public void parse(
InputStream stream, ContentHandler baseHandler, Metadata metadata, ParseContext context)
throws IOException, SAXException, TikaException {
// As we don't know which of the metadata or the content
// we'll hit first, catch the endDocument call initially
EndDocumentShieldingContentHandler handler =
new EndDocumentShieldingContentHandler(baseHandler);
// Process the file in turn
ZipInputStream zip = new ZipInputStream(stream);
ZipEntry entry = zip.getNextEntry();
while (entry != null) {
if (entry.getName().equals("mimetype")) {
String type = IOUtils.toString(zip, "UTF-8");
metadata.set(Metadata.CONTENT_TYPE, type);
} else if (entry.getName().equals("meta.xml")) {
meta.parse(zip, new DefaultHandler(), metadata, context);
} else if (entry.getName().endsWith("content.xml")) {
content.parse(zip, handler, metadata, context);
}
entry = zip.getNextEntry();
}
// Only now call the end document
if (handler.getEndDocumentWasCalled()) {
handler.reallyEndDocument();
}
}
public MediaType detect(TikaInputStream input, Metadata metadata) throws IOException {
ZipFile zip = new ZipFile(input.getFile());
for (ZipEntry entry : Collections.list(zip.entries())) {
// Is it an Open Document file?
if (entry.getName().equals("mimetype")) {
InputStream stream = zip.getInputStream(entry);
try {
return fromString(IOUtils.toString(stream, "UTF-8"));
} finally {
stream.close();
}
} else if (entry.getName().equals("_rels/.rels")
|| entry.getName().equals("[Content_Types].xml")) {
// Office Open XML File
// As POI to open and investigate it for us
try {
OPCPackage pkg = OPCPackage.open(input.getFile().toString());
input.setOpenContainer(pkg);
PackageRelationshipCollection core =
pkg.getRelationshipsByType(ExtractorFactory.CORE_DOCUMENT_REL);
if (core.size() != 1) {
throw new IOException(
"Invalid OOXML Package received - expected 1 core document, found " + core.size());
}
// Get the type of the core document part
PackagePart corePart = pkg.getPart(core.getRelationship(0));
String coreType = corePart.getContentType();
// Turn that into the type of the overall document
String docType = coreType.substring(0, coreType.lastIndexOf('.'));
return fromString(docType);
} catch (InvalidFormatException e) {
throw new IOException("Office Open XML File detected, but corrupted - " + e.getMessage());
}
} else if (entry.getName().equals("buildVersionHistory.plist")) {
// This is an iWork document
// Reset and ask
zip.close();
zip = new ZipFile(input.getFile());
return IWorkPackageParser.identifyType(zip);
} else if (entry.getName().equals("META-INF/")) {
// Java Jar
return MediaType.application("java-archive");
}
}
return MediaType.APPLICATION_ZIP;
}
