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

Example 1 with EmbeddedContentHandler

use of org.apache.tika.sax.EmbeddedContentHandler in project tika by apache.

the class XMLParser method parse.

public void parse(InputStream stream, ContentHandler handler, Metadata metadata, ParseContext context) throws IOException, SAXException, TikaException {
    if (metadata.get(Metadata.CONTENT_TYPE) == null) {
        metadata.set(Metadata.CONTENT_TYPE, "application/xml");
    }
    final XHTMLContentHandler xhtml = new XHTMLContentHandler(handler, metadata);
    xhtml.startDocument();
    xhtml.startElement("p");
    TaggedContentHandler tagged = new TaggedContentHandler(handler);
    try {
        context.getSAXParser().parse(new CloseShieldInputStream(stream), new OfflineContentHandler(new EmbeddedContentHandler(getContentHandler(tagged, metadata, context))));
    } catch (SAXException e) {
        tagged.throwIfCauseOf(e);
        throw new TikaException("XML parse error", e);
    } finally {
        xhtml.endElement("p");
        xhtml.endDocument();
    }
}

Example 2 with EmbeddedContentHandler

use of org.apache.tika.sax.EmbeddedContentHandler in project tika by apache.

the class MockParser method handleEmbedded.

private void handleEmbedded(Node action, XHTMLContentHandler handler, ParseContext context) throws TikaException, SAXException, IOException {
    String fileName = "";
    String contentType = "";
    NamedNodeMap attrs = action.getAttributes();
    if (attrs != null) {
        Node n = attrs.getNamedItem("filename");
        if (n != null) {
            fileName = n.getNodeValue();
        }
        n = attrs.getNamedItem("content-type");
        if (n != null) {
            contentType = n.getNodeValue();
        }
    }
    String embeddedText = action.getTextContent();
    EmbeddedDocumentExtractor extractor = getEmbeddedDocumentExtractor(context);
    Metadata m = new Metadata();
    m.set(TikaMetadataKeys.RESOURCE_NAME_KEY, fileName);
    if (!"".equals(contentType)) {
        m.set(Metadata.CONTENT_TYPE, contentType);
    }
    InputStream is = new ByteArrayInputStream(embeddedText.getBytes(UTF_8));
    extractor.parseEmbedded(is, new EmbeddedContentHandler(handler), m, true);
}

Example 3 with EmbeddedContentHandler

use of org.apache.tika.sax.EmbeddedContentHandler in project tika by apache.

the class OutlookExtractor method parse.

public void parse(XHTMLContentHandler xhtml, Metadata metadata) throws TikaException, SAXException, IOException {
    try {
        msg.setReturnNullOnMissingChunk(true);
        try {
            metadata.set(Office.MAPI_MESSAGE_CLASS, getMessageClass(msg.getMessageClass()));
        } catch (ChunkNotFoundException e) {
        }
        //  as Unicode, try to sort out an encoding for them
        if (msg.has7BitEncodingStrings()) {
            guess7BitEncoding(msg);
        }
        // Start with the metadata
        String subject = msg.getSubject();
        Map<String, String[]> headers = normalizeHeaders(msg.getHeaders());
        String from = msg.getDisplayFrom();
        handleFromTo(headers, metadata);
        metadata.set(TikaCoreProperties.TITLE, subject);
        // TODO: Move to description in Tika 2.0
        metadata.set(TikaCoreProperties.TRANSITION_SUBJECT_TO_DC_DESCRIPTION, msg.getConversationTopic());
        try {
            for (String recipientAddress : msg.getRecipientEmailAddressList()) {
                if (recipientAddress != null)
                    metadata.add(Metadata.MESSAGE_RECIPIENT_ADDRESS, recipientAddress);
            }
        } catch (ChunkNotFoundException he) {
        }
        for (Map.Entry<String, String[]> e : headers.entrySet()) {
            String headerKey = e.getKey();
            for (String headerValue : e.getValue()) {
                metadata.add(Metadata.MESSAGE_RAW_HEADER_PREFIX + headerKey, headerValue);
            }
        }
        // First try via the proper chunk
        if (msg.getMessageDate() != null) {
            metadata.set(TikaCoreProperties.CREATED, msg.getMessageDate().getTime());
            metadata.set(TikaCoreProperties.MODIFIED, msg.getMessageDate().getTime());
        } else {
            if (headers != null && headers.size() > 0) {
                for (Map.Entry<String, String[]> header : headers.entrySet()) {
                    String headerKey = header.getKey();
                    if (headerKey.toLowerCase(Locale.ROOT).startsWith("date:")) {
                        String date = headerKey.substring(headerKey.indexOf(':') + 1).trim();
                        // See if we can parse it as a normal mail date
                        try {
                            Date d = MboxParser.parseDate(date);
                            metadata.set(TikaCoreProperties.CREATED, d);
                            metadata.set(TikaCoreProperties.MODIFIED, d);
                        } catch (ParseException e) {
                            // Store it as-is, and hope for the best...
                            metadata.set(TikaCoreProperties.CREATED, date);
                            metadata.set(TikaCoreProperties.MODIFIED, date);
                        }
                        break;
                    }
                }
            }
        }
        xhtml.element("h1", subject);
        // Output the from and to details in text, as you
        //  often want them in text form for searching
        xhtml.startElement("dl");
        if (from != null) {
            header(xhtml, "From", from);
        }
        header(xhtml, "To", msg.getDisplayTo());
        header(xhtml, "Cc", msg.getDisplayCC());
        header(xhtml, "Bcc", msg.getDisplayBCC());
        try {
            header(xhtml, "Recipients", msg.getRecipientEmailAddress());
        } catch (ChunkNotFoundException e) {
        }
        xhtml.endElement("dl");
        // Get the message body. Preference order is: html, rtf, text
        Chunk htmlChunk = null;
        Chunk rtfChunk = null;
        Chunk textChunk = null;
        for (Chunk chunk : msg.getMainChunks().getChunks()) {
            if (chunk.getChunkId() == MAPIProperty.BODY_HTML.id) {
                htmlChunk = chunk;
            }
            if (chunk.getChunkId() == MAPIProperty.RTF_COMPRESSED.id) {
                rtfChunk = chunk;
            }
            if (chunk.getChunkId() == MAPIProperty.BODY.id) {
                textChunk = chunk;
            }
        }
        boolean doneBody = false;
        xhtml.startElement("div", "class", "message-body");
        if (htmlChunk != null) {
            byte[] data = null;
            if (htmlChunk instanceof ByteChunk) {
                data = ((ByteChunk) htmlChunk).getValue();
            } else if (htmlChunk instanceof StringChunk) {
                data = ((StringChunk) htmlChunk).getRawValue();
            }
            if (data != null) {
                Parser htmlParser = EmbeddedDocumentUtil.tryToFindExistingLeafParser(HtmlParser.class, parseContext);
                if (htmlParser == null) {
                    htmlParser = new HtmlParser();
                }
                htmlParser.parse(new ByteArrayInputStream(data), new EmbeddedContentHandler(new BodyContentHandler(xhtml)), new Metadata(), parseContext);
                doneBody = true;
            }
        }
        if (rtfChunk != null && !doneBody) {
            ByteChunk chunk = (ByteChunk) rtfChunk;
            MAPIRtfAttribute rtf = new MAPIRtfAttribute(MAPIProperty.RTF_COMPRESSED, Types.BINARY.getId(), chunk.getValue());
            Parser rtfParser = EmbeddedDocumentUtil.tryToFindExistingLeafParser(RTFParser.class, parseContext);
            if (rtfParser == null) {
                rtfParser = new RTFParser();
            }
            rtfParser.parse(new ByteArrayInputStream(rtf.getData()), new EmbeddedContentHandler(new BodyContentHandler(xhtml)), new Metadata(), parseContext);
            doneBody = true;
        }
        if (textChunk != null && !doneBody) {
            xhtml.element("p", ((StringChunk) textChunk).getValue());
        }
        xhtml.endElement("div");
        // Process the attachments
        for (AttachmentChunks attachment : msg.getAttachmentFiles()) {
            xhtml.startElement("div", "class", "attachment-entry");
            String filename = null;
            if (attachment.getAttachLongFileName() != null) {
                filename = attachment.getAttachLongFileName().getValue();
            } else if (attachment.getAttachFileName() != null) {
                filename = attachment.getAttachFileName().getValue();
            }
            if (filename != null && filename.length() > 0) {
                xhtml.element("h1", filename);
            }
            if (attachment.getAttachData() != null) {
                handleEmbeddedResource(TikaInputStream.get(attachment.getAttachData().getValue()), filename, null, null, xhtml, true);
            }
            if (attachment.getAttachmentDirectory() != null) {
                handleEmbeddedOfficeDoc(attachment.getAttachmentDirectory().getDirectory(), xhtml);
            }
            xhtml.endElement("div");
        }
    } catch (ChunkNotFoundException e) {
        throw new TikaException("POI MAPIMessage broken - didn't return null on missing chunk", e);
    } finally {
    //You'd think you'd want to call msg.close().
    //Don't do that.  That closes down the file system.
    //If an msg has multiple msg attachments, some of them
    //can reside in the same file system.  After the first
    //child is read, the fs is closed, and the other children
    //get a java.nio.channels.ClosedChannelException
    }
}

Example 4 with EmbeddedContentHandler

use of org.apache.tika.sax.EmbeddedContentHandler in project tika by apache.

the class SXSLFPowerPointExtractorDecorator method handleSlidePart.

private void handleSlidePart(PackagePart slidePart, XHTMLContentHandler xhtml) throws IOException, SAXException {
    Map<String, String> linkedRelationships = loadLinkedRelationships(slidePart, false, metadata);
    //        Map<String, String> hyperlinks = loadHyperlinkRelationships(packagePart);
    xhtml.startElement("div", "class", "slide-content");
    try (InputStream stream = slidePart.getInputStream()) {
        context.getSAXParser().parse(new CloseShieldInputStream(stream), new OfflineContentHandler(new EmbeddedContentHandler(new OOXMLWordAndPowerPointTextHandler(new OOXMLTikaBodyPartHandler(xhtml), linkedRelationships))));
    } catch (TikaException e) {
        metadata.add(TikaCoreProperties.TIKA_META_EXCEPTION_WARNING, ExceptionUtils.getStackTrace(e));
    }
    xhtml.endElement("div");
    handleBasicRelatedParts(XSLFRelation.SLIDE_LAYOUT.getRelation(), "slide-master-content", slidePart, new PlaceHolderSkipper(new OOXMLWordAndPowerPointTextHandler(new OOXMLTikaBodyPartHandler(xhtml), linkedRelationships)));
    handleBasicRelatedParts(XSLFRelation.NOTES.getRelation(), "slide-notes", slidePart, new OOXMLWordAndPowerPointTextHandler(new OOXMLTikaBodyPartHandler(xhtml), linkedRelationships));
    handleBasicRelatedParts(XSLFRelation.NOTES_MASTER.getRelation(), "slide-notes-master", slidePart, new OOXMLWordAndPowerPointTextHandler(new OOXMLTikaBodyPartHandler(xhtml), linkedRelationships));
    handleBasicRelatedParts(XSLFRelation.COMMENTS.getRelation(), null, slidePart, new XSLFCommentsHandler(xhtml));
}

Example 5 with EmbeddedContentHandler

use of org.apache.tika.sax.EmbeddedContentHandler in project tika by apache.

the class SXSLFPowerPointExtractorDecorator method handleBasicRelatedParts.

/**
     * This should handle the comments, master, notes, etc
     *
     * @param contentType
     * @param xhtmlClassLabel
     * @param parentPart
     * @param contentHandler
     */
private void handleBasicRelatedParts(String contentType, String xhtmlClassLabel, PackagePart parentPart, ContentHandler contentHandler) throws SAXException {
    PackageRelationshipCollection relatedPartPRC = null;
    try {
        relatedPartPRC = parentPart.getRelationshipsByType(contentType);
    } catch (InvalidFormatException e) {
        metadata.add(TikaCoreProperties.TIKA_META_EXCEPTION_WARNING, ExceptionUtils.getStackTrace(e));
    }
    if (relatedPartPRC != null && relatedPartPRC.size() > 0) {
        AttributesImpl attributes = new AttributesImpl();
        attributes.addAttribute("", "class", "class", "CDATA", xhtmlClassLabel);
        contentHandler.startElement("", "div", "div", attributes);
        for (int i = 0; i < relatedPartPRC.size(); i++) {
            PackageRelationship relatedPartPackageRelationship = relatedPartPRC.getRelationship(i);
            try {
                PackagePart relatedPartPart = parentPart.getRelatedPart(relatedPartPackageRelationship);
                try (InputStream stream = relatedPartPart.getInputStream()) {
                    context.getSAXParser().parse(stream, new OfflineContentHandler(new EmbeddedContentHandler(contentHandler)));
                } catch (IOException | TikaException e) {
                    metadata.add(TikaCoreProperties.TIKA_META_EXCEPTION_WARNING, ExceptionUtils.getStackTrace(e));
                }
            } catch (InvalidFormatException e) {
                metadata.add(TikaCoreProperties.TIKA_META_EXCEPTION_WARNING, ExceptionUtils.getStackTrace(e));
            }
        }
        contentHandler.endElement("", "div", "div");
    }
}