Search in sources :

Example 1 with MultiHandler

use of org.apache.stanbol.enhancer.engines.tika.handler.MultiHandler in project stanbol by apache.

the class TikaEngine method computeEnhancements.

@Override
public void computeEnhancements(ContentItem ci) throws EngineException {
    MediaTypeAndStream mtas = extractMediaType(ci);
    if (mtas.mediaType == null) {
        //unable to parse and detect content type
        return;
    }
    MediaType plainMediaType = mtas.mediaType.getBaseType();
    if (plainMediaType.equals(MediaType.TEXT_PLAIN)) {
        //we need not to process plain text!
        return;
    }
    final ParseContext context = new ParseContext();
    context.set(Parser.class, parser);
    Set<MediaType> supproted = parser.getSupportedTypes(context);
    if (supproted.contains(plainMediaType)) {
        final InputStream in;
        if (mtas.in == null) {
            in = ci.getStream();
        } else {
            in = mtas.in;
        }
        final Metadata metadata = new Metadata();
        //set the already parsed contentType
        metadata.set(Metadata.CONTENT_TYPE, mtas.mediaType.toString());
        //also explicitly set the charset as contentEncoding
        String charset = mtas.mediaType.getParameters().get("charset");
        if (charset != null) {
            metadata.set(Metadata.CONTENT_ENCODING, charset);
        }
        ContentSink plainTextSink;
        try {
            plainTextSink = ciFactory.createContentSink(TEXT_PLAIN + "; charset=" + UTF8.name());
        } catch (IOException e) {
            //close the input stream
            IOUtils.closeQuietly(in);
            throw new EngineException("Error while initialising Blob for" + "writing the text/plain version of the parsed content", e);
        }
        final Writer plainTextWriter = new OutputStreamWriter(plainTextSink.getOutputStream(), UTF8);
        final ContentHandler textHandler = new //only the Body
        BodyContentHandler(//skip ignoreable
        new PlainTextHandler(plainTextWriter, false, skipLinebreaks));
        final ToXMLContentHandler xhtmlHandler;
        final ContentHandler mainHandler;
        ContentSink xhtmlSink = null;
        try {
            if (!plainMediaType.equals(XHTML)) {
                //do not parse XHTML from XHTML
                try {
                    xhtmlSink = ciFactory.createContentSink(XHTML + "; charset=" + UTF8.name());
                } catch (IOException e) {
                    throw new EngineException("Error while initialising Blob for" + "writing the application/xhtml+xml version of the parsed content", e);
                }
                try {
                    xhtmlHandler = new ToXMLContentHandler(xhtmlSink.getOutputStream(), UTF8.name());
                } catch (UnsupportedEncodingException e) {
                    throw new EngineException("This system does not support the encoding " + UTF8, e);
                }
                mainHandler = new MultiHandler(textHandler, xhtmlHandler);
            } else {
                mainHandler = textHandler;
                xhtmlHandler = null;
                xhtmlSink = null;
            }
            try {
                AccessController.doPrivileged(new PrivilegedExceptionAction<Object>() {

                    public Object run() throws IOException, SAXException, TikaException {
                        /* 
                             * We need to replace the context Classloader with the Bundle ClassLoader
                             * to ensure that Singleton instances of XML frameworks (such as node4j) 
                             * do not leak into the OSGI environment.
                             * 
                             * Most Java XML libs prefer to load implementations by using the 
                             * {@link Thread#getContextClassLoader()}. However OSGI has no control over
                             * this {@link ClassLoader}. Because of that there can be situations where
                             * Interfaces are loaded via the Bundle Classloader and the implementations
                             * are taken from the context Classloader. What can cause 
                             * {@link ClassCastException}, {@link ExceptionInInitializerError}s, ...
                             * 
                             * Setting the context Classloader to the Bundle classloader helps to avoid
                             * those situations.
                             */
                        ClassLoader contextClassLoader = updateContextClassLoader();
                        try {
                            parser.parse(in, mainHandler, metadata, context);
                        } finally {
                            //reset the previous context ClassLoader
                            Thread.currentThread().setContextClassLoader(contextClassLoader);
                        }
                        return null;
                    }
                });
            } catch (PrivilegedActionException pae) {
                Exception e = pae.getException();
                if (e instanceof IOException || e instanceof SAXException || e instanceof TikaException) {
                    throw new EngineException("Unable to convert ContentItem " + ci.getUri() + " with mimeType '" + ci.getMimeType() + "' to " + "plain text!", e);
                } else {
                    //runtime exception
                    throw RuntimeException.class.cast(e);
                }
            }
        } finally {
            //ensure that the writers are closed correctly
            IOUtils.closeQuietly(in);
            IOUtils.closeQuietly(plainTextWriter);
            if (xhtmlSink != null) {
                IOUtils.closeQuietly(xhtmlSink.getOutputStream());
            }
        }
        String random = randomUUID().toString();
        IRI textBlobUri = new IRI("urn:tika:text:" + random);
        ci.addPart(textBlobUri, plainTextSink.getBlob());
        if (xhtmlHandler != null) {
            IRI xhtmlBlobUri = new IRI("urn:tika:xhtml:" + random);
            ci.addPart(xhtmlBlobUri, xhtmlSink.getBlob());
        }
        //add the extracted metadata
        if (log.isInfoEnabled()) {
            for (String name : metadata.names()) {
                log.info("{}: {}", name, Arrays.toString(metadata.getValues(name)));
            }
        }
        ci.getLock().writeLock().lock();
        try {
            Graph graph = ci.getMetadata();
            IRI id = ci.getUri();
            Set<String> mapped = ontologyMappings.apply(graph, id, metadata);
            if (includeUnmappedProperties) {
                Set<String> unmapped = new HashSet<String>(Arrays.asList(metadata.names()));
                unmapped.removeAll(mapped);
                for (String name : unmapped) {
                    if (name.indexOf(':') >= 0 || includeAllUnmappedProperties) {
                        //only mapped
                        IRI prop = new IRI(new StringBuilder(TIKA_URN_PREFIX).append(name).toString());
                        for (String value : metadata.getValues(name)) {
                            //TODO: without the Property for the name we have no datatype
                            //      information ... so we add PlainLiterals for now
                            graph.add(new TripleImpl(id, prop, new PlainLiteralImpl(value)));
                        }
                    }
                }
            }
        } finally {
            ci.getLock().writeLock().unlock();
        }
    }
//else not supported format
}
Also used : BodyContentHandler(org.apache.tika.sax.BodyContentHandler) IRI(org.apache.clerezza.commons.rdf.IRI) Metadata(org.apache.tika.metadata.Metadata) EngineException(org.apache.stanbol.enhancer.servicesapi.EngineException) ToXMLContentHandler(org.apache.tika.sax.ToXMLContentHandler) BodyContentHandler(org.apache.tika.sax.BodyContentHandler) ContentHandler(org.xml.sax.ContentHandler) SAXException(org.xml.sax.SAXException) MediaType(org.apache.tika.mime.MediaType) TripleImpl(org.apache.clerezza.commons.rdf.impl.utils.TripleImpl) HashSet(java.util.HashSet) MultiHandler(org.apache.stanbol.enhancer.engines.tika.handler.MultiHandler) ToXMLContentHandler(org.apache.tika.sax.ToXMLContentHandler) TikaException(org.apache.tika.exception.TikaException) PrivilegedActionException(java.security.PrivilegedActionException) PlainLiteralImpl(org.apache.clerezza.commons.rdf.impl.utils.PlainLiteralImpl) BufferedInputStream(java.io.BufferedInputStream) InputStream(java.io.InputStream) UnsupportedEncodingException(java.io.UnsupportedEncodingException) IOException(java.io.IOException) EngineException(org.apache.stanbol.enhancer.servicesapi.EngineException) ConfigurationException(org.osgi.service.cm.ConfigurationException) SAXException(org.xml.sax.SAXException) UnsupportedEncodingException(java.io.UnsupportedEncodingException) TikaException(org.apache.tika.exception.TikaException) PrivilegedActionException(java.security.PrivilegedActionException) IOException(java.io.IOException) Graph(org.apache.clerezza.commons.rdf.Graph) PlainTextHandler(org.apache.stanbol.enhancer.engines.tika.handler.PlainTextHandler) ParseContext(org.apache.tika.parser.ParseContext) OutputStreamWriter(java.io.OutputStreamWriter) ContentSink(org.apache.stanbol.enhancer.servicesapi.ContentSink) Writer(java.io.Writer) OutputStreamWriter(java.io.OutputStreamWriter)

Aggregations

BufferedInputStream (java.io.BufferedInputStream)1 IOException (java.io.IOException)1 InputStream (java.io.InputStream)1 OutputStreamWriter (java.io.OutputStreamWriter)1 UnsupportedEncodingException (java.io.UnsupportedEncodingException)1 Writer (java.io.Writer)1 PrivilegedActionException (java.security.PrivilegedActionException)1 HashSet (java.util.HashSet)1 Graph (org.apache.clerezza.commons.rdf.Graph)1 IRI (org.apache.clerezza.commons.rdf.IRI)1 PlainLiteralImpl (org.apache.clerezza.commons.rdf.impl.utils.PlainLiteralImpl)1 TripleImpl (org.apache.clerezza.commons.rdf.impl.utils.TripleImpl)1 MultiHandler (org.apache.stanbol.enhancer.engines.tika.handler.MultiHandler)1 PlainTextHandler (org.apache.stanbol.enhancer.engines.tika.handler.PlainTextHandler)1 ContentSink (org.apache.stanbol.enhancer.servicesapi.ContentSink)1 EngineException (org.apache.stanbol.enhancer.servicesapi.EngineException)1 TikaException (org.apache.tika.exception.TikaException)1 Metadata (org.apache.tika.metadata.Metadata)1 MediaType (org.apache.tika.mime.MediaType)1 ParseContext (org.apache.tika.parser.ParseContext)1