| /* |
| * Copyright (C) 2012 Google Inc. All rights reserved. |
| * |
| * Redistribution and use in source and binary forms, with or without |
| * modification, are permitted provided that the following conditions are |
| * met: |
| * |
| * * Redistributions of source code must retain the above copyright |
| * notice, this list of conditions and the following disclaimer. |
| * * 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. |
| * * Neither the name of Google Inc. nor the names of its |
| * contributors may be used to endorse or promote products derived from |
| * this software without specific prior written permission. |
| * |
| * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
| * "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 COPYRIGHT |
| * OWNER OR CONTRIBUTORS 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 "DOMPatchSupport.h" |
| |
| #include "Attribute.h" |
| #include "DOMEditor.h" |
| #include "Document.h" |
| #include "DocumentFragment.h" |
| #include "HTMLDocument.h" |
| #include "HTMLDocumentParser.h" |
| #include "HTMLElement.h" |
| #include "HTMLHeadElement.h" |
| #include "HTMLNames.h" |
| #include "InspectorHistory.h" |
| #include "Node.h" |
| #include "XMLDocument.h" |
| #include "XMLDocumentParser.h" |
| #include <wtf/Deque.h> |
| #include <wtf/HashTraits.h> |
| #include <wtf/RefPtr.h> |
| #include <wtf/SHA1.h> |
| #include <wtf/text/Base64.h> |
| #include <wtf/text/CString.h> |
| |
| namespace WebCore { |
| |
| using HTMLNames::bodyTag; |
| using HTMLNames::headTag; |
| using HTMLNames::htmlTag; |
| |
| struct DOMPatchSupport::Digest { |
| WTF_MAKE_STRUCT_FAST_ALLOCATED; |
| |
| String sha1; |
| String attrsSHA1; |
| Node* node; |
| Vector<std::unique_ptr<Digest>> children; |
| }; |
| |
| DOMPatchSupport::DOMPatchSupport(DOMEditor& domEditor, Document& document) |
| : m_domEditor(domEditor) |
| , m_document(document) |
| { |
| } |
| |
| void DOMPatchSupport::patchDocument(const String& markup) |
| { |
| RefPtr<Document> newDocument; |
| if (m_document.isHTMLDocument()) |
| newDocument = HTMLDocument::create(nullptr, URL()); |
| else if (m_document.isXHTMLDocument()) |
| newDocument = XMLDocument::createXHTML(nullptr, URL()); |
| else if (m_document.isSVGDocument()) |
| newDocument = XMLDocument::create(nullptr, URL()); |
| |
| ASSERT(newDocument); |
| RefPtr<DocumentParser> parser; |
| if (newDocument->isHTMLDocument()) |
| parser = HTMLDocumentParser::create(static_cast<HTMLDocument&>(*newDocument)); |
| else |
| parser = XMLDocumentParser::create(*newDocument, nullptr); |
| parser->insert(markup); // Use insert() so that the parser will not yield. |
| parser->finish(); |
| parser->detach(); |
| |
| if (!m_document.documentElement()) |
| return; |
| if (!newDocument->documentElement()) |
| return; |
| |
| std::unique_ptr<Digest> oldInfo = createDigest(*m_document.documentElement(), nullptr); |
| std::unique_ptr<Digest> newInfo = createDigest(*newDocument->documentElement(), &m_unusedNodesMap); |
| |
| if (innerPatchNode(*oldInfo, *newInfo).hasException()) { |
| // Fall back to rewrite. |
| m_document.write(nullptr, markup); |
| m_document.close(); |
| } |
| } |
| |
| ExceptionOr<Node*> DOMPatchSupport::patchNode(Node& node, const String& markup) |
| { |
| // Don't parse <html> as a fragment. |
| if (node.isDocumentNode() || (node.parentNode() && node.parentNode()->isDocumentNode())) { |
| patchDocument(markup); |
| return nullptr; |
| } |
| |
| Node* previousSibling = node.previousSibling(); |
| // FIXME: This code should use one of createFragment* in markup.h |
| auto fragment = DocumentFragment::create(m_document); |
| if (m_document.isHTMLDocument()) |
| fragment->parseHTML(markup, node.parentElement() ? node.parentElement() : m_document.documentElement()); |
| else |
| fragment->parseXML(markup, node.parentElement() ? node.parentElement() : m_document.documentElement()); |
| |
| // Compose the old list. |
| auto* parentNode = node.parentNode(); |
| Vector<std::unique_ptr<Digest>> oldList; |
| for (Node* child = parentNode->firstChild(); child; child = child->nextSibling()) |
| oldList.append(createDigest(*child, nullptr)); |
| |
| // Compose the new list. |
| Vector<std::unique_ptr<Digest>> newList; |
| for (Node* child = parentNode->firstChild(); child != &node; child = child->nextSibling()) |
| newList.append(createDigest(*child, nullptr)); |
| for (Node* child = fragment->firstChild(); child; child = child->nextSibling()) { |
| if (child->hasTagName(headTag) && !child->firstChild() && !markup.containsIgnoringASCIICase("</head>")) |
| continue; // HTML5 parser inserts empty <head> tag whenever it parses <body> |
| if (child->hasTagName(bodyTag) && !child->firstChild() && !markup.containsIgnoringASCIICase("</body>")) |
| continue; // HTML5 parser inserts empty <body> tag whenever it parses </head> |
| newList.append(createDigest(*child, &m_unusedNodesMap)); |
| } |
| for (Node* child = node.nextSibling(); child; child = child->nextSibling()) |
| newList.append(createDigest(*child, nullptr)); |
| |
| if (innerPatchChildren(*parentNode, oldList, newList).hasException()) { |
| // Fall back to total replace. |
| auto result = m_domEditor.replaceChild(*parentNode, fragment.get(), node); |
| if (result.hasException()) |
| return result.releaseException(); |
| } |
| return previousSibling ? previousSibling->nextSibling() : parentNode->firstChild(); |
| } |
| |
| ExceptionOr<void> DOMPatchSupport::innerPatchNode(Digest& oldDigest, Digest& newDigest) |
| { |
| if (oldDigest.sha1 == newDigest.sha1) |
| return { }; |
| |
| auto& oldNode = *oldDigest.node; |
| auto& newNode = *newDigest.node; |
| |
| if (newNode.nodeType() != oldNode.nodeType() || newNode.nodeName() != oldNode.nodeName()) |
| return m_domEditor.replaceChild(*oldNode.parentNode(), newNode, oldNode); |
| |
| if (oldNode.nodeValue() != newNode.nodeValue()) { |
| auto result = m_domEditor.setNodeValue(oldNode, newNode.nodeValue()); |
| if (result.hasException()) |
| return result.releaseException(); |
| } |
| |
| if (!is<Element>(oldNode)) |
| return { }; |
| |
| // Patch attributes |
| auto& oldElement = downcast<Element>(oldNode); |
| auto& newElement = downcast<Element>(newNode); |
| if (oldDigest.attrsSHA1 != newDigest.attrsSHA1) { |
| // FIXME: Create a function in Element for removing all properties. Take in account whether did/willModifyAttribute are important. |
| if (oldElement.hasAttributesWithoutUpdate()) { |
| while (oldElement.attributeCount()) { |
| auto result = m_domEditor.removeAttribute(oldElement, oldElement.attributeAt(0).localName()); |
| if (result.hasException()) |
| return result.releaseException(); |
| } |
| } |
| |
| // FIXME: Create a function in Element for copying properties. cloneDataFromElement() is close but not enough for this case. |
| if (newElement.hasAttributesWithoutUpdate()) { |
| for (auto& attribute : newElement.attributesIterator()) { |
| auto result = m_domEditor.setAttribute(oldElement, attribute.name().localName(), attribute.value()); |
| if (result.hasException()) |
| return result.releaseException(); |
| } |
| } |
| } |
| |
| auto result = innerPatchChildren(oldElement, oldDigest.children, newDigest.children); |
| m_unusedNodesMap.remove(newDigest.sha1); |
| return result; |
| } |
| |
| std::pair<DOMPatchSupport::ResultMap, DOMPatchSupport::ResultMap> |
| DOMPatchSupport::diff(const Vector<std::unique_ptr<Digest>>& oldList, const Vector<std::unique_ptr<Digest>>& newList) |
| { |
| ResultMap newMap(newList.size()); |
| ResultMap oldMap(oldList.size()); |
| |
| for (auto& result : oldMap) { |
| result.first = nullptr; |
| result.second = 0; |
| } |
| |
| for (auto& result : newMap) { |
| result.first = nullptr; |
| result.second = 0; |
| } |
| |
| // Trim head and tail. |
| for (size_t i = 0; i < oldList.size() && i < newList.size() && oldList[i]->sha1 == newList[i]->sha1; ++i) { |
| oldMap[i].first = oldList[i].get(); |
| oldMap[i].second = i; |
| newMap[i].first = newList[i].get(); |
| newMap[i].second = i; |
| } |
| for (size_t i = 0; i < oldList.size() && i < newList.size() && oldList[oldList.size() - i - 1]->sha1 == newList[newList.size() - i - 1]->sha1; ++i) { |
| size_t oldIndex = oldList.size() - i - 1; |
| size_t newIndex = newList.size() - i - 1; |
| oldMap[oldIndex].first = oldList[oldIndex].get(); |
| oldMap[oldIndex].second = newIndex; |
| newMap[newIndex].first = newList[newIndex].get(); |
| newMap[newIndex].second = oldIndex; |
| } |
| |
| typedef HashMap<String, Vector<size_t>> DiffTable; |
| DiffTable newTable; |
| DiffTable oldTable; |
| |
| for (size_t i = 0; i < newList.size(); ++i) |
| newTable.add(newList[i]->sha1, Vector<size_t>()).iterator->value.append(i); |
| |
| for (size_t i = 0; i < oldList.size(); ++i) |
| oldTable.add(oldList[i]->sha1, Vector<size_t>()).iterator->value.append(i); |
| |
| for (auto& newEntry : newTable) { |
| if (newEntry.value.size() != 1) |
| continue; |
| |
| auto oldIt = oldTable.find(newEntry.key); |
| if (oldIt == oldTable.end() || oldIt->value.size() != 1) |
| continue; |
| |
| newMap[newEntry.value[0]] = std::make_pair(newList[newEntry.value[0]].get(), oldIt->value[0]); |
| oldMap[oldIt->value[0]] = std::make_pair(oldList[oldIt->value[0]].get(), newEntry.value[0]); |
| } |
| |
| for (size_t i = 0; newList.size() > 0 && i < newList.size() - 1; ++i) { |
| if (!newMap[i].first || newMap[i + 1].first) |
| continue; |
| |
| size_t j = newMap[i].second + 1; |
| if (j < oldMap.size() && !oldMap[j].first && newList[i + 1]->sha1 == oldList[j]->sha1) { |
| newMap[i + 1] = std::make_pair(newList[i + 1].get(), j); |
| oldMap[j] = std::make_pair(oldList[j].get(), i + 1); |
| } |
| } |
| |
| for (size_t i = newList.size() - 1; newList.size() > 0 && i > 0; --i) { |
| if (!newMap[i].first || newMap[i - 1].first || newMap[i].second <= 0) |
| continue; |
| |
| size_t j = newMap[i].second - 1; |
| if (!oldMap[j].first && newList[i - 1]->sha1 == oldList[j]->sha1) { |
| newMap[i - 1] = std::make_pair(newList[i - 1].get(), j); |
| oldMap[j] = std::make_pair(oldList[j].get(), i - 1); |
| } |
| } |
| |
| #ifdef DEBUG_DOM_PATCH_SUPPORT |
| dumpMap(oldMap, "OLD"); |
| dumpMap(newMap, "NEW"); |
| #endif |
| |
| return std::make_pair(oldMap, newMap); |
| } |
| |
| ExceptionOr<void> DOMPatchSupport::innerPatchChildren(ContainerNode& parentNode, const Vector<std::unique_ptr<Digest>>& oldList, const Vector<std::unique_ptr<Digest>>& newList) |
| { |
| auto resultMaps = diff(oldList, newList); |
| ResultMap& oldMap = resultMaps.first; |
| ResultMap& newMap = resultMaps.second; |
| |
| Digest* oldHead = nullptr; |
| Digest* oldBody = nullptr; |
| |
| // 1. First strip everything except for the nodes that retain. Collect pending merges. |
| HashMap<Digest*, Digest*> merges; |
| HashSet<size_t, WTF::IntHash<size_t>, WTF::UnsignedWithZeroKeyHashTraits<size_t>> usedNewOrdinals; |
| for (size_t i = 0; i < oldList.size(); ++i) { |
| if (oldMap[i].first) { |
| if (!usedNewOrdinals.contains(oldMap[i].second)) { |
| usedNewOrdinals.add(oldMap[i].second); |
| continue; |
| } |
| oldMap[i].first = nullptr; |
| oldMap[i].second = 0; |
| } |
| |
| // Always match <head> and <body> tags with each other - we can't remove them from the DOM |
| // upon patching. |
| if (oldList[i]->node->hasTagName(headTag)) { |
| oldHead = oldList[i].get(); |
| continue; |
| } |
| if (oldList[i]->node->hasTagName(bodyTag)) { |
| oldBody = oldList[i].get(); |
| continue; |
| } |
| |
| // Check if this change is between stable nodes. If it is, consider it as "modified". |
| if (!m_unusedNodesMap.contains(oldList[i]->sha1) && (!i || oldMap[i - 1].first) && (i == oldMap.size() - 1 || oldMap[i + 1].first)) { |
| size_t anchorCandidate = i ? oldMap[i - 1].second + 1 : 0; |
| size_t anchorAfter = i == oldMap.size() - 1 ? anchorCandidate + 1 : oldMap[i + 1].second; |
| if (anchorAfter - anchorCandidate == 1 && anchorCandidate < newList.size()) |
| merges.set(newList[anchorCandidate].get(), oldList[i].get()); |
| else { |
| auto result = removeChildAndMoveToNew(*oldList[i]); |
| if (result.hasException()) |
| return result.releaseException(); |
| } |
| } else { |
| auto result = removeChildAndMoveToNew(*oldList[i]); |
| if (result.hasException()) |
| return result.releaseException(); |
| } |
| } |
| |
| // Mark retained nodes as used, do not reuse node more than once. |
| HashSet<size_t, WTF::IntHash<size_t>, WTF::UnsignedWithZeroKeyHashTraits<size_t>> usedOldOrdinals; |
| for (size_t i = 0; i < newList.size(); ++i) { |
| if (!newMap[i].first) |
| continue; |
| size_t oldOrdinal = newMap[i].second; |
| if (usedOldOrdinals.contains(oldOrdinal)) { |
| // Do not map node more than once |
| newMap[i].first = nullptr; |
| newMap[i].second = 0; |
| continue; |
| } |
| usedOldOrdinals.add(oldOrdinal); |
| markNodeAsUsed(*newMap[i].first); |
| } |
| |
| // Mark <head> and <body> nodes for merge. |
| if (oldHead || oldBody) { |
| for (size_t i = 0; i < newList.size(); ++i) { |
| if (oldHead && newList[i]->node->hasTagName(headTag)) |
| merges.set(newList[i].get(), oldHead); |
| if (oldBody && newList[i]->node->hasTagName(bodyTag)) |
| merges.set(newList[i].get(), oldBody); |
| } |
| } |
| |
| // 2. Patch nodes marked for merge. |
| for (auto& merge : merges) { |
| auto result = innerPatchNode(*merge.value, *merge.key); |
| if (result.hasException()) |
| return result.releaseException(); |
| } |
| |
| // 3. Insert missing nodes. |
| for (size_t i = 0; i < newMap.size(); ++i) { |
| if (newMap[i].first || merges.contains(newList[i].get())) |
| continue; |
| auto result = insertBeforeAndMarkAsUsed(parentNode, *newList[i], parentNode.traverseToChildAt(i)); |
| if (result.hasException()) |
| return result.releaseException(); |
| } |
| |
| // 4. Then put all nodes that retained into their slots (sort by new index). |
| for (size_t i = 0; i < oldMap.size(); ++i) { |
| if (!oldMap[i].first) |
| continue; |
| RefPtr<Node> node = oldMap[i].first->node; |
| auto* anchorNode = parentNode.traverseToChildAt(oldMap[i].second); |
| if (node == anchorNode) |
| continue; |
| if (node->hasTagName(bodyTag) || node->hasTagName(headTag)) |
| continue; // Never move head or body, move the rest of the nodes around them. |
| auto result = m_domEditor.insertBefore(parentNode, node.releaseNonNull(), anchorNode); |
| if (result.hasException()) |
| return result.releaseException(); |
| } |
| return { }; |
| } |
| |
| static void addStringToSHA1(SHA1& sha1, const String& string) |
| { |
| CString cString = string.utf8(); |
| sha1.addBytes(reinterpret_cast<const uint8_t*>(cString.data()), cString.length()); |
| } |
| |
| std::unique_ptr<DOMPatchSupport::Digest> DOMPatchSupport::createDigest(Node& node, UnusedNodesMap* unusedNodesMap) |
| { |
| auto digest = makeUnique<Digest>(); |
| digest->node = &node; |
| SHA1 sha1; |
| |
| auto nodeType = node.nodeType(); |
| sha1.addBytes(reinterpret_cast<const uint8_t*>(&nodeType), sizeof(nodeType)); |
| addStringToSHA1(sha1, node.nodeName()); |
| addStringToSHA1(sha1, node.nodeValue()); |
| |
| if (node.nodeType() == Node::ELEMENT_NODE) { |
| Node* child = node.firstChild(); |
| while (child) { |
| std::unique_ptr<Digest> childInfo = createDigest(*child, unusedNodesMap); |
| addStringToSHA1(sha1, childInfo->sha1); |
| child = child->nextSibling(); |
| digest->children.append(WTFMove(childInfo)); |
| } |
| auto& element = downcast<Element>(node); |
| |
| if (element.hasAttributesWithoutUpdate()) { |
| SHA1 attrsSHA1; |
| for (auto& attribute : element.attributesIterator()) { |
| addStringToSHA1(attrsSHA1, attribute.name().toString()); |
| addStringToSHA1(attrsSHA1, attribute.value()); |
| } |
| SHA1::Digest attrsHash; |
| attrsSHA1.computeHash(attrsHash); |
| digest->attrsSHA1 = base64Encode(attrsHash.data(), 10); |
| addStringToSHA1(sha1, digest->attrsSHA1); |
| } |
| } |
| |
| SHA1::Digest hash; |
| sha1.computeHash(hash); |
| digest->sha1 = base64Encode(hash.data(), 10); |
| if (unusedNodesMap) |
| unusedNodesMap->add(digest->sha1, digest.get()); |
| |
| return digest; |
| } |
| |
| ExceptionOr<void> DOMPatchSupport::insertBeforeAndMarkAsUsed(ContainerNode& parentNode, Digest& digest, Node* anchor) |
| { |
| ASSERT(digest.node); |
| auto result = m_domEditor.insertBefore(parentNode, *digest.node, anchor); |
| markNodeAsUsed(digest); |
| return result; |
| } |
| |
| ExceptionOr<void> DOMPatchSupport::removeChildAndMoveToNew(Digest& oldDigest) |
| { |
| Ref<Node> oldNode = *oldDigest.node; |
| ASSERT(oldNode->parentNode()); |
| auto result = m_domEditor.removeChild(*oldNode->parentNode(), oldNode); |
| if (result.hasException()) |
| return result.releaseException(); |
| |
| // Diff works within levels. In order not to lose the node identity when user |
| // prepends his HTML with "<div>" (i.e. all nodes are shifted to the next nested level), |
| // prior to dropping the original node on the floor, check whether new DOM has a digest |
| // with matching sha1. If it does, replace it with the original DOM chunk. Chances are |
| // high that it will get merged back into the original DOM during the further patching. |
| auto it = m_unusedNodesMap.find(oldDigest.sha1); |
| if (it != m_unusedNodesMap.end()) { |
| auto& newDigest = *it->value; |
| auto& newNode = *newDigest.node; |
| auto result = m_domEditor.replaceChild(*newNode.parentNode(), oldNode.get(), newNode); |
| if (result.hasException()) |
| return result.releaseException(); |
| newDigest.node = oldNode.ptr(); |
| markNodeAsUsed(newDigest); |
| return { }; |
| } |
| |
| for (auto& child : oldDigest.children) { |
| auto result = removeChildAndMoveToNew(*child); |
| if (result.hasException()) |
| return result.releaseException(); |
| } |
| return { }; |
| } |
| |
| void DOMPatchSupport::markNodeAsUsed(Digest& digest) |
| { |
| Deque<Digest*> queue; |
| queue.append(&digest); |
| while (!queue.isEmpty()) { |
| auto& first = *queue.takeFirst(); |
| m_unusedNodesMap.remove(first.sha1); |
| for (auto& child : first.children) |
| queue.append(child.get()); |
| } |
| } |
| |
| #ifdef DEBUG_DOM_PATCH_SUPPORT |
| |
| static String nodeName(Node* node) |
| { |
| if (node->document().isXHTMLDocument()) |
| return node->nodeName(); |
| return node->nodeName().convertToASCIILowercase(); |
| } |
| |
| void DOMPatchSupport::dumpMap(const ResultMap& map, const String& name) |
| { |
| fprintf(stderr, "\n\n"); |
| for (size_t i = 0; i < map.size(); ++i) |
| fprintf(stderr, "%s[%lu]: %s (%p) - [%lu]\n", name.utf8().data(), i, map[i].first ? nodeName(map[i].first->m_node).utf8().data() : "", map[i].first, map[i].second); |
| } |
| |
| #endif |
| |
| } // namespace WebCore |