net.sourceforge.phpeclipse/src/net/sourceforge/phpdt/internal/ui/text/HTML2TextReader.java

   1 package net.sourceforge.phpdt.internal.ui.text;
   2
   3
   4 /*
   5  * (c) Copyright IBM Corp. 2000, 2001.
   6  * All Rights Reserved.
   7  */
   8
   9
  10 import java.io.IOException;
  11 import java.io.PushbackReader;
  12 import java.io.Reader;
  13 import java.util.HashMap;
  14 import java.util.HashSet;
  15 import java.util.Map;
  16 import java.util.Set;
  17
  18 import net.sourceforge.phpdt.internal.ui.PHPUIMessages;
  19
  20 import org.eclipse.jface.text.TextPresentation;
  21 import org.eclipse.swt.SWT;
  22 import org.eclipse.swt.custom.StyleRange;
  23
  24
  25
  26
  27 /**
  28  * Reads the text contents from a reader of HTML contents and translates
  29  * the tags or cut them out.
  30  */
  31 public class HTML2TextReader extends SubstitutionTextReader {
  32
  33
  34         private static final String LINE_DELIM= System.getProperty("line.separator", "\n"); //$NON-NLS-1$ //$NON-NLS-2$
  35         private static final String EMPTY_STRING= ""; //$NON-NLS-1$
  36         private static final Map fgEntityLookup;
  37         private static final Set fgTags;
  38
  39         static {
  40
  41                 fgTags= new HashSet();
  42                 fgTags.add("b"); //$NON-NLS-1$
  43                 fgTags.add("br"); //$NON-NLS-1$
  44                 fgTags.add("h5"); //$NON-NLS-1$
  45                 fgTags.add("p"); //$NON-NLS-1$
  46                 fgTags.add("dl"); //$NON-NLS-1$
  47                 fgTags.add("dt"); //$NON-NLS-1$
  48                 fgTags.add("dd"); //$NON-NLS-1$
  49                 fgTags.add("li"); //$NON-NLS-1$
  50                 fgTags.add("ul"); //$NON-NLS-1$
  51                 fgTags.add("pre"); //$NON-NLS-1$
  52
  53                 fgEntityLookup= new HashMap(7);
  54                 fgEntityLookup.put("lt", "<"); //$NON-NLS-1$ //$NON-NLS-2$
  55                 fgEntityLookup.put("gt", ">"); //$NON-NLS-1$ //$NON-NLS-2$
  56                 fgEntityLookup.put("nbsp", " "); //$NON-NLS-1$ //$NON-NLS-2$
  57                 fgEntityLookup.put("amp", "&"); //$NON-NLS-1$ //$NON-NLS-2$
  58                 fgEntityLookup.put("circ", "^"); //$NON-NLS-1$ //$NON-NLS-2$
  59                 fgEntityLookup.put("tilde", "~"); //$NON-NLS-2$ //$NON-NLS-1$
  60                 fgEntityLookup.put("quot", "\"");                //$NON-NLS-1$ //$NON-NLS-2$
  61         }
  62
  63         private int fCounter= 0;
  64         private TextPresentation fTextPresentation;
  65         private int fBold= 0;
  66         private int fStartOffset= -1;
  67         private boolean fInParagraph= false;
  68         private boolean fIsPreformattedText= false;
  69
  70         /**
  71          * Transforms the html text from the reader to formatted text.
  72          * @param presentation If not <code>null</code>, formattings will be applied to
  73          * the presentation.
  74         */
  75         public HTML2TextReader(Reader reader, TextPresentation presentation) {
  76                 super(new PushbackReader(reader));
  77                 fTextPresentation= presentation;
  78         }
  79
  80         public int read() throws IOException {
  81                 int c= super.read();
  82                 if (c != -1)
  83                         ++ fCounter;
  84                 return c;
  85         }
  86
  87         protected void startBold() {
  88                 if (fBold == 0)
  89                         fStartOffset= fCounter;
  90                 ++ fBold;
  91         }
  92
  93         protected void startPreformattedText() {
  94                 fIsPreformattedText= true;
  95                 setSkipWhitespace(false);
  96         }
  97
  98         protected void stopPreformattedText() {
  99                 fIsPreformattedText= false;
 100                 setSkipWhitespace(true);
 101         }
 102
 103         protected void stopBold() {
 104                 -- fBold;
 105                 if (fBold == 0) {
 106                         if (fTextPresentation != null) {
 107                                 fTextPresentation.addStyleRange(new StyleRange(fStartOffset, fCounter - fStartOffset, null, null, SWT.BOLD));
 108                         }
 109                         fStartOffset= -1;
 110                 }
 111         }
 112
 113         /**
 114          * @see SubstitutionTextReader#computeSubstitution(char)
 115          */
 116         protected String computeSubstitution(int c) throws IOException {
 117
 118                 if (c == '<')
 119                         return  processHTMLTag();
 120                 else if (c == '&')
 121                         return processEntity();
 122                 else if (fIsPreformattedText)
 123                         return processPreformattedText(c);
 124
 125                 return null;
 126         }
 127
 128         private String html2Text(String html) {
 129
 130                 String tag= html;
 131                 if ('/' == tag.charAt(0))
 132                         tag= tag.substring(1);
 133
 134                 if (!fgTags.contains(tag))
 135                         return EMPTY_STRING;
 136
 137
 138                 if ("pre".equals(html)) { //$NON-NLS-1$
 139                         startPreformattedText();
 140                         return EMPTY_STRING;
 141                 }
 142
 143                 if ("/pre".equals(html)) { //$NON-NLS-1$
 144                         stopPreformattedText();
 145                         return EMPTY_STRING;
 146                 }
 147
 148                 if (fIsPreformattedText)
 149                         return EMPTY_STRING;
 150
 151                 if ("b".equals(html)) { //$NON-NLS-1$
 152                         startBold();
 153                         return EMPTY_STRING;
 154                 }
 155
 156                 if ("h5".equals(html) || "dt".equals(html)) { //$NON-NLS-1$ //$NON-NLS-2$
 157                         startBold();
 158                         return EMPTY_STRING;
 159                 }
 160
 161                 if ("dl".equals(html)) //$NON-NLS-1$
 162                         return LINE_DELIM;
 163
 164                 if ("dd".equals(html)) //$NON-NLS-1$
 165                         return "\t"; //$NON-NLS-1$
 166
 167                 if ("li".equals(html)) //$NON-NLS-1$
 168                         return LINE_DELIM + "\t" + PHPUIMessages.getString("HTML2TextReader.dash"); //$NON-NLS-1$ //$NON-NLS-2$
 169
 170                 if ("/b".equals(html)) { //$NON-NLS-1$
 171                         stopBold();
 172                         return EMPTY_STRING;
 173                 }
 174
 175                 if ("p".equals(html))  { //$NON-NLS-1$
 176                         fInParagraph= true;
 177                         return LINE_DELIM;
 178                 }
 179
 180                 if ("br".equals(html)) //$NON-NLS-1$
 181                         return LINE_DELIM;
 182
 183                 if ("/p".equals(html))  { //$NON-NLS-1$
 184                         boolean inParagraph= fInParagraph;
 185                         fInParagraph= false;
 186                         return inParagraph ? EMPTY_STRING : LINE_DELIM;
 187                 }
 188
 189                 if ("/h5".equals(html) || "/dt".equals(html)) { //$NON-NLS-1$ //$NON-NLS-2$
 190                         stopBold();
 191                         return LINE_DELIM;
 192                 }
 193
 194                 if ("/dd".equals(html)) //$NON-NLS-1$
 195                         return LINE_DELIM;
 196
 197                 return EMPTY_STRING;
 198         }
 199
 200         /*
 201          * A '<' has been read. Process a html tag
 202          */
 203         private String processHTMLTag() throws IOException {
 204
 205                 StringBuffer buf= new StringBuffer();
 206                 int ch;
 207                 do {
 208
 209                         ch= nextChar();
 210
 211                         while (ch != -1 && ch != '>') {
 212                                 buf.append(Character.toLowerCase((char) ch));
 213                                 ch= nextChar();
 214                                 if (ch == '"'){
 215                                         buf.append(Character.toLowerCase((char) ch));
 216                                         ch= nextChar();
 217                                         while (ch != -1 && ch != '"'){
 218                                                 buf.append(Character.toLowerCase((char) ch));
 219                                                 ch= nextChar();
 220                                         }
 221                                 }
 222                                 if (ch == '<'){
 223                                         unread(ch);
 224                                         return '<' + buf.toString();
 225                                 }
 226                         }
 227
 228                         if (ch == -1)
 229                                 return null;
 230
 231                         int tagLen= buf.length();
 232                         // needs special treatment for comments
 233                         if ((tagLen >= 3 && "!--".equals(buf.substring(0, 3))) //$NON-NLS-1$
 234                                 && !(tagLen >= 5 && "--!".equals(buf.substring(tagLen - 3)))) { //$NON-NLS-1$
 235                                 // unfinished comment
 236                                 buf.append(ch);
 237                         } else {
 238                                 break;
 239                         }
 240                 } while (true);
 241
 242                 return html2Text(buf.toString());
 243         }
 244
 245         private String processPreformattedText(int c) {
 246                 if  (c == '\r' || c == '\n')
 247                         fCounter++;
 248                 return null;
 249         }
 250
 251
 252         private void unread(int ch) throws IOException {
 253                 ((PushbackReader) getReader()).unread(ch);
 254         }
 255
 256         protected String entity2Text(String symbol) {
 257                 if (symbol.length() > 1 && symbol.charAt(0) == '#') {
 258                         int ch;
 259                         try {
 260                                 if (symbol.charAt(1) == 'x') {
 261                                         ch= Integer.parseInt(symbol.substring(2), 16);
 262                                 } else {
 263                                         ch= Integer.parseInt(symbol.substring(1), 10);
 264                                 }
 265                                 return EMPTY_STRING + (char)ch;
 266                         } catch (NumberFormatException e) {
 267                         }
 268                 } else {
 269                         String str= (String) fgEntityLookup.get(symbol);
 270                         if (str != null) {
 271                                 return str;
 272                         }
 273                 }
 274                 return "&" + symbol; // not found //$NON-NLS-1$
 275         }
 276
 277         /*
 278          * A '&' has been read. Process a entity
 279          */
 280         private String processEntity() throws IOException {
 281                 StringBuffer buf= new StringBuffer();
 282                 int ch= nextChar();
 283                 while (Character.isLetterOrDigit((char)ch) || ch == '#') {
 284                         buf.append((char) ch);
 285                         ch= nextChar();
 286                 }
 287
 288                 if (ch == ';')
 289                         return entity2Text(buf.toString());
 290
 291                 buf.insert(0, '&');
 292                 if (ch != -1)
 293                         buf.append((char) ch);
 294                 return buf.toString();
 295         }
 296 }