validationconsumer.java

gcc的组建

	} else {
	    Recognizer state = (Recognizer) contentStack.peek ();

	    if (state != null) {
		Recognizer newstate = state.acceptElement (qName);

		if (newstate == null)
		    error ("Element type '" + qName
			+ "' in element '" + state.type.name
			+ "' violates content model " + state.type.model
			);
		if (newstate != state) {
		    contentStack.pop ();
		    contentStack.push (newstate);
		}
	    }
	}

	//
	// Then check that this element was declared, and push the
	// object used to validate its content model onto our stack.
	//
	// This is where the recognizer gets created, if needed; if
	// it's a "children" (elements) content model, an NDFA is
	// created.  (One recognizer is used per content type, no
	// matter how complex that recognizer is.)
	//
	ElementInfo		info;

	info = (ElementInfo) elements.get (qName);
	if (info == null || info.model == null) {
	    // VC: Element Valid (base clause)
	    error ("Element type '" + qName + "' was not declared");
	    contentStack.push (null);

	    // for less diagnostic noise, fake a declaration.
	    elementDecl (qName, "ANY");
	} else
	    contentStack.push (info.getRecognizer (this));

	//
	// Then check each attribute present
	//
	int			len;
	String			aname;
	AttributeInfo		ainfo;

	if (atts != null)
	    len = atts.getLength ();
	else
	    len = 0;
	
	for (int i = 0; i < len; i++) {
	    aname = atts.getQName (i);

	    if (info == null
		    || (ainfo = (AttributeInfo) info.attributes.get (aname))
			    == null) {
		// VC: Attribute Value Type
		error ("Attribute '" + aname
		    + "' was not declared for element type " + qName);
		continue;
	    }

	    String value = atts.getValue (i);

	    // note that "==" for type names and "#FIXED" is correct
	    // (and fast) since we've interned those literals.

	    if ("#FIXED" == ainfo.mode) {
		String expanded = expandDefaultRefs (ainfo.value);

		// VC: Fixed Attribute Default
		if (!value.equals (expanded)) {
		    error ("Attribute '" + aname
			+ "' must match " + expanded
			);
		    continue;
		}
	    }

	    if ("CDATA" == ainfo.type)
		continue;
	    
	    //
	    // For all other attribute types, there are various
	    // rules to follow.
	    //
	    
	    if ("ID" == ainfo.type) {
		// VC: ID (must be a name)
		if (isName (value, "ID attribute", aname)) {
		    if (Boolean.TRUE == ids.get (value))
			// VC: ID (appears once)
			error ("ID attribute " + aname
			    + " uses an ID value '" + value
			    + "' which was already declared.");
		    else
			// any forward refs are no longer problems
			ids.put (value, Boolean.TRUE);
		}
		continue;
	    } 

	    if ("IDREF" == ainfo.type) {
		// VC: IDREF (value must be a name)
		if (isName (value, "IDREF attribute", aname)) {
		    // VC: IDREF (must match some ID attribute)
		    if (ids.get (value) == null)
			// new -- assume it's a forward ref
			ids.put (value, Boolean.FALSE);
		}
		continue;
	    } 

	    if ("IDREFS" == ainfo.type) {
		StringTokenizer	tokens = new StringTokenizer (value, " ");

		if (!tokens.hasMoreTokens ()) {
		    // VC: IDREF (one or more values)
		    error ("IDREFS attribute " + aname
			+ " must have at least one ID ref");
		} else do {
		    String id = tokens.nextToken ();

		    // VC: IDREF (value must be a name)
		    if (isName (id, "IDREFS attribute", aname)) {
			// VC: IDREF (must match some ID attribute)
			if (ids.get (id) == null)
			    // new -- assume it's a forward ref
			    ids.put (id, Boolean.FALSE);
		    }
		} while (tokens.hasMoreTokens ());
		continue;
	    }

	    if ("NMTOKEN" == ainfo.type) {
		// VC: Name Token (is a name token)
		isNmtoken (value, "NMTOKEN attribute", aname);
		continue;
	    }

	    if ("NMTOKENS" == ainfo.type) {
		StringTokenizer	tokens = new StringTokenizer (value, " ");

		if (!tokens.hasMoreTokens ()) {
		    // VC: Name Token (one or more values)
		    error ("NMTOKENS attribute " + aname
			+ " must have at least one name token");
		} else do {
		    String token = tokens.nextToken ();

		    // VC: Name Token (is a name token)
		    isNmtoken (token, "NMTOKENS attribute", aname);
		} while (tokens.hasMoreTokens ());
		continue;
	    }

	    if ("ENTITY" == ainfo.type) {
		if (!unparsed.contains (value))
		    // VC: Entity Name
		    error ("Value of attribute '" + aname
			+ "' refers to unparsed entity '" + value
			+ "' which was not declared.");
		continue;
	    }

	    if ("ENTITIES" == ainfo.type) {
		StringTokenizer	tokens = new StringTokenizer (value, " ");

		if (!tokens.hasMoreTokens ()) {
		    // VC: Entity Name (one or more values)
		    error ("ENTITIES attribute " + aname
			+ " must have at least one name token");
		} else do {
		    String entity = tokens.nextToken ();

		    if (!unparsed.contains (entity))
			// VC: Entity Name
			error ("Value of attribute '" + aname
			    + "' refers to unparsed entity '" + entity
			    + "' which was not declared.");
		} while (tokens.hasMoreTokens ());
		continue;
	    }

	    //
	    // check for enumerations last; more expensive
	    //
	    if (ainfo.type.charAt (0) == '(' /*)*/	
		    || ainfo.type.startsWith ("NOTATION ")
		    ) {
		// VC: Enumeration (value must be defined)
		checkEnumeration (value, ainfo.type, aname);
		continue;
	    }
	}

	//
	// Last, check that all #REQUIRED attributes were provided
	//
	if (info != null) {
	    Hashtable	table = info.attributes;

	    if (table.size () != 0) {
		Enumeration	e = table.keys ();

		// XXX table.keys uses the heap, bleech -- slows things

		while (e.hasMoreElements ()) {
		    aname = (String) e.nextElement ();
		    ainfo = (AttributeInfo) table.get (aname);

		    // "#REQUIRED" mode was interned in attributeDecl
		    if ("#REQUIRED" == ainfo.mode
			    && atts.getValue (aname) == null) {
			// VC: Required Attribute
			error ("Attribute '" + aname + "' must be specified "
			    + "for element type " + qName);
		    }
		}
	    }
	}
	super.startElement (uri, localName, qName, atts);
    }

    /**
     * <b>ContentHandler</b> Reports a validity error if the element's content
     * model does not permit character data.
     * Passed to the next consumer.
     */
    public void characters (char ch [], int start, int length)
    throws SAXException
    {
	Recognizer state;

	if (contentStack.empty ())
	    state = null;
	else
	    state = (Recognizer) contentStack.peek ();

	// NOTE:  if this ever supports with SAX parsers that don't
	// report ignorable whitespace as such (only XP?), this class
	// needs to morph it into ignorableWhitespace() as needed ...

	if (state != null && !state.acceptCharacters ())
	    // VC: Element Valid (clauses three, four -- see recognizer)
	    error ("Character content not allowed in element "
		+ state.type.name);
	
	super.characters (ch, start, length);
    }
	

    /**
     * <b>ContentHandler</b> Reports a validity error if the element's content
     * model does not permit end-of-element yet, or a well formedness error
     * if there was no matching startElement call.
     * Passed to the next consumer.
     */
    public void endElement (String uri, String localName, String qName)
    throws SAXException
    {
	try {
	    Recognizer state = (Recognizer) contentStack.pop ();

	    if (state != null && !state.completed ())
		// VC: Element valid (clauses two, three, four; see Recognizer)
		error ("Premature end for element '"
		    + state.type.name
		    + "', content model "
		    + state.type.model);
	    
	    // could insist on match of start element, but that's
	    // something the input stream must to guarantee.

	} catch (EmptyStackException e) {
	    fatalError ("endElement without startElement: " + qName
		+ ((uri == null)
		    ? ""
		    : ( " { '" + uri + "', " + localName + " }")));
	}
	super.endElement (uri, localName, qName);
    }

    /**
     * <b>ContentHandler</b> Checks whether all ID values that were
     * referenced have been declared, and releases all resources. 
     * Passed to the next consumer.
     * 
     * @see #setDocumentLocator
     */
    public void endDocument ()
    throws SAXException
    {
	for (Enumeration idNames = ids.keys ();
		idNames.hasMoreElements ();
		/* NOP */) {
	    String id = (String) idNames.nextElement ();
	    
	    if (Boolean.FALSE == ids.get (id)) {
		// VC: IDREF (must match ID)
		error ("Undeclared ID value '" + id
		    + "' was referred to by an IDREF/IDREFS attribute");
	    }
	}

	resetState ();
	super.endDocument ();
    }


    /** Holds per-element declarations */
    static private final class ElementInfo
    {
	String			name;
	String			model;

	// key = attribute name; value = AttributeInfo
	Hashtable		attributes = new Hashtable (11);

	ElementInfo (String n) { name = n; }

	private Recognizer	recognizer;

	// for validating content models:  one per type, shared,
	// and constructed only on demand ... so unused elements do
	// not need to consume resources.
	Recognizer	getRecognizer (ValidationConsumer consumer)
	throws SAXException
	{
	    if (recognizer == null) {
		if ("ANY".equals (model))
		    recognizer = ANY;
		else if ("EMPTY".equals (model))
		    recognizer = new EmptyRecognizer (this);
		else if ('#' == model.charAt (1))
		    // n.b. this constructor does a validity check
		    recognizer = new MixedRecognizer (this, consumer);
		else
		    recognizer = new ChildrenRecognizer (this, consumer);
	    }
	    return recognizer;
	}
    }

    /** Holds per-attribute declarations */
    static private final class AttributeInfo
    {
	String	type;
	String	mode;		// #REQUIRED, etc (or null)
	String	value;		// or null
    }


    //
    // Content model validation
    //

    static private final Recognizer	ANY = new Recognizer (null);


    // Base class defines the calls used to validate content,
    // and supports the "ANY" content model
    static private class Recognizer
    {
	final ElementInfo	type;

	Recognizer (ElementInfo t) { type = t; }

	// return true iff character data is legal here
	boolean acceptCharacters ()
	throws SAXException
	    // VC: Element Valid (third and fourth clauses)
	    { return true; }

	// null return = failure
	// otherwise, next state (like an FSM)
	// prerequisite: tested that name was declared
	Recognizer acceptElement (String name)
	throws SAXException
	    // VC: Element Valid (fourth clause)
	    { return this; }

	// return true iff model is completed, can finish
	boolean completed ()
	throws SAXException
	    // VC: Element Valid (fourth clause)
	    { return true; }
	
	public String toString ()
	    // n.b. "children" is the interesting case!
	    { return (type == null) ? "ANY" : type.model; }
    }

    // "EMPTY" content model -- no characters or elements
    private static final class EmptyRecognizer extends Recognizer
    {
	public EmptyRecognizer (ElementInfo type)
	    { super (type); }

	// VC: Element Valid (first clause)
	boolean acceptCharacters ()
	    { return false; }

	// VC: Element Valid (first clause)
	Recognizer acceptElement (String name)
	    { return null; }
    }

    // "Mixed" content model -- ANY, but restricts elements
    private static final class MixedRecognizer extends Recognizer
    {
	private String	permitted [];

	// N.B. constructor tests for duplicated element names (VC)
	public MixedRecognizer (ElementInfo t, ValidationConsumer v)
	throws SAXException
	{
	    super (t);

	    // (#PCDATA...)* or (#PCDATA) ==> ... or empty
	    // with the "..." being "|elname|..."
	    StringTokenizer	tokens = new StringTokenizer (
		t.model.substring (8, t.model.lastIndexOf (')')),
		"|");
	    Vector		vec = new Vector ();

	    while (tokens.hasMoreTokens ()) {
		String token = tokens.nextToken ();

		if (vec.contains (token))
		    v.error ("element " + token
			+ " is repeated in mixed content model: "
			+ t.model);
		else
		    vec.addElement (token.intern ());
	    }
	    permitted = new String [vec.size ()];
	    for (int i = 0; i < permitted.length; i++)
		permitted [i] = (String) vec.elementAt (i);
	    
	    // in one large machine-derived DTD sample, most of about
	    // 250 mixed content models were empty, and 25 had ten or
	    // more entries.  2 had over a hundred elements.  Linear
	    // search isn't obviously wrong.
	}

	// VC: Element Valid (third clause)
	Recognizer acceptElement (String name)
	{
	    int		length = permitted.length;

	    // first pass -- optimistic w.r.t. event source interning
	    // (and document validity)
	    for (int i = 0; i < length; i++)
		if (permitted [i] == name)
		    return this;
	    // second pass -- pessimistic w.r.t. event source interning
	    for (int i = 0; i < length; i++)
		if (permitted [i].equals (name))
		    return this;
	    return null;
	}
    }


    // recognizer loop flags, see later
    private static final int		F_LOOPHEAD = 0x01;
    private static final int		F_LOOPNEXT = 0x02;

    // for debugging -- used to label/count nodes in toString()
    private static int			nodeCount;

    /**
     * "Children" content model -- these are nodes in NDFA state graphs.
     * They work in fixed space.  Note that these graphs commonly have
     * cycles, handling features such as zero-or-more and one-or-more.
     *
     * <p>It's readonly, so only one copy is ever needed.  The content model
     * stack may have any number of pointers into each graph, when a model
     * happens to be needed more than once due to element nesting.  Since
     * traversing the graph just moves to another node, and never changes