blob: 722417b999cd3e30662e8de580b0885fb29f6e1b [file] [log] [blame]
/*
* Copyright (C) 2005 Frerich Raabe <raabe@kde.org>
* Copyright (C) 2006, 2009 Apple Inc. All rights reserved.
* Copyright (C) 2007 Alexey Proskuryakov <ap@webkit.org>
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "config.h"
#include "XPathStep.h"
#include "Attr.h"
#include "Document.h"
#include "HTMLElement.h"
#include "NodeTraversal.h"
#include "XMLNSNames.h"
#include "XPathParser.h"
#include "XPathUtil.h"
namespace WebCore {
namespace XPath {
Step::Step(Axis axis, NodeTest nodeTest)
: m_axis(axis)
, m_nodeTest(WTFMove(nodeTest))
{
}
Step::Step(Axis axis, NodeTest nodeTest, Vector<std::unique_ptr<Expression>> predicates)
: m_axis(axis)
, m_nodeTest(WTFMove(nodeTest))
, m_predicates(WTFMove(predicates))
{
}
Step::~Step() = default;
void Step::optimize()
{
// Evaluate predicates as part of node test if possible to avoid building unnecessary NodeSets.
// E.g., there is no need to build a set of all "foo" nodes to evaluate "foo[@bar]", we can check the predicate while enumerating.
// This optimization can be applied to predicates that are not context node list sensitive, or to first predicate that is only context position sensitive, e.g. foo[position() mod 2 = 0].
Vector<std::unique_ptr<Expression>> remainingPredicates;
for (auto& predicate : m_predicates) {
if ((!predicateIsContextPositionSensitive(*predicate) || m_nodeTest.m_mergedPredicates.isEmpty()) && !predicate->isContextSizeSensitive() && remainingPredicates.isEmpty())
m_nodeTest.m_mergedPredicates.append(WTFMove(predicate));
else
remainingPredicates.append(WTFMove(predicate));
}
m_predicates = WTFMove(remainingPredicates);
}
void optimizeStepPair(Step& first, Step& second, bool& dropSecondStep)
{
dropSecondStep = false;
if (first.m_axis != Step::DescendantOrSelfAxis)
return;
if (first.m_nodeTest.m_kind != Step::NodeTest::AnyNodeTest)
return;
if (!first.m_predicates.isEmpty())
return;
if (!first.m_nodeTest.m_mergedPredicates.isEmpty())
return;
ASSERT(first.m_nodeTest.m_data.isEmpty());
ASSERT(first.m_nodeTest.m_namespaceURI.isEmpty());
// Optimize the common case of "//" AKA /descendant-or-self::node()/child::NodeTest to /descendant::NodeTest.
if (second.m_axis != Step::ChildAxis)
return;
if (!second.predicatesAreContextListInsensitive())
return;
first.m_axis = Step::DescendantAxis;
first.m_nodeTest = WTFMove(second.m_nodeTest);
first.m_predicates = WTFMove(second.m_predicates);
first.optimize();
dropSecondStep = true;
}
bool Step::predicatesAreContextListInsensitive() const
{
for (auto& predicate : m_predicates) {
if (predicateIsContextPositionSensitive(*predicate) || predicate->isContextSizeSensitive())
return false;
}
for (auto& predicate : m_nodeTest.m_mergedPredicates) {
if (predicateIsContextPositionSensitive(*predicate) || predicate->isContextSizeSensitive())
return false;
}
return true;
}
void Step::evaluate(Node& context, NodeSet& nodes) const
{
EvaluationContext& evaluationContext = Expression::evaluationContext();
evaluationContext.position = 0;
nodesInAxis(context, nodes);
// Check predicates that couldn't be merged into node test.
for (auto& predicate : m_predicates) {
NodeSet newNodes;
if (!nodes.isSorted())
newNodes.markSorted(false);
for (unsigned j = 0; j < nodes.size(); j++) {
Node* node = nodes[j];
evaluationContext.node = node;
evaluationContext.size = nodes.size();
evaluationContext.position = j + 1;
if (evaluatePredicate(*predicate))
newNodes.append(node);
}
nodes = WTFMove(newNodes);
}
}
#if ASSERT_ENABLED
static inline Node::NodeType primaryNodeType(Step::Axis axis)
{
switch (axis) {
case Step::AttributeAxis:
return Node::ATTRIBUTE_NODE;
default:
return Node::ELEMENT_NODE;
}
}
#endif // ASSERT_ENABLED
// Evaluate NodeTest without considering merged predicates.
inline bool nodeMatchesBasicTest(Node& node, Step::Axis axis, const Step::NodeTest& nodeTest)
{
switch (nodeTest.m_kind) {
case Step::NodeTest::TextNodeTest:
return node.nodeType() == Node::TEXT_NODE || node.nodeType() == Node::CDATA_SECTION_NODE;
case Step::NodeTest::CommentNodeTest:
return node.nodeType() == Node::COMMENT_NODE;
case Step::NodeTest::ProcessingInstructionNodeTest: {
const AtomString& name = nodeTest.m_data;
return node.nodeType() == Node::PROCESSING_INSTRUCTION_NODE && (name.isEmpty() || node.nodeName() == name);
}
case Step::NodeTest::AnyNodeTest:
return true;
case Step::NodeTest::NameTest: {
const AtomString& name = nodeTest.m_data;
const AtomString& namespaceURI = nodeTest.m_namespaceURI;
if (axis == Step::AttributeAxis) {
ASSERT(node.isAttributeNode());
// In XPath land, namespace nodes are not accessible on the attribute axis.
if (node.namespaceURI() == XMLNSNames::xmlnsNamespaceURI)
return false;
if (name == starAtom())
return namespaceURI.isEmpty() || node.namespaceURI() == namespaceURI;
return node.localName() == name && node.namespaceURI() == namespaceURI;
}
// Node test on the namespace axis is not implemented yet, the caller has a check for it.
ASSERT(axis != Step::NamespaceAxis);
// For other axes, the principal node type is element.
ASSERT(primaryNodeType(axis) == Node::ELEMENT_NODE);
if (!is<Element>(node))
return false;
if (name == starAtom())
return namespaceURI.isEmpty() || namespaceURI == node.namespaceURI();
if (node.document().isHTMLDocument()) {
if (is<HTMLElement>(node)) {
// Paths without namespaces should match HTML elements in HTML documents despite those having an XHTML namespace. Names are compared case-insensitively.
return equalIgnoringASCIICase(downcast<HTMLElement>(node).localName(), name) && (namespaceURI.isNull() || namespaceURI == node.namespaceURI());
}
// An expression without any prefix shouldn't match no-namespace nodes (because HTML5 says so).
return downcast<Element>(node).hasLocalName(name) && namespaceURI == node.namespaceURI() && !namespaceURI.isNull();
}
return downcast<Element>(node).hasLocalName(name) && namespaceURI == node.namespaceURI();
}
}
ASSERT_NOT_REACHED();
return false;
}
inline bool nodeMatches(Node& node, Step::Axis axis, const Step::NodeTest& nodeTest)
{
if (!nodeMatchesBasicTest(node, axis, nodeTest))
return false;
EvaluationContext& evaluationContext = Expression::evaluationContext();
// Only the first merged predicate may depend on position.
++evaluationContext.position;
for (auto& predicate : nodeTest.m_mergedPredicates) {
// No need to set context size - we only get here when evaluating predicates that do not depend on it.
evaluationContext.node = &node;
if (!evaluatePredicate(*predicate))
return false;
}
return true;
}
// Result nodes are ordered in axis order. Node test (including merged predicates) is applied.
void Step::nodesInAxis(Node& context, NodeSet& nodes) const
{
ASSERT(nodes.isEmpty());
switch (m_axis) {
case ChildAxis:
if (context.isAttributeNode()) // In XPath model, attribute nodes do not have children.
return;
for (Node* node = context.firstChild(); node; node = node->nextSibling()) {
if (nodeMatches(*node, ChildAxis, m_nodeTest))
nodes.append(node);
}
return;
case DescendantAxis:
if (context.isAttributeNode()) // In XPath model, attribute nodes do not have children.
return;
for (Node* node = context.firstChild(); node; node = NodeTraversal::next(*node, &context)) {
if (nodeMatches(*node, DescendantAxis, m_nodeTest))
nodes.append(node);
}
return;
case ParentAxis:
if (context.isAttributeNode()) {
Element* node = static_cast<Attr&>(context).ownerElement();
if (nodeMatches(*node, ParentAxis, m_nodeTest))
nodes.append(node);
} else {
ContainerNode* node = context.parentNode();
if (node && nodeMatches(*node, ParentAxis, m_nodeTest))
nodes.append(node);
}
return;
case AncestorAxis: {
Node* node = &context;
if (context.isAttributeNode()) {
node = static_cast<Attr&>(context).ownerElement();
if (nodeMatches(*node, AncestorAxis, m_nodeTest))
nodes.append(node);
}
for (node = node->parentNode(); node; node = node->parentNode()) {
if (nodeMatches(*node, AncestorAxis, m_nodeTest))
nodes.append(node);
}
nodes.markSorted(false);
return;
}
case FollowingSiblingAxis:
if (context.isAttributeNode())
return;
for (Node* node = context.nextSibling(); node; node = node->nextSibling()) {
if (nodeMatches(*node, FollowingSiblingAxis, m_nodeTest))
nodes.append(node);
}
return;
case PrecedingSiblingAxis:
if (context.isAttributeNode())
return;
for (Node* node = context.previousSibling(); node; node = node->previousSibling()) {
if (nodeMatches(*node, PrecedingSiblingAxis, m_nodeTest))
nodes.append(node);
}
nodes.markSorted(false);
return;
case FollowingAxis:
if (context.isAttributeNode()) {
Node* node = static_cast<Attr&>(context).ownerElement();
while ((node = NodeTraversal::next(*node))) {
if (nodeMatches(*node, FollowingAxis, m_nodeTest))
nodes.append(node);
}
} else {
for (Node* parent = &context; !isRootDomNode(parent); parent = parent->parentNode()) {
for (Node* node = parent->nextSibling(); node; node = node->nextSibling()) {
if (nodeMatches(*node, FollowingAxis, m_nodeTest))
nodes.append(node);
for (Node* child = node->firstChild(); child; child = NodeTraversal::next(*child, node)) {
if (nodeMatches(*child, FollowingAxis, m_nodeTest))
nodes.append(child);
}
}
}
}
return;
case PrecedingAxis: {
Node* node;
if (context.isAttributeNode())
node = static_cast<Attr&>(context).ownerElement();
else
node = &context;
while (ContainerNode* parent = node->parentNode()) {
for (node = NodeTraversal::previous(*node); node != parent; node = NodeTraversal::previous(*node)) {
if (nodeMatches(*node, PrecedingAxis, m_nodeTest))
nodes.append(node);
}
node = parent;
}
nodes.markSorted(false);
return;
}
case AttributeAxis: {
if (!is<Element>(context))
return;
Element& contextElement = downcast<Element>(context);
// Avoid lazily creating attribute nodes for attributes that we do not need anyway.
if (m_nodeTest.m_kind == NodeTest::NameTest && m_nodeTest.m_data != starAtom()) {
auto attr = contextElement.getAttributeNodeNS(m_nodeTest.m_namespaceURI, m_nodeTest.m_data);
if (attr && attr->namespaceURI() != XMLNSNames::xmlnsNamespaceURI) { // In XPath land, namespace nodes are not accessible on the attribute axis.
if (nodeMatches(*attr, AttributeAxis, m_nodeTest)) // Still need to check merged predicates.
nodes.append(WTFMove(attr));
}
return;
}
if (!contextElement.hasAttributes())
return;
for (const Attribute& attribute : contextElement.attributesIterator()) {
auto attr = contextElement.ensureAttr(attribute.name());
if (nodeMatches(attr.get(), AttributeAxis, m_nodeTest))
nodes.append(WTFMove(attr));
}
return;
}
case NamespaceAxis:
// XPath namespace nodes are not implemented yet.
return;
case SelfAxis:
if (nodeMatches(context, SelfAxis, m_nodeTest))
nodes.append(&context);
return;
case DescendantOrSelfAxis:
if (nodeMatches(context, DescendantOrSelfAxis, m_nodeTest))
nodes.append(&context);
if (context.isAttributeNode()) // In XPath model, attribute nodes do not have children.
return;
for (Node* node = context.firstChild(); node; node = NodeTraversal::next(*node, &context)) {
if (nodeMatches(*node, DescendantOrSelfAxis, m_nodeTest))
nodes.append(node);
}
return;
case AncestorOrSelfAxis: {
if (nodeMatches(context, AncestorOrSelfAxis, m_nodeTest))
nodes.append(&context);
Node* node = &context;
if (context.isAttributeNode()) {
node = static_cast<Attr&>(context).ownerElement();
if (nodeMatches(*node, AncestorOrSelfAxis, m_nodeTest))
nodes.append(node);
}
for (node = node->parentNode(); node; node = node->parentNode()) {
if (nodeMatches(*node, AncestorOrSelfAxis, m_nodeTest))
nodes.append(node);
}
nodes.markSorted(false);
return;
}
}
ASSERT_NOT_REACHED();
}
}
}