private void crosswalkPDF(Context context, Item item, InputStream metadata) throws CrosswalkException, IOException, SQLException, AuthorizeException { COSDocument cos = null; try { PDFParser parser = new PDFParser(metadata); parser.parse(); cos = parser.getDocument(); // sanity check: PDFBox breaks on encrypted documents, so give up. if (cos.getEncryptionDictionary() != null) throw new MetadataValidationException( "This packager cannot accept an encrypted PDF document."); /* PDF to DC "crosswalk": * * NOTE: This is not in a crosswalk plugin because (a) it isn't * useful anywhere else, and more importantly, (b) the source * data is not XML so it doesn't fit the plugin's interface. * * pattern of crosswalk -- PDF dict entries to DC: * Title -> title.null * Author -> contributor.author * CreationDate -> date.created * ModDate -> date.created * Creator -> description.provenance (application that created orig) * Producer -> description.provenance (convertor to pdf) * Subject -> description.abstract * Keywords -> subject.other * date is java.util.Calendar */ PDDocument pd = new PDDocument(cos); PDDocumentInformation docinfo = pd.getDocumentInformation(); String title = docinfo.getTitle(); // sanity check: item must have a title. if (title == null) throw new MetadataValidationException( "This PDF file is unacceptable, it does not have a value for \"Title\" in its Info dictionary."); log.debug("PDF Info dict title=\"" + title + "\""); item.addDC("title", null, "en", title); String value; Calendar date; if ((value = docinfo.getAuthor()) != null) { item.addDC("contributor", "author", null, value); log.debug("PDF Info dict author=\"" + value + "\""); } if ((value = docinfo.getCreator()) != null) item.addDC( "description", "provenance", "en", "Application that created the original document: " + value); if ((value = docinfo.getProducer()) != null) item.addDC( "description", "provenance", "en", "Original document converted to PDF by: " + value); if ((value = docinfo.getSubject()) != null) item.addDC("description", "abstract", null, value); if ((value = docinfo.getKeywords()) != null) item.addDC("subject", "other", null, value); // Take either CreationDate or ModDate as "date.created", // Too bad there's no place to put "last modified" in the DC. Calendar calValue; if ((calValue = docinfo.getCreationDate()) == null) calValue = docinfo.getModificationDate(); if (calValue != null) item.addDC("date", "created", null, (new DCDate(calValue.getTime())).toString()); item.update(); } finally { if (cos != null) cos.close(); } }
package lius.index.pdf;