// nsIExtendedExpatSink
NS_IMETHODIMP
nsSAXXMLReader::HandleStartElement(const PRUnichar *aName,
const PRUnichar **aAtts,
uint32_t aAttsCount,
int32_t aIndex,
uint32_t aLineNumber)
{
if (!mContentHandler)
return NS_OK;
nsCOMPtr<nsSAXAttributes> atts = new nsSAXAttributes();
if (!atts)
return NS_ERROR_OUT_OF_MEMORY;
nsAutoString uri, localName, qName;
for (; *aAtts; aAtts += 2) {
SplitExpatName(aAtts[0], uri, localName, qName);
// XXX don't have attr type information
NS_NAMED_LITERAL_STRING(cdataType, "CDATA");
// could support xmlns reporting, it's a standard SAX feature
if (!uri.EqualsLiteral(XMLNS_URI)) {
NS_ASSERTION(aAtts[1], "null passed to handler");
atts->AddAttribute(uri, localName, qName, cdataType,
nsDependentString(aAtts[1]));
}
}
// Deal with the element name
SplitExpatName(aName, uri, localName, qName);
return mContentHandler->StartElement(uri, localName, qName, atts);
}
void
RDFContentSinkImpl::SetParseMode(const PRUnichar **aAttributes)
{
nsCOMPtr<nsIAtom> localName;
for (; *aAttributes; aAttributes += 2) {
const nsDependentSubstring& nameSpaceURI =
SplitExpatName(aAttributes[0], getter_AddRefs(localName));
if (localName == kParseTypeAtom) {
nsDependentString v(aAttributes[1]);
if (nameSpaceURI.IsEmpty() ||
nameSpaceURI.EqualsLiteral(RDF_NAMESPACE_URI)) {
if (v.EqualsLiteral("Resource"))
mParseMode = eRDFContentSinkParseMode_Resource;
break;
}
else if (nameSpaceURI.EqualsLiteral(NC_NAMESPACE_URI)) {
if (v.EqualsLiteral("Date"))
mParseMode = eRDFContentSinkParseMode_Date;
else if (v.EqualsLiteral("Integer"))
mParseMode = eRDFContentSinkParseMode_Int;
break;
}
}
}
}
nsresult
RDFContentSinkImpl::AddProperties(const PRUnichar** aAttributes,
nsIRDFResource* aSubject,
PRInt32* aCount)
{
if (aCount)
*aCount = 0;
nsCOMPtr<nsIAtom> localName;
for (; *aAttributes; aAttributes += 2) {
const nsDependentSubstring& nameSpaceURI =
SplitExpatName(aAttributes[0], getter_AddRefs(localName));
// skip 'xmlns' directives, these are "meta" information
if (nameSpaceURI.EqualsLiteral("http://www.w3.org/2000/xmlns/")) {
continue;
}
// skip `about', `ID', `resource', and 'nodeID' attributes (either with or
// without the `rdf:' prefix); these are all "special" and
// should've been dealt with by the caller.
if (localName == kAboutAtom || localName == kIdAtom ||
localName == kResourceAtom || localName == kNodeIdAtom) {
if (nameSpaceURI.IsEmpty() ||
nameSpaceURI.EqualsLiteral(RDF_NAMESPACE_URI))
continue;
}
// Skip `parseType', `RDF:parseType', and `NC:parseType'. This
// is meta-information that will be handled in SetParseMode.
if (localName == kParseTypeAtom) {
if (nameSpaceURI.IsEmpty() ||
nameSpaceURI.EqualsLiteral(RDF_NAMESPACE_URI) ||
nameSpaceURI.EqualsLiteral(NC_NAMESPACE_URI)) {
continue;
}
}
const char* attrName;
localName->GetUTF8String(&attrName);
NS_ConvertUTF16toUTF8 propertyStr(nameSpaceURI);
propertyStr.Append(attrName);
// Add the assertion to RDF
nsCOMPtr<nsIRDFResource> property;
gRDFService->GetResource(propertyStr, getter_AddRefs(property));
nsCOMPtr<nsIRDFLiteral> target;
gRDFService->GetLiteral(aAttributes[1],
getter_AddRefs(target));
mDataSource->Assert(aSubject, property, target, PR_TRUE);
}
return NS_OK;
}
NS_IMETHODIMP
nsSAXXMLReader::HandleEndElement(const PRUnichar *aName)
{
if (mContentHandler) {
nsAutoString uri, localName, qName;
SplitExpatName(aName, uri, localName, qName);
return mContentHandler->EndElement(uri, localName, qName);
}
return NS_OK;
}
nsresult
RDFContentSinkImpl::OpenMember(const PRUnichar* aName,
const PRUnichar** aAttributes)
{
// ensure that we're actually reading a member element by making
// sure that the opening tag is <rdf:li>, where "rdf:" corresponds
// to whatever they've declared the standard RDF namespace to be.
nsresult rv;
nsCOMPtr<nsIAtom> localName;
const nsDependentSubstring& nameSpaceURI =
SplitExpatName(aName, getter_AddRefs(localName));
if (!nameSpaceURI.EqualsLiteral(RDF_NAMESPACE_URI) ||
localName != kLiAtom) {
PR_LOG(gLog, PR_LOG_ALWAYS,
("rdfxml: expected RDF:li at line %d",
-1)); // XXX pass in line number
return NS_ERROR_UNEXPECTED;
}
// The parent element is the container.
nsIRDFResource* container = GetContextElement(0);
if (! container)
return NS_ERROR_NULL_POINTER;
nsIRDFResource* resource;
if (NS_SUCCEEDED(rv = GetResourceAttribute(aAttributes, &resource))) {
// Okay, this node has an RDF:resource="..." attribute. That
// means that it's a "referenced item," as covered in [6.29].
nsCOMPtr<nsIRDFContainer> c;
NS_NewRDFContainer(getter_AddRefs(c));
c->Init(mDataSource, container);
c->AppendElement(resource);
// XXX Technically, we should _not_ fall through here and push
// the element onto the stack: this is supposed to be a closed
// node. But right now I'm lazy and the code will just Do The
// Right Thing so long as the RDF is well-formed.
NS_RELEASE(resource);
}
// Change state. Pushing a null context element is a bit weird,
// but the idea is that there really is _no_ context "property".
// The contained element will use nsIRDFContainer::AppendElement() to add
// the element to the container, which requires only the container
// and the element to be added.
PushContext(nsnull, mState, mParseMode);
mState = eRDFContentSinkState_InMemberElement;
SetParseMode(aAttributes);
return NS_OK;
}
nsresult
RDFContentSinkImpl::OpenRDF(const PRUnichar* aName)
{
// ensure that we're actually reading RDF by making sure that the
// opening tag is <rdf:RDF>, where "rdf:" corresponds to whatever
// they've declared the standard RDF namespace to be.
nsCOMPtr<nsIAtom> localName;
const nsDependentSubstring& nameSpaceURI =
SplitExpatName(aName, getter_AddRefs(localName));
if (!nameSpaceURI.EqualsLiteral(RDF_NAMESPACE_URI) || localName != kRDFAtom) {
// PR_LOG(gLog, PR_LOG_ALWAYS,
// ("rdfxml: expected RDF:RDF at line %d",
// aNode.GetSourceLineNumber()));
return NS_ERROR_UNEXPECTED;
}
PushContext(nsnull, mState, mParseMode);
mState = eRDFContentSinkState_InDocumentElement;
return NS_OK;
}
nsresult
RDFContentSinkImpl::GetResourceAttribute(const PRUnichar** aAttributes,
nsIRDFResource** aResource)
{
nsCOMPtr<nsIAtom> localName;
nsAutoString nodeID;
for (; *aAttributes; aAttributes += 2) {
const nsDependentSubstring& nameSpaceURI =
SplitExpatName(aAttributes[0], getter_AddRefs(localName));
// We'll accept `resource' or `rdf:resource', under the spirit
// that we should be liberal towards the input that we
// receive.
if (!nameSpaceURI.IsEmpty() &&
!nameSpaceURI.EqualsLiteral(RDF_NAMESPACE_URI)) {
continue;
}
// XXX you can't specify both, but we'll just pick up the
// first thing that was specified and ignore the other.
if (localName == kResourceAtom) {
// XXX Take the URI and make it fully qualified by
// sticking it into the document's URL. This may not be
// appropriate...
nsAutoString relURI(aAttributes[1]);
if (rdf_RequiresAbsoluteURI(relURI)) {
nsresult rv;
nsCAutoString uri;
rv = mDocumentURL->Resolve(NS_ConvertUTF16toUTF8(aAttributes[1]), uri);
if (NS_FAILED(rv)) return rv;
return gRDFService->GetResource(uri, aResource);
}
return gRDFService->GetResource(NS_ConvertUTF16toUTF8(aAttributes[1]),
aResource);
}
else if (localName == kNodeIdAtom) {
nodeID.Assign(aAttributes[1]);
}
}
// If nodeID is present, check if we already know about it. If we've seen
// the nodeID before, use the same resource, otherwise generate a new one.
if (!nodeID.IsEmpty()) {
mNodeIDMap.Get(nodeID,aResource);
if (!*aResource) {
nsresult rv;
rv = gRDFService->GetAnonymousResource(aResource);
if (NS_FAILED(rv)) {
return rv;
}
mNodeIDMap.Put(nodeID,*aResource);
}
return NS_OK;
}
return NS_ERROR_FAILURE;
}
nsresult
RDFContentSinkImpl::GetIdAboutAttribute(const PRUnichar** aAttributes,
nsIRDFResource** aResource,
PRBool* aIsAnonymous)
{
// This corresponds to the dirty work of production [6.5]
nsresult rv = NS_OK;
nsAutoString nodeID;
nsCOMPtr<nsIAtom> localName;
for (; *aAttributes; aAttributes += 2) {
const nsDependentSubstring& nameSpaceURI =
SplitExpatName(aAttributes[0], getter_AddRefs(localName));
// We'll accept either `ID' or `rdf:ID' (ibid with `about' or
// `rdf:about') in the spirit of being liberal towards the
// input that we receive.
if (!nameSpaceURI.IsEmpty() &&
!nameSpaceURI.EqualsLiteral(RDF_NAMESPACE_URI)) {
continue;
}
// XXX you can't specify both, but we'll just pick up the
// first thing that was specified and ignore the other.
if (localName == kAboutAtom) {
if (aIsAnonymous)
*aIsAnonymous = PR_FALSE;
nsAutoString relURI(aAttributes[1]);
if (rdf_RequiresAbsoluteURI(relURI)) {
nsCAutoString uri;
rv = mDocumentURL->Resolve(NS_ConvertUTF16toUTF8(aAttributes[1]), uri);
if (NS_FAILED(rv)) return rv;
return gRDFService->GetResource(uri,
aResource);
}
return gRDFService->GetResource(NS_ConvertUTF16toUTF8(aAttributes[1]),
aResource);
}
else if (localName == kIdAtom) {
if (aIsAnonymous)
*aIsAnonymous = PR_FALSE;
// In the spirit of leniency, we do not bother trying to
// enforce that this be a valid "XML Name" (see
// http://www.w3.org/TR/REC-xml#NT-Nmtoken), as per
// 6.21. If we wanted to, this would be where to do it.
// Construct an in-line resource whose URI is the
// document's URI plus the XML name specified in the ID
// attribute.
nsCAutoString name;
nsCAutoString ref('#');
AppendUTF16toUTF8(aAttributes[1], ref);
rv = mDocumentURL->Resolve(ref, name);
if (NS_FAILED(rv)) return rv;
return gRDFService->GetResource(name, aResource);
}
else if (localName == kNodeIdAtom) {
nodeID.Assign(aAttributes[1]);
}
else if (localName == kAboutEachAtom) {
// XXX we don't deal with aboutEach...
//PR_LOG(gLog, PR_LOG_WARNING,
// ("rdfxml: ignoring aboutEach at line %d",
// aNode.GetSourceLineNumber()));
}
}
// Otherwise, we couldn't find anything, so just gensym one...
if (aIsAnonymous)
*aIsAnonymous = PR_TRUE;
// If nodeID is present, check if we already know about it. If we've seen
// the nodeID before, use the same resource, otherwise generate a new one.
if (!nodeID.IsEmpty()) {
mNodeIDMap.Get(nodeID,aResource);
if (!*aResource) {
rv = gRDFService->GetAnonymousResource(aResource);
mNodeIDMap.Put(nodeID,*aResource);
}
}
else {
rv = gRDFService->GetAnonymousResource(aResource);
}
return rv;
}
nsresult
RDFContentSinkImpl::OpenProperty(const PRUnichar* aName, const PRUnichar** aAttributes)
{
nsresult rv;
// an "object" non-terminal is either a "description", a "typed
// node", or a "container", so this change the content sink's
// state appropriately.
nsCOMPtr<nsIAtom> localName;
const nsDependentSubstring& nameSpaceURI =
SplitExpatName(aName, getter_AddRefs(localName));
const char* attrName;
localName->GetUTF8String(&attrName);
NS_ConvertUTF16toUTF8 propertyStr(nameSpaceURI);
propertyStr.Append(attrName);
nsCOMPtr<nsIRDFResource> property;
rv = gRDFService->GetResource(propertyStr, getter_AddRefs(property));
if (NS_FAILED(rv)) return rv;
// See if they've specified a 'resource' attribute, in which case
// they mean *that* to be the object of this property.
nsCOMPtr<nsIRDFResource> target;
GetResourceAttribute(aAttributes, getter_AddRefs(target));
PRBool isAnonymous = PR_FALSE;
if (! target) {
// See if an 'ID' attribute has been specified, in which case
// this corresponds to the fourth form of [6.12].
// XXX strictly speaking, we should reject the RDF/XML as
// invalid if they've specified both an 'ID' and a 'resource'
// attribute. Bah.
// XXX strictly speaking, 'about=' isn't allowed here, but
// what the hell.
GetIdAboutAttribute(aAttributes, getter_AddRefs(target), &isAnonymous);
}
if (target) {
// They specified an inline resource for the value of this
// property. Create an RDF resource for the inline resource
// URI, add the properties to it, and attach the inline
// resource to its parent.
PRInt32 count;
rv = AddProperties(aAttributes, target, &count);
NS_ASSERTION(NS_SUCCEEDED(rv), "problem adding properties");
if (NS_FAILED(rv)) return rv;
if (count || !isAnonymous) {
// If the resource was "anonymous" (i.e., they hadn't
// explicitly set an ID or resource attribute), then we'll
// only assert this property from the context element *if*
// there were properties specified on the anonymous
// resource.
rv = mDataSource->Assert(GetContextElement(0), property, target, PR_TRUE);
if (NS_FAILED(rv)) return rv;
}
// XXX Technically, we should _not_ fall through here and push
// the element onto the stack: this is supposed to be a closed
// node. But right now I'm lazy and the code will just Do The
// Right Thing so long as the RDF is well-formed.
}
// Push the element onto the context stack and change state.
PushContext(property, mState, mParseMode);
mState = eRDFContentSinkState_InPropertyElement;
SetParseMode(aAttributes);
return NS_OK;
}
nsresult
RDFContentSinkImpl::OpenObject(const PRUnichar* aName,
const PRUnichar** aAttributes)
{
// an "object" non-terminal is either a "description", a "typed
// node", or a "container", so this change the content sink's
// state appropriately.
nsCOMPtr<nsIAtom> localName;
const nsDependentSubstring& nameSpaceURI =
SplitExpatName(aName, getter_AddRefs(localName));
// Figure out the URI of this object, and create an RDF node for it.
nsCOMPtr<nsIRDFResource> source;
GetIdAboutAttribute(aAttributes, getter_AddRefs(source));
// If there is no `ID' or `about', then there's not much we can do.
if (! source)
return NS_ERROR_FAILURE;
// Push the element onto the context stack
PushContext(source, mState, mParseMode);
// Now figure out what kind of state transition we need to
// make. We'll either be going into a mode where we parse a
// description or a container.
PRBool isaTypedNode = PR_TRUE;
if (nameSpaceURI.EqualsLiteral(RDF_NAMESPACE_URI)) {
isaTypedNode = PR_FALSE;
if (localName == kDescriptionAtom) {
// it's a description
mState = eRDFContentSinkState_InDescriptionElement;
}
else if (localName == kBagAtom) {
// it's a bag container
InitContainer(kRDF_Bag, source);
mState = eRDFContentSinkState_InContainerElement;
}
else if (localName == kSeqAtom) {
// it's a seq container
InitContainer(kRDF_Seq, source);
mState = eRDFContentSinkState_InContainerElement;
}
else if (localName == kAltAtom) {
// it's an alt container
InitContainer(kRDF_Alt, source);
mState = eRDFContentSinkState_InContainerElement;
}
else {
// heh, that's not *in* the RDF namespace: just treat it
// like a typed node
isaTypedNode = PR_TRUE;
}
}
if (isaTypedNode) {
const char* attrName;
localName->GetUTF8String(&attrName);
NS_ConvertUTF16toUTF8 typeStr(nameSpaceURI);
typeStr.Append(attrName);
nsCOMPtr<nsIRDFResource> type;
nsresult rv = gRDFService->GetResource(typeStr, getter_AddRefs(type));
if (NS_FAILED(rv)) return rv;
rv = mDataSource->Assert(source, kRDF_type, type, PR_TRUE);
if (NS_FAILED(rv)) return rv;
mState = eRDFContentSinkState_InDescriptionElement;
}
AddProperties(aAttributes, source);
return NS_OK;
}
