All Downloads are FREE. Search and download functionalities are using the official Maven repository.

org.htmlunit.cyberneko.xerces.dom.NodeImpl Maven / Gradle / Ivy

There is a newer version: 6.0.36
Show newest version
/*
 * Copyright 2017-2023 Ronald Brill
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 * https://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */
package org.htmlunit.cyberneko.xerces.dom;

import org.w3c.dom.DOMException;
import org.w3c.dom.Document;
import org.w3c.dom.DocumentType;
import org.w3c.dom.Element;
import org.w3c.dom.NamedNodeMap;
import org.w3c.dom.Node;
import org.w3c.dom.NodeList;
import org.w3c.dom.UserDataHandler;
import org.w3c.dom.events.Event;
import org.w3c.dom.events.EventListener;
import org.w3c.dom.events.EventTarget;

/**
 * NodeImpl provides the basic structure of a DOM tree. It is never used
 * directly, but instead is subclassed to add type and data information, and
 * additional methods, appropriate to each node of the tree. Only its subclasses
 * should be instantiated -- and those, with the exception of Document itself,
 * only through a specific Document's factory methods.
 * 

* The Node interface provides shared behaviors such as siblings and children, * both for consistancy and so that the most common tree operations may be * performed without constantly having to downcast to specific node types. When * there is no obvious mapping for one of these queries, it will respond with * null. Note that the default behavior is that children are forbidden. To * permit them, the subclass ParentNode overrides several methods. *

* NodeImpl also implements NodeList, so it can return itself in response to the * getChildNodes() query. This eliminiates the need for a separate ChildNodeList * object. Note that this is an IMPLEMENTATION DETAIL; applications should * _never_ assume that this identity exists. *

* All nodes in a single document must originate in that document. (Note that * this is much tighter than "must be same implementation") Nodes are all aware * of their ownerDocument, and attempts to mismatch will throw * WRONG_DOCUMENT_ERR. *

* However, to save memory not all nodes always have a direct reference to their * ownerDocument. When a node is owned by another node it relies on its owner to * store its ownerDocument. Parent nodes always store it though, so there is * never more than one level of indirection. And when a node doesn't have an * owner, ownerNode refers to its ownerDocument. *

* This class doesn't directly support mutation events, however, it still * implements the EventTarget interface and forward all related calls to the * document so that the document class do so. *

* * @author Arnaud Le Hors, IBM * @author Joe Kesselman, IBM */ public abstract class NodeImpl implements Node, NodeList, EventTarget, Cloneable { // TreePosition Constants. // Taken from DOM L3 Node interface. // DocumentPosition public static final short DOCUMENT_POSITION_DISCONNECTED = 0x01; public static final short DOCUMENT_POSITION_PRECEDING = 0x02; public static final short DOCUMENT_POSITION_FOLLOWING = 0x04; public static final short DOCUMENT_POSITION_CONTAINS = 0x08; public static final short DOCUMENT_POSITION_IS_CONTAINED = 0x10; public static final short DOCUMENT_POSITION_IMPLEMENTATION_SPECIFIC = 0x20; /** Element definition node type. */ public static final short ELEMENT_DEFINITION_NODE = 21; protected NodeImpl ownerNode; // typically the parent but not always! private short flags; protected static final short READONLY = 0x1 << 0; protected static final short SYNCDATA = 0x1 << 1; protected static final short SYNCCHILDREN = 0x1 << 2; protected static final short OWNED = 0x1 << 3; protected static final short FIRSTCHILD = 0x1 << 4; protected static final short SPECIFIED = 0x1 << 5; protected static final short IGNORABLEWS = 0x1 << 6; protected static final short HASSTRING = 0x1 << 7; protected static final short NORMALIZED = 0x1 << 8; protected static final short ID = 0x1 << 9; // // Constructors // /** * No public constructor; only subclasses of Node should be instantiated, and * those normally via a Document's factory methods *

* Every Node knows what Document it belongs to. * * @param ownerDocument the owner document */ protected NodeImpl(final CoreDocumentImpl ownerDocument) { // as long as we do not have any owner, ownerNode is our ownerDocument ownerNode = ownerDocument; } // (CoreDocumentImpl) /** * {@inheritDoc} * * A short integer indicating what type of node this is. The named constants for * this value are defined in the org.w3c.dom.Node interface. */ @Override public abstract short getNodeType(); /** * {@inheritDoc} * * the name of this node. */ @Override public abstract String getNodeName(); /** * {@inheritDoc} * * Returns the node value. * * @throws DOMException DOMSTRING_SIZE_ERR */ @Override public String getNodeValue() throws DOMException { return null; // overridden in some subclasses } /** * {@inheritDoc} * * Sets the node value. * * @throws DOMException NO_MODIFICATION_ALLOWED_ERR */ @Override public void setNodeValue(final String x) throws DOMException { // Default behavior is to do nothing, overridden in some subclasses } /** * {@inheritDoc} * * Adds a child node to the end of the list of children for this node. * Convenience shorthand for insertBefore(newChild,null). * * @see #insertBefore(Node, Node) *

* By default we do not accept any children, ParentNode overrides this. * @see ParentNode * * @return newChild, in its new state (relocated, or emptied in the case of * DocumentNode.) * * @throws DOMException HIERARCHY_REQUEST_ERR if newChild is of a type that * shouldn't be a child of this node. * * @throws DOMException WRONG_DOCUMENT_ERR if newChild has a different owner * document than we do. * * @throws DOMException NO_MODIFICATION_ALLOWED_ERR if this node is read-only. */ @Override public Node appendChild(final Node newChild) throws DOMException { return insertBefore(newChild, null); } /** * {@inheritDoc} * * Returns a duplicate of a given node. You can consider this a generic "copy * constructor" for nodes. The newly returned object should be completely * independent of the source object's subtree, so changes in one after the clone * has been made will not affect the other. *

* Note: since we never have any children deep is meaningless here, ParentNode * overrides this behavior. * * @see ParentNode * *

* Example: Cloning a Text node will copy both the node and the text it * contains. *

* Example: Cloning something that has children -- Element or Attr, for * example -- will _not_ clone those children unless a "deep clone" has * been requested. A shallow clone of an Attr node will yield an empty Attr * of the same name. *

* NOTE: Clones will always be read/write, even if the node being cloned is * read-only, to permit applications using only the DOM API to obtain * editable copies of locked portions of the tree. */ @Override public Node cloneNode(final boolean deep) { if (needsSyncData()) { synchronizeData(); } final NodeImpl newnode; try { newnode = (NodeImpl) clone(); } catch (final CloneNotSupportedException e) { // if we get here we have an error in our program we may as well // be vocal about it, so that people can take appropriate action. throw new RuntimeException("**Internal Error**" + e); } // Need to break the association w/ original kids newnode.ownerNode = ownerDocument(); newnode.isOwned(false); return newnode; } // cloneNode(boolean):Node /** * {@inheritDoc} * * Find the Document that this Node belongs to (the document in whose context * the Node was created). The Node may or may not currently be part of that * Document's actual contents. */ @Override public Document getOwnerDocument() { // if we have an owner simply forward the request // otherwise ownerNode is our ownerDocument if (isOwned()) { return ownerNode.ownerDocument(); } return (Document) ownerNode; } /** * same as above but returns internal type and this one is not overridden by * CoreDocumentImpl to return null */ CoreDocumentImpl ownerDocument() { // if we have an owner simply forward the request // otherwise ownerNode is our ownerDocument if (isOwned()) { return ownerNode.ownerDocument(); } return (CoreDocumentImpl) ownerNode; } // NON-DOM set the ownerDocument of this node protected void setOwnerDocument(final CoreDocumentImpl doc) { if (needsSyncData()) { synchronizeData(); } // if we have an owner we rely on it to have it right // otherwise ownerNode is our ownerDocument if (!isOwned()) { ownerNode = doc; } } /** * @return the node number */ protected int getNodeNumber() { final int nodeNumber; final CoreDocumentImpl cd = (CoreDocumentImpl) (this.getOwnerDocument()); nodeNumber = cd.getNodeNumber(this); return nodeNumber; } /** * Obtain the DOM-tree parent of this node, or null if it is not currently * active in the DOM tree (perhaps because it has just been created or removed). * Note that Document, DocumentFragment, and Attribute will never have parents. */ @Override public Node getParentNode() { return null; // overriden by ChildNode } /* * same as above but returns internal type */ NodeImpl parentNode() { return null; } /** The next child of this node's parent, or null if none */ @Override public Node getNextSibling() { return null; // default behavior, overriden in ChildNode } /** The previous child of this node's parent, or null if none */ @Override public Node getPreviousSibling() { return null; // default behavior, overriden in ChildNode } ChildNode previousSibling() { return null; // default behavior, overriden in ChildNode } /** * Return the collection of attributes associated with this node, or null if * none. At this writing, Element is the only type of node which will ever have * attributes. * * @see ElementImpl */ @Override public NamedNodeMap getAttributes() { return null; // overridden in ElementImpl } /** * Returns whether this node (if it is an element) has any attributes. * * @return true if this node has any attributes, false * otherwise. * @see ElementImpl */ @Override public boolean hasAttributes() { return false; // overridden in ElementImpl } /** * Test whether this node has any children. Convenience shorthand for * (Node.getFirstChild()!=null) *

* By default we do not have any children, ParentNode overrides this. * * @see ParentNode */ @Override public boolean hasChildNodes() { return false; } /** * Obtain a NodeList enumerating all children of this node. If there are none, * an (initially) empty NodeList is returned. *

* NodeLists are "live"; as children are added/removed the NodeList will * immediately reflect those changes. Also, the NodeList refers to the actual * nodes, so changes to those nodes made via the DOM tree will be reflected in * the NodeList and vice versa. *

* In this implementation, Nodes implement the NodeList interface and provide * their own getChildNodes() support. Other DOMs may solve this differently. */ @Override public NodeList getChildNodes() { return this; } /** * The first child of this Node, or null if none. *

* By default we do not have any children, ParentNode overrides this. * * @see ParentNode */ @Override public Node getFirstChild() { return null; } /** * The first child of this Node, or null if none. *

* By default we do not have any children, ParentNode overrides this. * * @see ParentNode */ @Override public Node getLastChild() { return null; } /** * Move one or more node(s) to our list of children. Note that this implicitly * removes them from their previous parent. *

* By default we do not accept any children, ParentNode overrides this. * * @see ParentNode * * @param newChild The Node to be moved to our subtree. As a convenience * feature, inserting a DocumentNode will instead insert all its * children. * * @param refChild Current child which newChild should be placed immediately * before. If refChild is null, the insertion occurs after all * existing Nodes, like appendChild(). * * @return newChild, in its new state (relocated, or emptied in the case of * DocumentNode.) * * @throws DOMException HIERARCHY_REQUEST_ERR if newChild is of a type that * shouldn't be a child of this node, or if newChild is an * ancestor of this node. * * @throws DOMException WRONG_DOCUMENT_ERR if newChild has a different owner * document than we do. * * @throws DOMException NOT_FOUND_ERR if refChild is not a child of this node. * * @throws DOMException NO_MODIFICATION_ALLOWED_ERR if this node is read-only. */ @Override public Node insertBefore(final Node newChild, final Node refChild) throws DOMException { throw new DOMException(DOMException.HIERARCHY_REQUEST_ERR, DOMMessageFormatter.formatMessage(DOMMessageFormatter.DOM_DOMAIN, "HIERARCHY_REQUEST_ERR", null)); } /** * Remove a child from this Node. The removed child's subtree remains intact so * it may be re-inserted elsewhere. *

* By default we do not have any children, ParentNode overrides this. * * @see ParentNode * * @return oldChild, in its new state (removed). * * @throws DOMException NOT_FOUND_ERR if oldChild is not a child of this node. * * @throws DOMException NO_MODIFICATION_ALLOWED_ERR if this node is read-only. */ @Override public Node removeChild(final Node oldChild) throws DOMException { throw new DOMException(DOMException.NOT_FOUND_ERR, DOMMessageFormatter.formatMessage(DOMMessageFormatter.DOM_DOMAIN, "NOT_FOUND_ERR", null)); } /** * Make newChild occupy the location that oldChild used to have. Note that * newChild will first be removed from its previous parent, if any. Equivalent * to inserting newChild before oldChild, then removing oldChild. *

* By default we do not have any children, ParentNode overrides this. * * @see ParentNode * * @return oldChild, in its new state (removed). * * @throws DOMException HIERARCHY_REQUEST_ERR if newChild is of a type that * shouldn't be a child of this node, or if newChild is one * of our ancestors. * * @throws DOMException WRONG_DOCUMENT_ERR if newChild has a different owner * document than we do. * * @throws DOMException NOT_FOUND_ERR if oldChild is not a child of this node. * * @throws DOMException NO_MODIFICATION_ALLOWED_ERR if this node is read-only. */ @Override public Node replaceChild(final Node newChild, final Node oldChild) throws DOMException { throw new DOMException(DOMException.HIERARCHY_REQUEST_ERR, DOMMessageFormatter.formatMessage(DOMMessageFormatter.DOM_DOMAIN, "HIERARCHY_REQUEST_ERR", null)); } // // NodeList methods // /** * {@inheritDoc} NodeList method: Count the immediate children of this node *

* By default we do not have any children, ParentNode overrides this. */ @Override public int getLength() { return 0; } /** * {@inheritDoc} NodeList method: Return the Nth immediate child of this node, * or null if the index is out of bounds. *

* By default we do not have any children, ParentNode overrides this. */ @Override public Node item(final int index) { return null; } // // DOM2: methods, getters, setters // /** * Puts all Text nodes in the full depth of the sub-tree underneath * this Node, including attribute nodes, into a "normal" form where * only markup (e.g., tags, comments, processing instructions, CDATA sections, * and entity references) separates Text nodes, i.e., there are no * adjacent Text nodes. This can be used to ensure that the DOM * view of a document is the same as if it were saved and re-loaded, and is * useful when operations (such as XPointer lookups) that depend on a particular * document tree structure are to be used.In cases where the document contains * CDATASections, the normalize operation alone may not be * sufficient, since XPointers do not differentiate between Text * nodes and CDATASection nodes. *

* Note that this implementation simply calls normalize() on this Node's * children. It is up to implementors or Node to override normalize() to take * action. */ @Override public void normalize() { /* * by default we do not have any children, ParentNode overrides this behavior */ } /** * Introduced in DOM Level 2. *

* Tests whether the DOM implementation implements a specific feature and that * feature is supported by this node. * * @param feature The package name of the feature to test. This is the same name * as what can be passed to the method hasFeature on * DOMImplementation. * @param version This is the version number of the package name to test. In * Level 2, version 1, this is the string "2.0". If the version * is not specified, supporting any version of the feature will * cause the method to return true. * @return boolean Returns true if this node defines a subtree within which the * specified feature is supported, false otherwise. */ @Override public boolean isSupported(final String feature, final String version) { return ownerDocument().getImplementation().hasFeature(feature, version); } /** * Introduced in DOM Level 2. *

* * The namespace URI of this node, or null if it is unspecified. When this node * is of any type other than ELEMENT_NODE and ATTRIBUTE_NODE, this is always * null and setting it has no effect. *

* * This is not a computed value that is the result of a namespace lookup based * on an examination of the namespace declarations in scope. It is merely the * namespace URI given at creation time. *

* * For nodes created with a DOM Level 1 method, such as createElement from the * Document interface, this is null. * * @see AttrNSImpl * @see ElementNSImpl */ @Override public String getNamespaceURI() { return null; } /** * {@inheritDoc} * * Introduced in DOM Level 2. *

* * The namespace prefix of this node, or null if it is unspecified. When this * node is of any type other than ELEMENT_NODE and ATTRIBUTE_NODE this is always * null and setting it has no effect. *

* * For nodes created with a DOM Level 1 method, such as createElement from the * Document interface, this is null. * * @see AttrNSImpl * @see ElementNSImpl */ @Override public String getPrefix() { return null; } /** * {@inheritDoc} * * Introduced in DOM Level 2. *

* * The namespace prefix of this node, or null if it is unspecified. When this * node is of any type other than ELEMENT_NODE and ATTRIBUTE_NODE this is always * null and setting it has no effect. *

* * For nodes created with a DOM Level 1 method, such as createElement from the * Document interface, this is null. *

* * Note that setting this attribute changes the nodeName attribute, which holds * the qualified name, as well as the tagName and name attributes of the Element * and Attr interfaces, when applicable. * * @throws DOMException INVALID_CHARACTER_ERR Raised if the specified prefix * contains an invalid character. * * @see AttrNSImpl * @see ElementNSImpl */ @Override public void setPrefix(final String prefix) throws DOMException { throw new DOMException(DOMException.NAMESPACE_ERR, DOMMessageFormatter.formatMessage(DOMMessageFormatter.DOM_DOMAIN, "NAMESPACE_ERR", null)); } /** * {@inheritDoc} * * Introduced in DOM Level 2. *

* * Returns the local part of the qualified name of this node. For nodes created * with a DOM Level 1 method, such as createElement from the Document interface, * and for nodes of any type other than ELEMENT_NODE and ATTRIBUTE_NODE this is * the same as the nodeName attribute. * * @see AttrNSImpl * @see ElementNSImpl */ @Override public String getLocalName() { return null; } // // EventTarget support // @Override public void addEventListener(final String type, final EventListener listener, final boolean useCapture) { // simply forward to Document ownerDocument().addEventListener(this, type, listener, useCapture); } @Override public void removeEventListener(final String type, final EventListener listener, final boolean useCapture) { // simply forward to Document ownerDocument().removeEventListener(this, type, listener, useCapture); } @Override public boolean dispatchEvent(final Event event) { return false; } // // Public DOM Level 3 methods // /** * The absolute base URI of this node or null if undefined. This * value is computed according to . However, when the Document * supports the feature "HTML" , the base URI is computed using first the value * of the href attribute of the HTML BASE element if any, and the value of the * documentURI attribute from the Document interface * otherwise.
* When the node is an Element, a Document or a a * ProcessingInstruction, this attribute represents the properties * [base URI] defined in . When the node is a Notation, an * Entity, or an EntityReference, this attribute * represents the properties [declaration base URI] in the . How will this be * affected by resolution of relative namespace URIs issue?It's not.Should this * only be on Document, Element, ProcessingInstruction, Entity, and Notation * nodes, according to the infoset? If not, what is it equal to on other nodes? * Null? An empty string? I think it should be the parent's.No.Should this be * read-only and computed or and actual read-write attribute?Read-only and * computed (F2F 19 Jun 2000 and teleconference 30 May 2001).If the base HTML * element is not yet attached to a document, does the insert change the * Document.baseURI? Yes. (F2F 26 Sep 2001) */ @Override public String getBaseURI() { return null; } /** * Compares a node with this node with regard to their position in the document. * * @param other The node to compare against this node. * @return Returns how the given node is positioned relatively to this node. */ @Override public short compareDocumentPosition(final Node other) throws DOMException { // If the nodes are the same, no flags should be set if (this == other) { return 0; } // check if other is from a different implementation if (other != null && !(other instanceof NodeImpl)) { // other comes from a different implementation final String msg = DOMMessageFormatter.formatMessage(DOMMessageFormatter.DOM_DOMAIN, "NOT_SUPPORTED_ERR", null); throw new DOMException(DOMException.NOT_SUPPORTED_ERR, msg); } final Document thisOwnerDoc; final Document otherOwnerDoc; // get the respective Document owners. if (this.getNodeType() == Node.DOCUMENT_NODE) { thisOwnerDoc = (Document) this; } else { thisOwnerDoc = this.getOwnerDocument(); } if (other.getNodeType() == Node.DOCUMENT_NODE) { otherOwnerDoc = (Document) other; } else { otherOwnerDoc = other.getOwnerDocument(); } // If from different documents, we know they are disconnected. // and have an implementation dependent order if (thisOwnerDoc != otherOwnerDoc && thisOwnerDoc != null && otherOwnerDoc != null) { final int otherDocNum = ((CoreDocumentImpl) otherOwnerDoc).getNodeNumber(); final int thisDocNum = ((CoreDocumentImpl) thisOwnerDoc).getNodeNumber(); if (otherDocNum > thisDocNum) { return DOCUMENT_POSITION_DISCONNECTED | DOCUMENT_POSITION_FOLLOWING | DOCUMENT_POSITION_IMPLEMENTATION_SPECIFIC; } return DOCUMENT_POSITION_DISCONNECTED | DOCUMENT_POSITION_PRECEDING | DOCUMENT_POSITION_IMPLEMENTATION_SPECIFIC; } // Find the ancestor of each node, and the distance each node is from // its ancestor. // During this traversal, look for ancestor/descendent relationships // between the 2 nodes in question. // We do this now, so that we get this info correct for attribute nodes // and their children. Node node; Node thisAncestor = this; Node otherAncestor = other; int thisDepth = 0; int otherDepth = 0; for (node = this; node != null; node = node.getParentNode()) { thisDepth += 1; if (node == other) { // The other node is an ancestor of this one. return DOCUMENT_POSITION_CONTAINS | DOCUMENT_POSITION_PRECEDING; } thisAncestor = node; } for (node = other; node != null; node = node.getParentNode()) { otherDepth += 1; if (node == this) { // The other node is a descendent of the reference node. return DOCUMENT_POSITION_IS_CONTAINED | DOCUMENT_POSITION_FOLLOWING; } otherAncestor = node; } final int thisAncestorType = thisAncestor.getNodeType(); final int otherAncestorType = otherAncestor.getNodeType(); Node thisNode = this; Node otherNode = other; // Special casing for ENTITY, NOTATION, DOCTYPE and ATTRIBUTES // LM: should rewrite this. switch (thisAncestorType) { case Node.NOTATION_NODE: case Node.ENTITY_NODE: { final DocumentType container = thisOwnerDoc.getDoctype(); if (container == otherAncestor) { return DOCUMENT_POSITION_CONTAINS | DOCUMENT_POSITION_PRECEDING; } switch (otherAncestorType) { case Node.NOTATION_NODE: case Node.ENTITY_NODE: { if (thisAncestorType != otherAncestorType) { // the nodes are of different types return (thisAncestorType > otherAncestorType) ? DOCUMENT_POSITION_PRECEDING : DOCUMENT_POSITION_FOLLOWING; } // the nodes are of the same type. Find order. if (thisAncestorType == Node.NOTATION_NODE) { if (((NamedNodeMapImpl) container.getNotations()).precedes(otherAncestor, thisAncestor)) { return DOCUMENT_POSITION_PRECEDING | DOCUMENT_POSITION_IMPLEMENTATION_SPECIFIC; } return DOCUMENT_POSITION_FOLLOWING | DOCUMENT_POSITION_IMPLEMENTATION_SPECIFIC; } if (((NamedNodeMapImpl) container.getEntities()).precedes(otherAncestor, thisAncestor)) { return DOCUMENT_POSITION_PRECEDING | DOCUMENT_POSITION_IMPLEMENTATION_SPECIFIC; } return DOCUMENT_POSITION_FOLLOWING | DOCUMENT_POSITION_IMPLEMENTATION_SPECIFIC; } } thisNode = thisAncestor = thisOwnerDoc; break; } case Node.DOCUMENT_TYPE_NODE: { if (otherNode == thisOwnerDoc) { return DOCUMENT_POSITION_PRECEDING | DOCUMENT_POSITION_CONTAINS; } else if (thisOwnerDoc != null && thisOwnerDoc == otherOwnerDoc) { return DOCUMENT_POSITION_FOLLOWING; } break; } case Node.ATTRIBUTE_NODE: { thisNode = ((AttrImpl) thisAncestor).getOwnerElement(); if (otherAncestorType == Node.ATTRIBUTE_NODE) { otherNode = ((AttrImpl) otherAncestor).getOwnerElement(); if (otherNode == thisNode) { if (((NamedNodeMapImpl) thisNode.getAttributes()).precedes(other, this)) { return DOCUMENT_POSITION_PRECEDING | DOCUMENT_POSITION_IMPLEMENTATION_SPECIFIC; } return DOCUMENT_POSITION_FOLLOWING | DOCUMENT_POSITION_IMPLEMENTATION_SPECIFIC; } } // Now, find the ancestor of the element thisDepth = 0; for (node = thisNode; node != null; node = node.getParentNode()) { thisDepth += 1; if (node == otherNode) { // The other node is an ancestor of the owning element return DOCUMENT_POSITION_CONTAINS | DOCUMENT_POSITION_PRECEDING; } thisAncestor = node; } } } switch (otherAncestorType) { case Node.NOTATION_NODE: case Node.ENTITY_NODE: { final DocumentType container = thisOwnerDoc.getDoctype(); if (container == this) { return DOCUMENT_POSITION_IS_CONTAINED | DOCUMENT_POSITION_FOLLOWING; } otherNode = otherAncestor = thisOwnerDoc; break; } case Node.DOCUMENT_TYPE_NODE: { if (thisNode == otherOwnerDoc) { return DOCUMENT_POSITION_FOLLOWING | DOCUMENT_POSITION_IS_CONTAINED; } else if (otherOwnerDoc != null && thisOwnerDoc == otherOwnerDoc) { return DOCUMENT_POSITION_PRECEDING; } break; } case Node.ATTRIBUTE_NODE: { otherDepth = 0; otherNode = ((AttrImpl) otherAncestor).getOwnerElement(); for (node = otherNode; node != null; node = node.getParentNode()) { otherDepth += 1; if (node == thisNode) { // The other node is a descendent of the reference // node's element return DOCUMENT_POSITION_FOLLOWING | DOCUMENT_POSITION_IS_CONTAINED; } otherAncestor = node; } } } // thisAncestor and otherAncestor must be the same at this point, // otherwise, the original nodes are disconnected if (thisAncestor != otherAncestor) { final int thisAncestorNum; final int otherAncestorNum; thisAncestorNum = ((NodeImpl) thisAncestor).getNodeNumber(); otherAncestorNum = ((NodeImpl) otherAncestor).getNodeNumber(); if (thisAncestorNum > otherAncestorNum) { return DOCUMENT_POSITION_DISCONNECTED | DOCUMENT_POSITION_FOLLOWING | DOCUMENT_POSITION_IMPLEMENTATION_SPECIFIC; } return DOCUMENT_POSITION_DISCONNECTED | DOCUMENT_POSITION_PRECEDING | DOCUMENT_POSITION_IMPLEMENTATION_SPECIFIC; } // Go up the parent chain of the deeper node, until we find a node // with the same depth as the shallower node if (thisDepth > otherDepth) { for (int i = 0; i < thisDepth - otherDepth; i++) { thisNode = thisNode.getParentNode(); } // Check if the node we have reached is in fact "otherNode". This can // happen in the case of attributes. In this case, otherNode // "precedes" this. if (thisNode == otherNode) { return DOCUMENT_POSITION_PRECEDING; } } else { for (int i = 0; i < otherDepth - thisDepth; i++) { otherNode = otherNode.getParentNode(); } // Check if the node we have reached is in fact "thisNode". This can // happen in the case of attributes. In this case, otherNode // "follows" this. if (otherNode == thisNode) { return DOCUMENT_POSITION_FOLLOWING; } } // We now have nodes at the same depth in the tree. Find a common // ancestor. Node thisNodeP; Node otherNodeP; for (thisNodeP = thisNode.getParentNode(), otherNodeP = otherNode.getParentNode(); thisNodeP != otherNodeP;) { thisNode = thisNodeP; otherNode = otherNodeP; thisNodeP = thisNodeP.getParentNode(); otherNodeP = otherNodeP.getParentNode(); } // At this point, thisNode and otherNode are direct children of // the common ancestor. // See whether thisNode or otherNode is the leftmost for (Node current = thisNodeP.getFirstChild(); current != null; current = current.getNextSibling()) { if (current == otherNode) { return DOCUMENT_POSITION_PRECEDING; } else if (current == thisNode) { return DOCUMENT_POSITION_FOLLOWING; } } // REVISIT: shouldn't get here. Should probably throw an // exception return 0; } /** * This attribute returns the text content of this node and its descendants. * When it is defined to be null, setting it has no effect. When set, any * possible children this node may have are removed and replaced by a single * Text node containing the string this attribute is set to. On * getting, no serialization is performed, the returned string does not contain * any markup. No whitespace normalization is performed, the returned string * does not contain the element content whitespaces . Similarly, on setting, no * parsing is performed either, the input string is taken as pure textual * content.
* The string returned is made of the text content of this node depending on its * type, as defined below: *

* * * * * * * * * * * * * * * * *
Node typeContent
ELEMENT_NODE, ENTITY_NODE, * ENTITY_REFERENCE_NODE, DOCUMENT_FRAGMENT_NODEconcatenation of the * textContent attribute value of every child node, excluding * COMMENT_NODE and PROCESSING_INSTRUCTION_NODE nodes
ATTRIBUTE_NODE, TEXT_NODE, * CDATA_SECTION_NODE, COMMENT_NODE, PROCESSING_INSTRUCTION_NODEnodeValue
DOCUMENT_NODE, DOCUMENT_TYPE_NODE, * NOTATION_NODEnull
* * @exception DOMException DOMSTRING_SIZE_ERR: Raised when it would return more * characters than fit in a DOMString * variable on the implementation platform. */ @Override public String getTextContent() throws DOMException { return getNodeValue(); // overriden in some subclasses } // internal method taking a StringBuilder in parameter void getTextContent(final StringBuilder builder) throws DOMException { final String content = getNodeValue(); if (content != null) { builder.append(content); } } /** * This attribute returns the text content of this node and its descendants. * When it is defined to be null, setting it has no effect. When set, any * possible children this node may have are removed and replaced by a single * Text node containing the string this attribute is set to. On * getting, no serialization is performed, the returned string does not contain * any markup. No whitespace normalization is performed, the returned string * does not contain the element content whitespaces . Similarly, on setting, no * parsing is performed either, the input string is taken as pure textual * content.
* The string returned is made of the text content of this node depending on its * type, as defined below: * * * * * * * * * * * * * * * * * *
Node typeContent
ELEMENT_NODE, ENTITY_NODE, * ENTITY_REFERENCE_NODE, DOCUMENT_FRAGMENT_NODEconcatenation of the * textContent attribute value of every child node, excluding * COMMENT_NODE and PROCESSING_INSTRUCTION_NODE nodes
ATTRIBUTE_NODE, TEXT_NODE, * CDATA_SECTION_NODE, COMMENT_NODE, PROCESSING_INSTRUCTION_NODEnodeValue
DOCUMENT_NODE, DOCUMENT_TYPE_NODE, * NOTATION_NODEnull
* * @exception DOMException DOMSTRING_SIZE_ERR: Raised when it would return more * characters than fit in a DOMString * variable on the implementation platform. */ @Override public void setTextContent(final String textContent) throws DOMException { setNodeValue(textContent); } /** * Returns whether this node is the same node as the given one.
* This method provides a way to determine whether two Node * references returned by the implementation reference the same object. When two * Node references are references to the same object, even if * through a proxy, the references may be used completely interchangably, such * that all attributes have the same values and calling the same DOM method on * either reference always has exactly the same effect. * * @param other The node to test against. * @return Returns true if the nodes are the same, * false otherwise. */ @Override public boolean isSameNode(final Node other) { // we do not use any wrapper so the answer is obvious return this == other; } /** * DOM Level 3: Experimental This method checks if the specified * namespaceURI is the default namespace or not. * * @param namespaceURI The namespace URI to look for. * @return true if the specified namespaceURI is the * default namespace, false otherwise. */ @Override public boolean isDefaultNamespace(final String namespaceURI) { // REVISIT: remove casts when DOM L3 becomes REC. final short type = this.getNodeType(); switch (type) { case Node.ELEMENT_NODE: { final String namespace = this.getNamespaceURI(); final String prefix = this.getPrefix(); // REVISIT: is it possible that prefix is empty string? if (prefix == null || prefix.length() == 0) { if (namespaceURI == null) { return namespace == namespaceURI; } return namespaceURI.equals(namespace); } if (this.hasAttributes()) { final ElementImpl elem = (ElementImpl) this; final NodeImpl attr = (NodeImpl) elem.getAttributeNodeNS("http://www.w3.org/2000/xmlns/", "xmlns"); if (attr != null) { final String value = attr.getNodeValue(); if (namespaceURI == null) { return namespace == value; } return namespaceURI.equals(value); } } final NodeImpl ancestor = (NodeImpl) getElementAncestor(this); if (ancestor != null) { return ancestor.isDefaultNamespace(namespaceURI); } return false; } case Node.DOCUMENT_NODE: { final Element docElement = ((Document) this).getDocumentElement(); if (docElement != null) { return docElement.isDefaultNamespace(namespaceURI); } return false; } case Node.ENTITY_NODE: case Node.NOTATION_NODE: case Node.DOCUMENT_FRAGMENT_NODE: case Node.DOCUMENT_TYPE_NODE: // type is unknown return false; case Node.ATTRIBUTE_NODE: { if (this.ownerNode.getNodeType() == Node.ELEMENT_NODE) { return ownerNode.isDefaultNamespace(namespaceURI); } return false; } default: { final NodeImpl ancestor = (NodeImpl) getElementAncestor(this); if (ancestor != null) { return ancestor.isDefaultNamespace(namespaceURI); } return false; } } } /** * {@inheritDoc} * * DOM Level 3 - Experimental: Look up the prefix associated to the given * namespace URI, starting from this node. * * @param namespaceURI the namespace uri * @return the prefix for the namespace */ @Override public String lookupPrefix(final String namespaceURI) { // REVISIT: When Namespaces 1.1 comes out this may not be true // Prefix can't be bound to null namespace if (namespaceURI == null) { return null; } final short type = this.getNodeType(); switch (type) { case Node.ELEMENT_NODE: { this.getNamespaceURI(); // to flip out children return lookupNamespacePrefix(namespaceURI, (ElementImpl) this); } case Node.DOCUMENT_NODE: { final Element docElement = ((Document) this).getDocumentElement(); if (docElement != null) { return docElement.lookupPrefix(namespaceURI); } return null; } case Node.ENTITY_NODE: case Node.NOTATION_NODE: case Node.DOCUMENT_FRAGMENT_NODE: case Node.DOCUMENT_TYPE_NODE: // type is unknown return null; case Node.ATTRIBUTE_NODE: { if (this.ownerNode.getNodeType() == Node.ELEMENT_NODE) { return ownerNode.lookupPrefix(namespaceURI); } return null; } default: { final NodeImpl ancestor = (NodeImpl) getElementAncestor(this); if (ancestor != null) { return ancestor.lookupPrefix(namespaceURI); } return null; } } } /** * {@inheritDoc} * * DOM Level 3 - Experimental: Look up the namespace URI associated to the given * prefix, starting from this node. Use lookupNamespaceURI(null) to lookup the * default namespace */ @Override public String lookupNamespaceURI(final String specifiedPrefix) { final short type = this.getNodeType(); switch (type) { case Node.ELEMENT_NODE: { String namespace = this.getNamespaceURI(); final String prefix = this.getPrefix(); if (namespace != null) { // REVISIT: is it possible that prefix is empty string? if (specifiedPrefix == null && prefix == specifiedPrefix) { // looking for default namespace return namespace; } else if (prefix != null && prefix.equals(specifiedPrefix)) { // non default namespace return namespace; } } if (this.hasAttributes()) { final NamedNodeMap map = this.getAttributes(); final int length = map.getLength(); for (int i = 0; i < length; i++) { final Node attr = map.item(i); namespace = attr.getNamespaceURI(); if (namespace != null && "http://www.w3.org/2000/xmlns/".equals(namespace)) { final String attrPrefix = attr.getPrefix(); final String value = attr.getNodeValue(); // at this point we are dealing with DOM Level 2 nodes only if (specifiedPrefix == null && "xmlns".equals(attr.getNodeName())) { // default namespace return value.length() > 0 ? value : null; } else if (attrPrefix != null && "xmlns".equals(attrPrefix) && attr.getLocalName().equals(specifiedPrefix)) { // non default namespace return value.length() > 0 ? value : null; } } } } final NodeImpl ancestor = (NodeImpl) getElementAncestor(this); if (ancestor != null) { return ancestor.lookupNamespaceURI(specifiedPrefix); } return null; } case Node.DOCUMENT_NODE: { final Element docElement = ((Document) this).getDocumentElement(); if (docElement != null) { return docElement.lookupNamespaceURI(specifiedPrefix); } return null; } case Node.ENTITY_NODE: case Node.NOTATION_NODE: case Node.DOCUMENT_FRAGMENT_NODE: case Node.DOCUMENT_TYPE_NODE: // type is unknown return null; case Node.ATTRIBUTE_NODE: { if (this.ownerNode.getNodeType() == Node.ELEMENT_NODE) { return ownerNode.lookupNamespaceURI(specifiedPrefix); } return null; } default: { final NodeImpl ancestor = (NodeImpl) getElementAncestor(this); if (ancestor != null) { return ancestor.lookupNamespaceURI(specifiedPrefix); } return null; } } } Node getElementAncestor(final Node currentNode) { Node parent = currentNode.getParentNode(); while (parent != null) { final short type = parent.getNodeType(); if (type == Node.ELEMENT_NODE) { return parent; } parent = parent.getParentNode(); } return null; } String lookupNamespacePrefix(final String namespaceURI, final ElementImpl el) { String namespace = this.getNamespaceURI(); // REVISIT: if no prefix is available is it null or empty string, or // could be both? final String prefix = this.getPrefix(); if (namespace != null && namespace.equals(namespaceURI)) { if (prefix != null) { final String foundNamespace = el.lookupNamespaceURI(prefix); if (foundNamespace != null && foundNamespace.equals(namespaceURI)) { return prefix; } } } if (this.hasAttributes()) { final NamedNodeMap map = this.getAttributes(); final int length = map.getLength(); for (int i = 0; i < length; i++) { final Node attr = map.item(i); namespace = attr.getNamespaceURI(); if (namespace != null && "http://www.w3.org/2000/xmlns/".equals(namespace)) { final String attrPrefix = attr.getPrefix(); final String value = attr.getNodeValue(); // DOM Level 2 nodes if (("xmlns".equals(attr.getNodeName())) || (attrPrefix != null && "xmlns".equals(attrPrefix)) && value.equals(namespaceURI)) { final String localname = attr.getLocalName(); final String foundNamespace = el.lookupNamespaceURI(localname); if (foundNamespace != null && foundNamespace.equals(namespaceURI)) { return localname; } } } } } final NodeImpl ancestor = (NodeImpl) getElementAncestor(this); if (ancestor != null) { return ancestor.lookupNamespacePrefix(namespaceURI, el); } return null; } /** * Tests whether two nodes are equal.
* This method tests for equality of nodes, not sameness (i.e., whether the two * nodes are references to the same object) which can be tested with * Node.isSameNode. All nodes that are the same will also be equal, * though the reverse may not be true.
* Two nodes are equal if and only if the following conditions are satisfied: * The two nodes are of the same type.The following string attributes are equal: * nodeName, localName, namespaceURI, * prefix, nodeValue , baseURI. This is: * they are both null, or they have the same length and are * character for character identical. The attributes * NamedNodeMaps are equal. This is: they are both * null, or they have the same length and for each node that exists * in one map there is a node that exists in the other map and is equal, * although not necessarily at the same index.The childNodes * NodeLists are equal. This is: they are both null, * or they have the same length and contain equal nodes at the same index. This * is true for Attr nodes as for any other type of node. Note that * normalization can affect equality; to avoid this, nodes should be normalized * before being compared.
* For two DocumentType nodes to be equal, the following conditions * must also be satisfied: The following string attributes are equal: * publicId, systemId, internalSubset.The * entities NamedNodeMaps are equal.The * notations NamedNodeMaps are equal.
* On the other hand, the following do not affect equality: the * ownerDocument attribute, the specified attribute * for Attr nodes, the isWhitespaceInElementContent * attribute for Text nodes, as well as any user data or event * listeners registered on the nodes. * * @param arg The node to compare equality with. * @return If the nodes, and possibly subtrees are equal, true * otherwise false. */ @Override public boolean isEqualNode(final Node arg) { if (arg == this) { return true; } if (arg.getNodeType() != getNodeType()) { return false; } // in theory nodeName can't be null but better be careful // who knows what other implementations may be doing?... if (getNodeName() == null) { if (arg.getNodeName() != null) { return false; } } else if (!getNodeName().equals(arg.getNodeName())) { return false; } if (getLocalName() == null) { if (arg.getLocalName() != null) { return false; } } else if (!getLocalName().equals(arg.getLocalName())) { return false; } if (getNamespaceURI() == null) { if (arg.getNamespaceURI() != null) { return false; } } else if (!getNamespaceURI().equals(arg.getNamespaceURI())) { return false; } if (getPrefix() == null) { if (arg.getPrefix() != null) { return false; } } else if (!getPrefix().equals(arg.getPrefix())) { return false; } if (getNodeValue() == null) { if (arg.getNodeValue() != null) { return false; } } else if (!getNodeValue().equals(arg.getNodeValue())) { return false; } return true; } @Override public Object getFeature(final String feature, final String version) { // we don't have any alternate node, either this node does the job // or we don't have anything that does return isSupported(feature, version) ? this : null; } /** * Associate an object to a key on this node. The object can later be retrieved * from this node by calling getUserData with the same key. * * @param key The key to associate the object to. * @param data The object to associate to the given key, or null * to remove any existing association to that key. * @param handler The handler to associate to that key, or null. * @return Returns the DOMObject previously associated to the given * key on this node, or null if there was none. */ @Override public Object setUserData(final String key, final Object data, final UserDataHandler handler) { return null; } /** * Retrieves the object associated to a key on a this node. The object must * first have been set to this node by calling setUserData with the * same key. * * @param key The key the object is associated to. * @return Returns the DOMObject associated to the given key on * this node, or null if there was none. */ @Override public Object getUserData(final String key) { return null; } // // Protected methods // /** * Denotes that this node has changed. */ protected void changed() { // we do not actually store this information on every node, we only // have a global indicator on the Document. Doing otherwise cost us too // much for little gain. ownerDocument().changed(); } /** * @return the number of changes to this node. */ protected int changes() { // we do not actually store this information on every node, we only // have a global indicator on the Document. Doing otherwise cost us too // much for little gain. return ownerDocument().changes(); } /** * Override this method in subclass to hook in efficient internal data * structure. */ protected void synchronizeData() { // By default just change the flag to avoid calling this method again needsSyncData(false); } final boolean needsSyncData() { return (flags & SYNCDATA) != 0; } final void needsSyncData(final boolean value) { flags = (short) (value ? flags | SYNCDATA : flags & ~SYNCDATA); } final boolean needsSyncChildren() { return (flags & SYNCCHILDREN) != 0; } public final void needsSyncChildren(final boolean value) { flags = (short) (value ? flags | SYNCCHILDREN : flags & ~SYNCCHILDREN); } final boolean isOwned() { return (flags & OWNED) != 0; } final void isOwned(final boolean value) { flags = (short) (value ? flags | OWNED : flags & ~OWNED); } final boolean isFirstChild() { return (flags & FIRSTCHILD) != 0; } final void isFirstChild(final boolean value) { flags = (short) (value ? flags | FIRSTCHILD : flags & ~FIRSTCHILD); } final boolean isSpecified() { return (flags & SPECIFIED) != 0; } final void isSpecified(final boolean value) { flags = (short) (value ? flags | SPECIFIED : flags & ~SPECIFIED); } // inconsistent name to avoid clash with public method on TextImpl final boolean internalIsIgnorableWhitespace() { return (flags & IGNORABLEWS) != 0; } final void isIgnorableWhitespace(final boolean value) { flags = (short) (value ? flags | IGNORABLEWS : flags & ~IGNORABLEWS); } final boolean hasStringValue() { return (flags & HASSTRING) != 0; } final void hasStringValue(final boolean value) { flags = (short) (value ? flags | HASSTRING : flags & ~HASSTRING); } final boolean isNormalized() { return (flags & NORMALIZED) != 0; } final void isNormalized(final boolean value) { // See if flag should propagate to parent. if (!value && isNormalized() && ownerNode != null) { ownerNode.isNormalized(false); } flags = (short) (value ? flags | NORMALIZED : flags & ~NORMALIZED); } final boolean isIdAttribute() { return (flags & ID) != 0; } final void isIdAttribute(final boolean value) { flags = (short) (value ? flags | ID : flags & ~ID); } // NON-DOM method for debugging convenience. @Override public String toString() { return "[" + getNodeName() + ": " + getNodeValue() + "]"; } }




© 2015 - 2024 Weber Informatics LLC | Privacy Policy