1 // Copyright 2003 Adam Megacz, see the COPYING file for licensing [GPL]
2 package org.xwt.translators;
11 * While entities are limited to a subset of Unicode characters ,
12 * numeric character references can specify any character. Numeric
13 * character references may be given in decimal or hexadecimal, though
14 * browser support is stronger for decimal references. Decimal
15 * references are of the form &#number; while hexadecimal references
16 * take the case-insensitive form &#xnumber;. Examples of numeric
17 * character references include © or © for the copyright
18 * symbol, Α or Α for the Greek capital letter alpha, and
19 * ا or ا for the Arabic letter ALEF.
21 * http://www.htmlhelp.com/reference/html40/entities/special.html
22 * http://www.htmlhelp.com/reference/html40/entities/symbols.html
23 * http://www.htmlhelp.com/reference/html40/entities/latin1.html
27 * This class parses an InputStream containing HTML and returns it
28 * as an XWT DOM tree. Each HTML Element is returned as a struct,
29 * with the following members:
31 * Since HTML may have multiple top level elements (unlike XML),
32 * this class will search all top level elements for one with a tag
33 * name 'html'. If such a node is found, only it is returned. If no
34 * top-level element has the tag name 'html', such a node is
35 * fabricated, and all top level elements become the children of
36 * that node, which is then returned.
40 private final static String[] noEndTag =
41 new String[] { "area", "base", "basefont", "br", "col", "frame", "hr", "img",
42 "input", "isindex", "link", "meta", "param" };
44 /** we keep a char[] around for use by removeRedundantWhitespace() */
45 private static char[] cbuf = null;
47 /** we keep a StringBuffer around for use by removeRedundantWhitespace() */
48 private static StringBuffer sbuf = null;
50 /** true iff we have encountered an LI more recently than the last OL/UL */
51 private static boolean withinLI = false;
53 public static synchronized JS parseReader(Reader r) throws IOException, JSExn {
54 CharStream cs = new CharStream(r);
58 h.put("$name", "html");
61 while (true) parseBody(cs, h, null);
62 } catch (EOFException e) {
63 // continue until we get an EOFException
67 Object[] ids = h.keys();
68 for(int i=0; i<ids.length; i++) {
69 Object el = h.get((String)ids[i]);
70 if (el instanceof JS && "html".equals(((JS)el).get("$name")))
78 * Parses a single element and stores it in <tt>h</tt>. The
79 * CharStream should be positioned immediately <i>after</i> the
82 * If a close tag not matching this open tag is found, the
83 * tagname on the close tag will be returned in order to
84 * facilitate correcting broken HTML. Otherwise, this returns
87 private static String parseElement(CharStream cs, JS h) throws IOException, JSExn {
89 while(Character.isSpace(cs.peek())) cs.get();
90 String elementName = parseElementName(cs);
92 boolean saveWithinLI = withinLI;
93 if (elementName.equals("li")) {
95 cs.unread(new char[] { '<', 'l', 'i', ' ' });
100 } else if (elementName.equals("ol") || elementName.equals("ul")) {
104 h.put("$name", elementName);
105 if (elementName.equals("!--")) {
106 h.put("0", parseComment(cs));
107 h.put("$numchildren", new Integer(0));
112 while (cs.peek() != '>') {
113 String name = parseAttributeName(cs);
114 if (name.equals("")) break;
115 String value = expandEntities(parseAttributeValue(cs));
119 // eat the close-angle bracket
122 // bodyless tags return here
123 for(int i=0; i<noEndTag.length; i++)
124 if (noEndTag[i].equals(elementName))
128 String ret = parseBody(cs, h, elementName);
129 withinLI = saveWithinLI;
134 * Parses the body of an element. The CharStream should be
135 * positioned at the character immediately after the right
136 * bracket closing the start-tag
138 private static String parseBody(CharStream cs, JS h, String elementName) throws IOException, JSExn {
140 int length = h.get("$numchildren") == null ? 0 : Integer.parseInt(h.get("$numchildren").toString());
142 String closetag = null;
146 if (c != '<') { cdata += c; continue; }
147 String expanded = removeRedundantWhitespace(expandEntities(cdata));
148 if (expanded.length() > 0) {
149 h.put(String.valueOf(length), expanded);
150 h.put("$numchildren", new Integer(++length));
154 } catch (EOFException e) {
155 String expanded = removeRedundantWhitespace(expandEntities(cdata));
156 if (expanded.length() > 0) {
157 h.put(String.valueOf(length), expanded);
158 h.put("$numchildren", new Integer(++length));
165 if (cs.peek() != '/') {
167 closetag = parseElement(cs, kid);
168 h.put(String.valueOf(length), kid);
169 h.put("$numchildren", new Integer(++length));
173 cs.get(); // drop the slash
174 closetag = parseElementName(cs);
175 while(cs.get() != '>');
177 } catch (EOFException e) {
182 if (closetag != null)
183 return closetag.equals(elementName) ? null : closetag;
187 /** Parses an element name and returns it. The CharStream should
188 * be positioned at the first character of the name.
190 private static String parseElementName(CharStream cs) throws IOException, JSExn {
192 while (cs.peek() != '>' && !Character.isSpace(cs.peek())) ret += cs.get();
193 return ret.toLowerCase();
196 /** Parses an attribute name and returns it. The CharStream should
197 * be positioned at the first character of the name, possibly
198 * with intervening whitespace.
200 private static String parseAttributeName(CharStream cs) throws IOException, JSExn {
201 while(Character.isSpace(cs.peek())) cs.get();
203 while(!Character.isSpace(cs.peek()) && cs.peek() != '=' && cs.peek() != '>') ret += cs.get();
204 return ret.toLowerCase();
207 /** Parses an attribute value and returns it. The CharStream
208 * should be positioned at the equals sign, possibly with
209 * intervening whitespace.
211 private static String parseAttributeValue(CharStream cs) throws IOException, JSExn {
213 // eat whitespace and equals sign
214 while(Character.isSpace(cs.peek())) cs.get();
215 if (cs.peek() != '=') return "";
217 while(Character.isSpace(cs.peek())) cs.get();
219 boolean doublequoted = false;
220 boolean singlequoted = false;
223 if (cs.peek() == '\"') { doublequoted = true; cs.get(); }
224 else if (cs.peek() == '\'') { singlequoted = true; cs.get(); }
228 if (!doublequoted && !singlequoted && (Character.isSpace(c) || c == '>')) break;
229 if (singlequoted && c == '\'') { cs.get(); break; }
230 if (doublequoted && c == '\"') { cs.get(); break; }
236 /** Parses a comment and returns its body. The CharStream should
237 * be positioned immediately after the <!--
239 private static String parseComment(CharStream cs) throws IOException, JSExn {
244 if (c == '>' && dashes == 2) return ret.substring(0, ret.length() - 2);
245 if (c == '-') dashes++;
251 /** Expands all SGML entities in string <tt>s</tt> */
252 public static String expandEntities(String s) throws IOException, JSExn {
253 if (s.indexOf('&') == -1) return s;
254 StringBuffer sb = new StringBuffer();
257 while(nextamp != -1) {
258 nextamp = s.indexOf('&', i);
259 sb.append(nextamp == -1 ? s.substring(i) : s.substring(i, nextamp));
260 if (nextamp == -1) break;
261 if (s.regionMatches(nextamp, "&", 0, 5)) {
264 } else if (s.regionMatches(nextamp, ">", 0, 4)) {
267 } else if (s.regionMatches(nextamp, "<", 0, 4)) {
270 } else if (s.regionMatches(nextamp, """, 0, 6)) {
273 } else if (s.regionMatches(nextamp, " ", 0, 6)) {
274 // FEATURE: perhaps we should distinguish this somehow
282 return sb.toString();
285 /** removes all redundant whitespace */
286 private static String removeRedundantWhitespace(String s) throws JSExn {
288 if (s.indexOf(' ') == -1 && s.indexOf('\n') == -1 && s.indexOf('\t') == -1 && s.indexOf('\r') == -1) return s;
290 int len = s.length();
291 if (cbuf == null || cbuf.length < len) {
292 cbuf = new char[len * 2];
293 sbuf = new StringBuffer(len * 2);
296 s.getChars(0, len, cbuf, 0);
299 boolean lastWasWhitespace = false;
300 for(int i=0; i<len; i++) {
301 boolean lastlast = lastWasWhitespace;
303 case '\n': case '\r': case '\t':
306 lastWasWhitespace = true;
309 lastWasWhitespace = false;
312 if (lastWasWhitespace && lastlast) {
313 if (last != i) sbuf.append(cbuf, last, i - last);
318 if (last != len) sbuf.append(cbuf, last, len - last);
319 return sbuf.toString().trim();
322 // CharStream /////////////////////////////////////////////////////////////////////
324 private static class CharStream extends PushbackReader {
325 public CharStream(Reader r) { super(r, 1024); }
327 public char peek() throws IOException {
333 public char get() throws IOException {
335 if (i == -1) throw new EOFException();