<?xml version="1.0" standalone="no"?>
<!DOCTYPE s1 SYSTEM "../../style/dtd/document.dtd">
<!--
 * Copyright 2001-2004 The Apache Software Foundation.
 *
 * 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
 *
 *     http://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.
-->
<!-- $Id: xsltc_iterators.xml 337884 2004-02-17 19:29:35Z minchau $ -->

<s1 title="XSLTC node iterators">

  <s2 title="Contents">

  <p>This document describes the function of XSLTC's node iterators. It also
  describes the <code>NodeIterator</code> interface and some implementations of
  this interface are described in detail:</p>

  <ul>
    <li><link anchor="purpose">Node iterator function</link></li>
    <li><link anchor="interface">NodeIterator interface</link></li>
    <li><link anchor="baseclass">Node iterator base class</link></li>    
    <li><link anchor="details">Implementation details</link></li>    
  </ul>

  </s2>

  <!--=================== OVERVIEW SECTION ===========================--> 

  <anchor name="purpose"/>
  <s2 title="Node Iterator Function">

    <p>Node iterators have several functions in XSLTC. The most obvious is
    acting as a placeholder for node-sets. Node iterators also act as a link
    between the translet and the DOM(s), they can act as filters (implementing
    predicates), they contain the functionality necessary to cover all XPath
    axes and they even serve as a front-end to XSLTC's node-indexing mechanism
    (for the <code>id()</code> and <code>key()</code> functions).</p>
  </s2>

  <!--=================== INTERFACE SECTION ==========================--> 

  <anchor name="interface"/>
  <s2 title="Node Iterator Interface">

    <p>The node iterator interface is defined in
    <code>org.apache.xalan.xsltc.NodeIterator</code>.</p>

    <p>The most basic operations in the <code>NodeIterator</code> interface are
    for setting the iterators start-node. The &quot;start-node&quot; is
    an index into the DOM. This index, and the axis of the iterator, determine
    the node-set that the iterator contains. The axis is programmed into the
    various node iterator implementations, while the start-node can be set by
    calling:</p><source>
    public NodeIterator setStartNode(int node);</source>

    <p>Once the start node is set the node-set can be traversed by a sequence of
    calls to:</p><source>
    public int next();</source>

    <p>This method will return the constant <code>NodeIterator.END</code> when
    the whole node-set has been returned. The iterator can be reset to the start
    of the node-set by calling:</p><source>
    public NodeIterator reset();</source>

    <p>Two additional methods are provided to set the position within the
    node-set. The first method below will  mark  the current node in the
    node-set, while the second will (at any point) set the iterators position
    back to that node.</p><source>
    public void setMark();
    public void gotoMark();</source>

    <p>Every node iterator implements two functions that make up the
    functionality behind XPath's <code>getPosition()</code> and
    <code>getLast()</code> functions.</p><source>
    public int getPosition();
    public int getLast();</source>

    <p>The <code>getLast()</code> function returns the number of nodes in the
    set, while the <code>getPosition()</code> returns the current position
    within the node-set. The value returned by <code>getPosition()</code> for
    the first node in the set is always 1 (one), and the value returned for the
    last node in the set is always the same value as is returned by
    <code>getLast()</code>.</p>

    <p>All node iterators that implement an XPath axis will return the node-set
    in the natural order of the axis. For example, the iterator implementing the
     ancestor  axis will return nodes in reverse document order (bottom to
    top), while the iterator implementing the  descendant  will return
    nodes in document order. The node iterator interface has a method that can
    be used to determine if an iterator returns nodes in reverse document order:
    </p><source>
    public boolean isReverse();</source>

    <p>Two methods are provided for when node iterators are encapsulated inside
    a variable or parameter. To understand the purpose behind these two methods
    we should have a look at a sample XML document and stylesheet first:</p>
    <source>
    &lt;?xml version="1.0"?>
    &lt;foo>
        &lt;bar>
            &lt;baz>A&lt;/baz>
            &lt;baz>B&lt;/baz>
        &lt;/bar>
        &lt;bar>
            &lt;baz>C&lt;/baz>
            &lt;baz>D&lt;/baz>
        &lt;/bar>
    &lt;/foo>

    &lt;xsl:stylesheet version="1.0" xmlns:xsl="http://www.w3.org/1999/XSL/Transform">

        &lt;xsl:template match="foo">
            &lt;xsl:variable name="my-nodes" select="//foo/bar/baz"/>
            &lt;xsl:for-each select="bar">
                &lt;xsl:for-each select="baz">
                    &lt;xsl:value-of select="."/>
                &lt;/xsl:for-each>
                &lt;xsl:for-each select="$my-nodes">
                    &lt;xsl:value-of select="."/>
                &lt;/xsl:for-each>
            &lt;/xsl:for-each>
        &lt;/xsl:template>

    &lt;/xsl:stylesheet></source>

    <p>Now, there are three iterators at work here. The first iterator is the
    one that is wrapped inside the variable <code>my-nodes</code> - this
    iterator contains  all  <code>&lt;baz/&gt;</code> elements in the
    document. The second iterator contains all <code>&lt;bar&gt;</code>
    elements under the current element (this is the iterator used by the
    outer <code>for-each</code> loop). The third and last iterator is the one
    used by the first of the inner <code>for-each</code> loops. When the outer
    loop is run the first time, this third iterator will be initialized to
    contain the first two <code>&lt;baz&gt;</code> elements under the context
    node (the first <code>&lt;bar&gt;</code> element). Iterators are by default
    restarted from the current node when used inside a <code>for-each</code>
    loop like this. But what about the iterator inside the variable
    <code>my-nodes</code>? The variable should keep its assigned value, no
    matter what the context node is. In able to prevent the iterator from being
    reset, we must use a mechanism to block calls to the
    <code>setStartNode()</code> method. This is done in three steps:</p>

    <ul>
      <li>The iterator is created and initialized when the variable gets
      assigned its value (node-set).</li>
      <li>When the variable is read, the iterator is copied (cloned). The
      original iterator inside the variable is never used directly. This is
      to make sure that the iterator inside the variable is always in its
      original state when read.</li>
      <li>The iterator clone is marked as not restartable to prevent it from
      being restarted when used to iterate the <code>&lt;xsl:for-each&gt;</code>
      element loop.</li>
    </ul>

    <p>These are the two methods used for the three steps above:</p><source>
    public NodeIterator cloneIterator();
    public void setRestartable(boolean isRestartable);</source>

    <p>Special care must be taken when implementing these methods in some
    iterators. The <code>StepIterator</code> class is the best example of this.
    This iterator wraps two other iterators; one of which is used to generate
    start-nodes for the other - so one of the encapsulated node iterators must
    always remain restartable - even when used inside variables. The
    <code>StepIterator</code> class is described in detail later in this
    document.</p>

  </s2>


  <!--================= BASE CLASS SECTION ===========================--> 

  <anchor name="baseclass"/>
  <s2 title="Node Iterator Base Class">

    <p>A node iterator base class is provided to contain some common
    functionality. The base class implements the node iterator interface, and
    has a few additional methods:</p><source>
    public NodeIterator includeSelf();
    protected final int returnNode(final int node);
    protected final NodeIterator resetPosition();</source>

    <p>The <code>includeSelf()</code> is used with certain axis iterators that
    implement both the <code>ancestor</code> and <code>ancestor-or-self</code>
    axis and similar. One common implementation is used for these axes and
    this method is used to signal that the  &quot;self&quot;  node should
    also be included in the node-set.</p>

    <p>The <code>returnNode()</code> method is called by the implementation of
    the <code>next()</code> method. <code>returnNode()</code> increments an
    internal node counter/cursor that keeps track of the current position within
    the node set. This counter/cursor is then used by the 
    <code>getPosition()</code> implementation to return the current position.
    The node cursor can be reset by calling <code>resetPosition()</code>. This
    method is normally called by an iterator's <code>reset()</code> method.</p>

  </s2>

  <!--==================== DETAILS SECTION ===========================--> 

  <anchor name="details"/>
  <s2 title="Node Iterator Implementation Details">

    <s3 title="Axis iterators">

    <p>All axis iterators are implemented as inner classes of the internal
    DOM implementation <code>org.apache.xalan.xsltc.dom.DOMImpl</code>. In this
    way all axis iterator classes have direct access to the internal node
    type- and navigation arrays of the DOM:</p><source>
    private short[]   _type;          // Node types
    private short[]   _namespace;     // Namespace URI types
    private short[]   _prefix;        // Namespace prefix types

    private int[]     _parent;        // Index of a node's parent
    private int[]     _nextSibling;   // Index of a node's next sibling node
    private int[]     _offsetOrChild; // Index of an elements first child node
    private int[]     _lengthOrAttr;  // Index of an elements first attribute node</source>

    <p>The axis iterators can be instanciated by calling either of these two
    methods of the DOM:</p><source>
    public NodeIterator getAxisIterator(final int axis);
    public NodeIterator getTypedAxisIterator(final int axis, final int type);</source>

    </s3>

    <s3 title="StepIterator">
    
    <p>The <code>StepIterator</code> is used to  chain  other iterators. A
    very basic example is this XPath expression:</p><source>
    &lt;xsl:for-each select="foo/bar"></source>

    <p>To generate the appropriate node-set for this loop we need three
    iterators. The compiler will generate code that first creates a typed axis
    iterator; the axis will be  child  and the type will be that assigned
    to <code>&lt;foo&gt;</code> elements. Then a second typed axis iterator will
    be created; this also a  child -iterator, but this one with the type
    assigned to <code>&lt;bar&gt;</code> elements. The third iterator is a
    step iterator that encapsulates the two axis iterators. The step iterator is
    the initialized with the context node.</p>

    <p>The step iterator will use the first axis iterator to generate
    start-nodes for the second axis iterator. In plain english this means that
    the step iterator will scan all <code>foo</code> elements for any
    <code>bar</code> child elements. When a <code>StepIterator</code> is
    initialized with a start-node it passes the start node to the
    <code>setStartNode()</code> method of its  source -iterator (left).
    It then calls <code>next()</code> on that iterator to get the start-node
    for the  iterator  iterator (right):</p><source>
    // Set start node for left-hand iterator...
    _source.setStartNode(_startNode);
    // ... and get start node for right-hand iterator from left-hand,
    _iterator.setStartNode(_source.next());</source>

    <p>The step iterator will keep returning nodes from its right iterator until
    it runs out of nodes. Then a new start-node is retrieved by again calling
    <code>next()</code> on the  source -iterator. This is why the
    right-hand iterator always has to be restartable - even if the step iterator
    is placed inside a variable or parameter. This becomes even more complicated
    for step iterators that encapsulate other step iterators. We'll make our
    previous example a bit more interesting:</p><source>
    &lt;xsl:for-each select="foo/bar[@name='cat and cage']/baz"></source>

    <p>This will result in an iterator-tree similar to this:</p>

    <p><img src="iterator_stack.gif" alt="iterator_stack.gif"/></p>
    <p><ref>Figure 1: Stacked step iterators</ref></p>

    <p>The  &quot;foo&quot;  iterator is used to supply the second step
    iterator with start nodes. The second step iterator will pass these start
    nodes to the  &quot;bar&quot;  iterator, which will be used to get the
    start nodes for the third step iterator, and so on....</p>

    </s3>

    <s3 title="Iterators for Filtering/Predicates">

    <p>The <code>org.apache.xalan.xsltc.dom</code> package contains a few
    iterators that are used to implement predicates and filters. Such iterators
    are normally placed on top of another iterator, and return only those nodes
    that match a specific node value, position, etc.
    These iterators include:</p>

    <ul>
      <li>NthIterator</li>
      <li>NodeValueIterator</li>
      <li>FilteredStepIterator</li>
      <li>CurrentNodeListIterator</li>
    </ul>

    <p>The last one is the most interesting. This iterator is used to implement
    chained predicates, such as:</p><source>
    &lt;xsl:value-of select="foo[@blob='boo'][2]"></source>

    <p>The first predicate reduces the node set from containing all
    <code>&lt;foo&gt;</code> elements, to containing only those elements that
    have a  &quot;blob&quot;  attribute with the value 'boo'. The
    <code>CurrentNodeListIterator</code> is used to contain this reduced
    node-set. The iterator is constructed by passing it a source iterator (in
    this case an iterator that contains all <code>&lt;foo&gt;</code> elements)
    and a filter that implements the predicate (<code>@blob = 'boo'</code>).</p>

    </s3>

    <s3 title="SortingIterator">

    <p>The sorting iterator is one of the main functional components behind the
    implementation of the <code>&lt;xsl:sort&gt;</code> element. This element,
    including the sorting iterator, is described in detail in the
    <code>&lt;xsl:sort&gt;</code>
    <link idref="xsl_sort_design">design document</link>.</p>

    </s3>

    <s3 title="SingletonIterator"></s3>

    <p>The singleton iterator is a wrapper for a single node. The node passed
    in to the <code>setStartNode()</code> method is the only node that will be
    returned by the <code>next()</code> method. The singleton iterator is used
    mainly for node to node-set type conversions.</p>

    <s3 title="UnionIterator">

    <p>The union iterator is used to contain unions of node-sets contained in
    other iterators. Some of the methods in this iterator are unnecessary
    comlicated. The <code>next()</code> method contains an algorithm for
    ensuring that the union node-set is returned in document order. We might be
    better off by simply wrapping the union iterator inside a duplicate filter
    iterator, but there could be some performance implications. Worth checking.
    </p>

    </s3>

    <s3 title="KeyIndex">

    <p>This is not just an node iterator. An index used for keys and ids will
    return a set of nodes that are contained within the named index and that
    share a certain property. The <code>KeyIndex</code> implements the node
    iterator interface, so that these nodes can be returned and handled just
    like any other node set. See the
    <link idref="xsl_key_design">design document</link> for 
    <code>&lt;xsl:key&gt;</code>, <code>key()</code> and <code>id()</code>
    for further details.</p>

    </s3>

  </s2>

</s1>