}", e, configName);
}
}
Settings c2Settings = builder.build();
// Parse suite descriptors with loggers turned off (shut them up a bit).
final Path suitePath = pluginConfigPath.resolve(c2Settings.get(DEFAULT_SUITE_PROPERTY_NAME));
if (!Files.isRegularFile(suitePath)) {
throw new ElasticsearchException(
"Could not find algorithm suite: " + suitePath.toAbsolutePath().normalize());
}
final ResourceLookup suiteLookup =
new ResourceLookup(new DirLocator(suitePath.getParent().toFile()));
final IResource suiteResource = suiteLookup.getFirst(suitePath.getFileName().toString());
final List<String> failed = Lists.newArrayList();
final ProcessingComponentSuite suite =
LoggerUtils.quietCall(
new Callable<ProcessingComponentSuite>() {
public ProcessingComponentSuite call() throws Exception {
ProcessingComponentSuite suite =
ProcessingComponentSuite.deserialize(suiteResource, suiteLookup);
for (ProcessingComponentDescriptor desc : suite.removeUnavailableComponents()) {
failed.add(desc.getId());
if (isNoClassDefFound(desc.getInitializationFailure())) {
logger.debug("Algorithm not available on classpath: {}", desc.getId());
} else {
logger.warn(
"Algorithm initialization failed: {}",
desc.getInitializationFailure(),
desc.getId());
}
}
return suite;
}
},
Logger.getLogger(ProcessingComponentDescriptor.class),
Logger.getLogger(ReflectionUtils.class));
algorithms = Lists.newArrayList();
for (ProcessingComponentDescriptor descriptor : suite.getAlgorithms()) {
algorithms.add(descriptor.getId());
}
algorithms = Collections.unmodifiableList(algorithms);
if (!algorithms.isEmpty()) {
logger.info("Available clustering components: {}", Joiner.on(", ").join(algorithms));
}
if (!failed.isEmpty()) {
logger.info("Unavailable clustering components: {}", Joiner.on(", ").join(failed));
}
final Path resourcesPath =
pluginConfigPath
.resolve(c2Settings.get(DEFAULT_RESOURCES_PROPERTY_NAME, "."))
.toAbsolutePath()
.normalize();
logger.info("Lexical resources dir: {}", resourcesPath);
final ResourceLookup resourceLookup =
new ResourceLookup(
new DirLocator(resourcesPath.toFile()),
new ClassLoaderLocator(ControllerSingleton.class.getClassLoader()));
// Change the default resource lookup to include the configured location.
Map<String, Object> c2SettingsAsMap = Maps.newHashMap();
DefaultLexicalDataFactoryDescriptor.attributeBuilder(c2SettingsAsMap)
.resourceLookup(resourceLookup);
c2SettingsAsMap.putAll(c2Settings.getAsMap());
// Set up the license for Lingo3G, if it's available.
Path lingo3gLicense = scanForLingo3GLicense(environment, pluginConfigPath);
if (lingo3gLicense != null && Files.isReadable(lingo3gLicense)) {
c2SettingsAsMap.put("license", new FileResource(lingo3gLicense.toFile()));
} else if (algorithms.contains("lingo3g")) {
logger.warn(
"Lingo3G is on classpath, but no licenses have been found. Check out the documentation.");
}
// Create component pool.
Integer poolSize = c2Settings.getAsInt(DEFAULT_COMPONENT_SIZE_PROPERTY_NAME, 0);
if (poolSize > 0) {
controller = ControllerFactory.createPooling(poolSize);
} else {
controller = ControllerFactory.createPooling();
}
controller.init(c2SettingsAsMap, suite.getComponentConfigurations());
} catch (Exception e) {
throw new ElasticsearchException("Could not start Carrot2 controller.", e);
}
if (algorithms == null || algorithms.isEmpty()) {
throw new ElasticsearchException(
"No registered/ available clustering algorithms? Check the logs, it's odd.");
}
}
