/** * This will add the contents to the lucene document. * * @param document The document to add the contents to. * @param is The stream to get the contents from. * @param documentLocation The location of the document, used just for debug messages. * @throws IOException If there is an error parsing the document. */ private void addContent(Document document, InputStream is, String documentLocation) throws IOException { PDDocument pdfDocument = null; PDFTextStripper stripper; try { pdfDocument = PDDocument.load(is); if (pdfDocument.isEncrypted()) { // Just try using the default password and move on pdfDocument.decrypt(""); } // create a writer where to append the text content. StringWriter writer = new StringWriter(); stripper = new PDFTextStripper(); try { stripper.writeText(pdfDocument, writer); } catch (Exception e) { System.out.println("Error in stripper.writeText()"); } String contents = writer.getBuffer().toString(); StringReader reader = new StringReader(contents); addTextField(document, Indexer.contents, reader); PDDocumentInformation info = pdfDocument.getDocumentInformation(); if (info != null) { addTextField(document, Indexer.Author, info.getAuthor()); try { addTextField(document, Indexer.created, info.getCreationDate()); } catch (IOException io) { // ignore, bad date but continue with indexing } addTextField(document, Indexer.keywords, info.getKeywords()); try { addTextField(document, Indexer.modified, info.getModificationDate()); } catch (IOException io) { // ignore, bad date but continue with indexing } addTextField(document, "Subject", info.getSubject()); addTextField(document, Indexer.Title, info.getTitle()); } int summarySize = Math.min(contents.length(), 500); String summary = contents.substring(0, summarySize); // Add the summary as an UnIndexed field, so that it is stored and // returned // with hit documents for display. addUnindexedField(document, Indexer.summary, summary); } catch (CryptographyException e) { throw new IOException("Error decrypting document(" + documentLocation + "): " + e); } catch (InvalidPasswordException e) { // they didn't suppply a password and the default of "" was wrong. throw new IOException( "Error: The document(" + documentLocation + ") is encrypted and will not be indexed."); } finally { if (pdfDocument != null) { pdfDocument.close(); } } }
/** * This will add the contents to the lucene document. * * @param document The document to add the contents to. * @param is The stream to get the contents from. * @param documentLocation The location of the document, used just for debug messages. * @throws IOException If there is an error parsing the document. */ private void addContent(Document document, InputStream is, String documentLocation) throws IOException { PDDocument pdfDocument = null; try { pdfDocument = PDDocument.load(is, ""); // create a writer where to append the text content. StringWriter writer = new StringWriter(); if (stripper == null) { stripper = new PDFTextStripper(); } stripper.writeText(pdfDocument, writer); // Note: the buffer to string operation is costless; // the char array value of the writer buffer and the content string // is shared as long as the buffer content is not modified, which will // not occur here. String contents = writer.getBuffer().toString(); StringReader reader = new StringReader(contents); // Add the tag-stripped contents as a Reader-valued Text field so it will // get tokenized and indexed. addTextField(document, "contents", reader); PDDocumentInformation info = pdfDocument.getDocumentInformation(); if (info != null) { addTextField(document, "Author", info.getAuthor()); addTextField(document, "CreationDate", info.getCreationDate()); addTextField(document, "Creator", info.getCreator()); addTextField(document, "Keywords", info.getKeywords()); addTextField(document, "ModificationDate", info.getModificationDate()); addTextField(document, "Producer", info.getProducer()); addTextField(document, "Subject", info.getSubject()); addTextField(document, "Title", info.getTitle()); addTextField(document, "Trapped", info.getTrapped()); } int summarySize = Math.min(contents.length(), 500); String summary = contents.substring(0, summarySize); // Add the summary as an UnIndexed field, so that it is stored and returned // with hit documents for display. addUnindexedField(document, "summary", summary); } catch (InvalidPasswordException e) { // they didn't suppply a password and the default of "" was wrong. throw new IOException( "Error: The document(" + documentLocation + ") is encrypted and will not be indexed.", e); } finally { if (pdfDocument != null) { pdfDocument.close(); } } }
@Override public Document[] parse( final AnchorURL location, final String mimeType, final String charset, final VocabularyScraper scraper, final int timezoneOffset, final InputStream source) throws Parser.Failure, InterruptedException { // check memory for parser if (!MemoryControl.request(200 * 1024 * 1024, false)) throw new Parser.Failure( "Not enough Memory available for pdf parser: " + MemoryControl.available(), location); // create a pdf parser PDDocument pdfDoc; try { Thread.currentThread().setPriority(Thread.MIN_PRIORITY); // the pdfparser is a big pain // pdfDoc = PDDocument.load(source); final PDFParser pdfParser = new PDFParser(source); pdfParser.setTempDirectory(new File(System.getProperty("java.io.tmpdir"))); pdfParser.parse(); pdfDoc = pdfParser.getPDDocument(); } catch (final IOException e) { throw new Parser.Failure(e.getMessage(), location); } finally { Thread.currentThread().setPriority(Thread.NORM_PRIORITY); } if (pdfDoc.isEncrypted()) { try { pdfDoc.openProtection(new StandardDecryptionMaterial("")); } catch (final BadSecurityHandlerException e) { try { pdfDoc.close(); } catch (final IOException ee) { } throw new Parser.Failure("Document is encrypted (1): " + e.getMessage(), location); } catch (final IOException e) { try { pdfDoc.close(); } catch (final IOException ee) { } throw new Parser.Failure("Document is encrypted (2): " + e.getMessage(), location); } catch (final CryptographyException e) { try { pdfDoc.close(); } catch (final IOException ee) { } throw new Parser.Failure("Document is encrypted (3): " + e.getMessage(), location); } final AccessPermission perm = pdfDoc.getCurrentAccessPermission(); if (perm == null || !perm.canExtractContent()) { try { pdfDoc.close(); } catch (final IOException ee) { } throw new Parser.Failure("Document is encrypted and cannot be decrypted", location); } } // extracting some metadata PDDocumentInformation info = pdfDoc.getDocumentInformation(); String docTitle = null, docSubject = null, docAuthor = null, docPublisher = null, docKeywordStr = null; Date docDate = new Date(); if (info != null) { docTitle = info.getTitle(); docSubject = info.getSubject(); docAuthor = info.getAuthor(); docPublisher = info.getProducer(); if (docPublisher == null || docPublisher.isEmpty()) docPublisher = info.getCreator(); docKeywordStr = info.getKeywords(); try { if (info.getModificationDate() != null) docDate = info.getModificationDate().getTime(); } catch (IOException e) { } // unused: // info.getTrapped()); } info = null; if (docTitle == null || docTitle.isEmpty()) { docTitle = MultiProtocolURL.unescape(location.getFileName()); } if (docTitle == null) { docTitle = docSubject; } String[] docKeywords = null; if (docKeywordStr != null) { docKeywords = docKeywordStr.split(" |,"); } Collection<AnchorURL>[] pdflinks = null; Document[] result = null; try { // get the links pdflinks = extractPdfLinks(pdfDoc); // get the fulltext (either per document or for each page) final PDFTextStripper stripper = new PDFTextStripper(StandardCharsets.UTF_8.name()); if (individualPages) { // this is a hack which stores individual pages of the source pdf into individual index // documents // the new documents will get a virtual link with a post argument page=X appended to the // original url // collect text int pagecount = pdfDoc.getNumberOfPages(); String[] pages = new String[pagecount]; for (int page = 1; page <= pagecount; page++) { stripper.setStartPage(page); stripper.setEndPage(page); pages[page - 1] = stripper.getText(pdfDoc); // System.out.println("PAGE " + page + ": " + pages[page - 1]); } // create individual documents for each page assert pages.length == pdflinks.length : "pages.length = " + pages.length + ", pdflinks.length = " + pdflinks.length; result = new Document[Math.min(pages.length, pdflinks.length)]; String loc = location.toNormalform(true); for (int page = 0; page < result.length; page++) { result[page] = new Document( new AnchorURL( loc + (loc.indexOf('?') > 0 ? '&' : '?') + individualPagePropertyname + '=' + (page + 1)), // these are virtual new pages; we cannot combine them with '#' // as that would be removed when computing the urlhash mimeType, StandardCharsets.UTF_8.name(), this, null, docKeywords, singleList(docTitle), docAuthor, docPublisher, null, null, 0.0f, 0.0f, pages == null || page > pages.length ? new byte[0] : UTF8.getBytes(pages[page]), pdflinks == null || page >= pdflinks.length ? null : pdflinks[page], null, null, false, docDate); } } else { // collect the whole text at once final CharBuffer writer = new CharBuffer(odtParser.MAX_DOCSIZE); byte[] contentBytes = new byte[0]; stripper.setEndPage(3); // get first 3 pages (always) writer.append(stripper.getText(pdfDoc)); contentBytes = writer.getBytes(); // remember text in case of interrupting thread if (pdfDoc.getNumberOfPages() > 3) { // spare creating/starting thread if all pages read stripper.setStartPage(4); // continue with page 4 (terminated, resulting in no text) stripper.setEndPage(Integer.MAX_VALUE); // set to default // we start the pdf parsing in a separate thread to ensure that it can be terminated final PDDocument pdfDocC = pdfDoc; final Thread t = new Thread() { @Override public void run() { Thread.currentThread().setName("pdfParser.getText:" + location); try { writer.append(stripper.getText(pdfDocC)); } catch (final Throwable e) { } } }; t.start(); t.join(3000); // pdfbox likes to forget to terminate ... (quite often) if (t.isAlive()) t.interrupt(); } contentBytes = writer.getBytes(); // get final text before closing writer Collection<AnchorURL> pdflinksCombined = new HashSet<AnchorURL>(); for (Collection<AnchorURL> pdflinksx : pdflinks) if (pdflinksx != null) pdflinksCombined.addAll(pdflinksx); result = new Document[] { new Document( location, mimeType, StandardCharsets.UTF_8.name(), this, null, docKeywords, singleList(docTitle), docAuthor, docPublisher, null, null, 0.0f, 0.0f, contentBytes, pdflinksCombined, null, null, false, docDate) }; } } catch (final Throwable e) { // close the writer (in finally) // throw new Parser.Failure(e.getMessage(), location); } finally { try { pdfDoc.close(); } catch (final Throwable e) { } } // clear resources in pdfbox. they say that is resolved but it's not. see: // https://issues.apache.org/jira/browse/PDFBOX-313 // https://issues.apache.org/jira/browse/PDFBOX-351 // https://issues.apache.org/jira/browse/PDFBOX-441 // the pdfbox still generates enormeous number of object allocations and don't delete these // the following Object are statically stored and never flushed: // COSFloat, COSArray, COSInteger, COSObjectKey, COSObject, COSDictionary, // COSStream, COSString, COSName, COSDocument, COSInteger[], COSNull // the great number of these objects can easily be seen in Java Visual VM // we try to get this shit out of the memory here by forced clear calls, hope the best the // rubbish gets out. pdfDoc = null; clean_up_idiotic_PDFParser_font_cache_which_eats_up_tons_of_megabytes(); return result; }