Examples with ToTextContentHandler org.apache.tika.sax.ToTextContentHandler used on opensource projects

Example 1 with ToTextContentHandler

use of org.apache.tika.sax.ToTextContentHandler in project ddf by codice.

the class TikaInputTransformer method transform.

@Override
public Metacard transform(InputStream input, String id) throws IOException, CatalogTransformerException {
    LOGGER.debug("Transforming input stream using Tika.");
    if (input == null) {
        throw new CatalogTransformerException("Cannot transform null input.");
    }
    try (TemporaryFileBackedOutputStream fileBackedOutputStream = new TemporaryFileBackedOutputStream()) {
        try {
            IOUtils.copy(input, fileBackedOutputStream);
        } catch (IOException e) {
            throw new CatalogTransformerException("Could not copy bytes of content message.", e);
        }
        Parser parser = new AutoDetectParser();
        ToXMLContentHandler xmlContentHandler = new ToXMLContentHandler();
        ToTextContentHandler textContentHandler = null;
        ContentHandler contentHandler;
        if (!contentMetadataExtractors.isEmpty()) {
            textContentHandler = new ToTextContentHandler();
            contentHandler = new TeeContentHandler(xmlContentHandler, textContentHandler);
        } else {
            contentHandler = xmlContentHandler;
        }
        TikaMetadataExtractor tikaMetadataExtractor = new TikaMetadataExtractor(parser, contentHandler);
        Metadata metadata;
        try (InputStream inputStreamCopy = fileBackedOutputStream.asByteSource().openStream()) {
            metadata = tikaMetadataExtractor.parseMetadata(inputStreamCopy, new ParseContext());
        }
        String metadataText = xmlContentHandler.toString();
        if (templates != null) {
            metadataText = transformToXml(metadataText);
        }
        String metacardContentType = metadata.get(Metadata.CONTENT_TYPE);
        MetacardType metacardType = getMetacardTypeFromMimeType(metacardContentType);
        if (metacardType == null) {
            metacardType = commonTikaMetacardType;
        }
        Metacard metacard;
        if (textContentHandler != null) {
            String plainText = textContentHandler.toString();
            Set<AttributeDescriptor> attributes = contentMetadataExtractors.values().stream().map(ContentMetadataExtractor::getMetacardAttributes).flatMap(Collection::stream).collect(Collectors.toSet());
            MetacardTypeImpl extendedMetacardType = new MetacardTypeImpl(metacardType.getName(), metacardType, attributes);
            metacard = MetacardCreator.createMetacard(metadata, id, metadataText, extendedMetacardType);
            for (ContentMetadataExtractor contentMetadataExtractor : contentMetadataExtractors.values()) {
                contentMetadataExtractor.process(plainText, metacard);
            }
        } else {
            metacard = MetacardCreator.createMetacard(metadata, id, metadataText, metacardType);
        }
        if (StringUtils.isNotBlank(metacardContentType)) {
            metacard.setAttribute(new AttributeImpl(Core.DATATYPE, getDatatype(metacardContentType)));
        }
        if (StringUtils.startsWith(metacardContentType, "image")) {
            try (InputStream inputStreamCopy = fileBackedOutputStream.asByteSource().openStream()) {
                createThumbnail(inputStreamCopy, metacard);
            }
        }
        LOGGER.debug("Finished transforming input stream using Tika.");
        return metacard;
    }
}

Example 2 with ToTextContentHandler

use of org.apache.tika.sax.ToTextContentHandler in project ddf by codice.

the class PdfInputTransformer method transformWithExtractors.

private Metacard transformWithExtractors(InputStream input, String id) throws IOException, CatalogTransformerException {
    try (TemporaryFileBackedOutputStream fbos = new TemporaryFileBackedOutputStream()) {
        try {
            IOUtils.copy(input, fbos);
        } catch (IOException e) {
            throw new CatalogTransformerException("Could not copy bytes of content message.", e);
        }
        String plainText = null;
        try (InputStream isCopy = fbos.asByteSource().openStream()) {
            Parser parser = new AutoDetectParser();
            ContentHandler contentHandler = new ToTextContentHandler();
            TikaMetadataExtractor tikaMetadataExtractor = new TikaMetadataExtractor(parser, contentHandler);
            tikaMetadataExtractor.parseMetadata(isCopy, new ParseContext());
            plainText = contentHandler.toString();
        } catch (CatalogTransformerException e) {
            LOGGER.warn("Cannot extract metadata from pdf", e);
        }
        try (InputStream isCopy = fbos.asByteSource().openStream();
            PDDocument pdfDocument = pdDocumentGenerator.apply(isCopy)) {
            return transformPdf(id, pdfDocument, plainText);
        } catch (InvalidPasswordException e) {
            LOGGER.debug("Cannot transform encrypted pdf", e);
            return initializeMetacard(id);
        }
    }
}