|
1 package org.jahia.services.htmlparser;
2
3 import java.io.*;
4 import java.util.*;
5
6 import javax.xml.parsers.*;
7 import javax.xml.transform.*;
8 import javax.xml.transform.dom.*;
9 import javax.xml.transform.stream.*;
10
11 import org.jahia.registries.*;
12 import org.jahia.settings.*;
13 import org.jahia.utils.*;
14 import org.jahia.utils.fileparsers.*;
15 import org.jahia.utils.properties.*;
16 import org.w3c.dom.*;
17 import org.w3c.tidy.*;
18 import org.xml.sax.*;
19
20
29 public class TidyHtmlParser implements HtmlParser {
30
31 private static org.apache.log4j.Logger logger =
32 org.apache.log4j.Logger.getLogger(TidyHtmlParser.class);
33
34 public static String  AMPERSAND = "$$$amp$$$";
35 public static String AMPERSAND_SECONDPASS = "$$$amp_secondpass$$$";
36 public static String TIDYERRORS_TAG = "TIDYERRORS";
37
38 private static Vector newInlineTags = new Vector();
39 private static Vector newBlockLevelTags = new Vector();
40 private static Vector unrecognizedTags = new Vector();
41
42 private Properties config = new Properties();
43
44 public TidyHtmlParser(){}
45
46 public TidyHtmlParser(Properties config){
47 this.config = config;
48 if ( this.config == null ){
49 this.config = new Properties();
50 }
51 }
52
53
57 public void init(HtmlParserService htmlParserService){
58
59 SettingsBean settings = htmlParserService.getSettingsBean();
60 if ( settings == null ){
61 return;
62 }
63 String fileName = settings.getPropertiesFile().getProperty("tidyConfig");
64 if ( fileName == null || "".equals(fileName.trim()) ){
65 fileName = "tidy.properties";
66 }
67 StringBuffer buff = new StringBuffer (settings.getJahiaEtcDiskPath());
68 buff.append(File.separator);
69 buff.append("config");
70 buff.append(File.separator);
71 buff.append(fileName);
72 try {
73 PropertiesManager propManager = new PropertiesManager(buff.toString());
74 this.config = propManager.getPropertiesObject();
75 } catch ( Throwable t ){
76 logger.debug("Error loading tidy config file, use default settings ",t);
77 }
78 if ( this.config == null ){
79 this.config = new Properties();
80 }
81 }
82
83
91 public String parse(String inputString, Vector DOMVisitors){
92 return parse(inputString,-1,config,DOMVisitors);
93 }
94
95
104 public String parse(String inputString, Vector DOMVisitors,
105 int siteId){
106 if ( inputString == null || inputString.trim().equals("") ){
107 return inputString;
108 }
109 return parse(inputString,siteId,config,DOMVisitors);
110 }
111
112
122 public static String parse( String input,
123 int siteId,
124 Properties tidyConfig,
125 Vector DOMVisitors){
126
127 if (input == null || "".equals(input.trim())) {
128 return input;
129 }
130 String result = new String (input);
131 result = JahiaTools.replacePattern(result, "&", AMPERSAND);
132
133 ByteArrayInputStream strIn;
134 ByteArrayOutputStream strOut;
135 BufferedInputStream urlIn;
136 Tidy tidy = new Tidy();
137
138 Properties config = (Properties) tidyConfig.clone();
139 String val = tidyConfig.getProperty(TidyConfig.NEW_BLOCK_LEVEL_TAGS);
140 if (val == null) {
141 val = "";
142 }
143 String tag = null;
144 int size = newBlockLevelTags.size();
145 for (int i = 0; i < size; i++) {
146 tag = (String ) newBlockLevelTags.get(i);
147 if (val.length() == 0) {
148 val = tag;
149 } else {
150 val += "," + tag;
151 }
152 }
153 config.setProperty(TidyConfig.NEW_BLOCK_LEVEL_TAGS, val);
154
155 val = config.getProperty(TidyConfig.NEW_INLINE_TAGS);
156 if (val == null) {
157 val = "";
158 }
159 size = newInlineTags.size();
160 for (int i = 0; i < size; i++) {
161 tag = (String ) newInlineTags.get(i);
162 if (val.length() == 0) {
163 val = tag;
164 } else {
165 val += "," + tag;
166 }
167 }
168
169 config.setProperty(TidyConfig.NEW_INLINE_TAGS, val);
170
171 byte[] strByte = null;
173 String charSet = null; CharsetDetection charsetDet = new CharsetDetection();
175 try {
176 strByte = org.apache.commons.io.IOUtils.toByteArray(result);
177 strIn = new ByteArrayInputStream(strByte);
178 charsetDet.charsetDetection(strIn);
179 charSet = charsetDet.getCharset();
180
181 if ((config.getProperty(TidyConfig.CHAR_ENCODING) == null)
182 && "UTF-8".equalsIgnoreCase(charSet)) {
183 config.setProperty(TidyConfig.CHAR_ENCODING, "utf8");
184 }
185 } catch (Throwable t) {
186 }
187
188 tidy.setConfigurationFromProps(config);
189
190 try {
191 if (charSet == null) {
192 strByte = result.getBytes();
193 } else {
194 strByte = result.getBytes(charSet);
195 }
196 strIn = new ByteArrayInputStream(strByte);
197 strOut = new ByteArrayOutputStream();
198 ByteArrayOutputStream strErr = new ByteArrayOutputStream();
199 tidy.setErrout(new PrintWriter(strErr, true));
200 tidy.setShowWarnings(false);
201 tidy.parse(strIn, strOut);
202 strIn.reset();
203 String tmpValue = null;
204 if (charSet == null) {
205 tmpValue = strOut.toString();
206 } else {
207 tmpValue = strOut.toString(charSet);
208 }
209 tmpValue = JahiaTools.replacePattern(tmpValue, "&",
210 AMPERSAND_SECONDPASS);
211
212 if (tmpValue == null) {
213 tmpValue = "";
214 }
215 if (!"".equals(tmpValue.trim())) {
216 if (charSet == null) {
217 strByte = tmpValue.getBytes();
218 } else {
219 strByte = tmpValue.getBytes(charSet);
220 }
221 strIn = new ByteArrayInputStream(strByte);
222
231 DocumentBuilderFactory dfactory = DocumentBuilderFactory.
232 newInstance();
233
234 EntityResolver et = null;
235 try {
236 et = ServicesRegistry.getInstance().
237 getJahiaWebAppsDeployerService().getDtdEntityResolver();
238 } catch (Throwable t) {
239 }
240 DocumentBuilder docBuilder = dfactory.newDocumentBuilder();
241 if (et != null) {
242 docBuilder.setEntityResolver(et);
243 }
244 Document doc = docBuilder.parse(strIn);
245
246 TagRemover tagRemover =
247 new TagRemover();
248
249 synchronized (unrecognizedTags) {
250 size = unrecognizedTags.size();
251 for (int i = 0; i < size; i++) {
252 tagRemover.addTag((String ) unrecognizedTags.get(i));
253 }
254 }
255 doc = tagRemover.parseDOM(doc);
257
258 size = DOMVisitors.size();
259 for (int i = 0; i < size; i++) {
260 HtmlDOMVisitor visitor = (HtmlDOMVisitor) DOMVisitors.get(i);
261 doc = visitor.parseDOM(doc);
262 }
263
264 doc.normalize();
265
266 TransformerFactory tfactory = TransformerFactory.newInstance();
267
268 Transformer serializer = tfactory.newTransformer();
271
272 serializer.setOutputProperty(OutputKeys.METHOD, "html");
273 serializer.setOutputProperty(OutputKeys.INDENT, "yes");
274 if (charSet != null) {
275 serializer.setOutputProperty(OutputKeys.ENCODING, charSet);
276 }
277 strOut.reset();
280 serializer.transform(new DOMSource(doc),
281 new StreamResult(strOut));
282
283 if (tidy.getParseErrors() > 0) {
284 result = "<TIDYERRORS>\n" + strErr.toString() +
285 "</TIDYERRORS>";
286 } else {
287 if (charSet == null) {
288 result = strOut.toString();
289 } else {
290 result = strOut.toString(charSet);
291 }
292 }
293 result = JahiaTools.replacePattern(result, AMPERSAND_SECONDPASS,
294 "&");
295 result = JahiaTools.text2XMLEntityRef(result, 1);
296 result = JahiaTools.replacePattern(result, AMPERSAND, "&");
297
298 } else if (tidy.getParseErrors() > 0) {
299 String err = strErr.toString();
300 result = "<TIDYERRORS>\n" + err + "</TIDYERRORS>";
301 if (err.indexOf("is not recognized!") != -1) {
302 err = JahiaTools.replacePatternIgnoreCase(err.toLowerCase(),
303 " - error: ", "@@@");
304 String [] errors = org.jahia.utils.JahiaTools.getTokens(
305 err, "@@@");
306 if (errors.length > 0) {
307 String token = "";
308 ArrayList tags = new ArrayList();
309 tag = null;
310 String newInput = input;
311 int pos = -1;
312 for (int i = 0; i < errors.length; i++) {
313 token = errors[i];
314 pos = token.indexOf(" is not recognized!");
315 if (pos != -1) {
316 try {
317 tag = token.substring(0, pos);
318 if (!tag.startsWith("<")) {
319 synchronized (unrecognizedTags) {
321 if (unrecognizedTags.contains(tag)) {
322 continue;
323 } else {
324 unrecognizedTags.add(tag);
325 newInlineTags.add(tag);
326 }
327 }
328 } else {
329 tag = tag.substring(1, tag.length() - 1);
330 synchronized (unrecognizedTags) {
331 if (unrecognizedTags.contains(tag)) {
332 continue;
333 } else {
334 unrecognizedTags.add(tag);
335 newBlockLevelTags.add(tag);
336 }
337 }
338 }
339 } catch (Throwable t) {
340 }
341 }
342 }
343 result = parse(input, siteId, tidyConfig,
344 DOMVisitors);
345 }
346 }
347 }
348 } catch (Exception e) {
349 e.printStackTrace();
350 return input;
351 }
352 return result;
353 }
354
355 }
356 |
Popular Tags |
Java API By Example, From Geeks To Geeks.