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

   1 package net.sourceforge.phpdt.internal.ui.text;
   2
   3 /*
   4  * (c) Copyright IBM Corp. 2000, 2001.
   5  * All Rights Reserved.
   6  */
   7
   8 import java.io.IOException;
   9 import java.io.PushbackReader;
  10 import java.io.Reader;
  11 import java.util.HashMap;
  12 import java.util.HashSet;
  13 import java.util.Map;
  14 import java.util.Set;
  15
  16 import net.sourceforge.phpdt.internal.ui.PHPUIMessages;
  17
  18 import org.eclipse.jface.text.TextPresentation;
  19 import org.eclipse.swt.SWT;
  20 import org.eclipse.swt.custom.StyleRange;
  21
  22 /**
  23  * Reads the text contents from a reader of HTML contents and translates the
  24  * tags or cut them out.
  25  */
  26 public class HTML2TextReader extends SubstitutionTextReader {
  27
  28         private static final String LINE_DELIM = System.getProperty(
  29                         "line.separator", "\n"); //$NON-NLS-1$ //$NON-NLS-2$
  30
  31         private static final String EMPTY_STRING = ""; //$NON-NLS-1$
  32
  33         private static final Map fgEntityLookup;
  34
  35         private static final Set fgTags;
  36
  37         static {
  38
  39                 fgTags = new HashSet();
  40                 fgTags.add("b"); //$NON-NLS-1$
  41                 fgTags.add("br"); //$NON-NLS-1$
  42                 fgTags.add("h5"); //$NON-NLS-1$
  43                 fgTags.add("p"); //$NON-NLS-1$
  44                 fgTags.add("dl"); //$NON-NLS-1$
  45                 fgTags.add("dt"); //$NON-NLS-1$
  46                 fgTags.add("dd"); //$NON-NLS-1$
  47                 fgTags.add("li"); //$NON-NLS-1$
  48                 fgTags.add("ul"); //$NON-NLS-1$
  49                 fgTags.add("pre"); //$NON-NLS-1$
  50
  51                 fgEntityLookup = new HashMap(7);
  52                 fgEntityLookup.put("lt", "<"); //$NON-NLS-1$ //$NON-NLS-2$
  53                 fgEntityLookup.put("gt", ">"); //$NON-NLS-1$ //$NON-NLS-2$
  54                 fgEntityLookup.put("nbsp", " "); //$NON-NLS-1$ //$NON-NLS-2$
  55                 fgEntityLookup.put("amp", "&"); //$NON-NLS-1$ //$NON-NLS-2$
  56                 fgEntityLookup.put("circ", "^"); //$NON-NLS-1$ //$NON-NLS-2$
  57                 fgEntityLookup.put("tilde", "~"); //$NON-NLS-2$ //$NON-NLS-1$
  58                 fgEntityLookup.put("quot", "\""); //$NON-NLS-1$ //$NON-NLS-2$
  59         }
  60
  61         private int fCounter = 0;
  62
  63         private TextPresentation fTextPresentation;
  64
  65         private int fBold = 0;
  66
  67         private int fStartOffset = -1;
  68
  69         private boolean fInParagraph = false;
  70
  71         private boolean fIsPreformattedText = false;
  72
  73         /**
  74          * Transforms the html text from the reader to formatted text.
  75          *
  76          * @param presentation
  77          *            If not <code>null</code>, formattings will be applied to
  78          *            the presentation.
  79          */
  80         public HTML2TextReader(Reader reader, TextPresentation presentation) {
  81                 super(new PushbackReader(reader));
  82                 fTextPresentation = presentation;
  83         }
  84
  85         public int read() throws IOException {
  86                 int c = super.read();
  87                 if (c != -1)
  88                         ++fCounter;
  89                 return c;
  90         }
  91
  92         protected void startBold() {
  93                 if (fBold == 0)
  94                         fStartOffset = fCounter;
  95                 ++fBold;
  96         }
  97
  98         protected void startPreformattedText() {
  99                 fIsPreformattedText = true;
 100                 setSkipWhitespace(false);
 101         }
 102
 103         protected void stopPreformattedText() {
 104                 fIsPreformattedText = false;
 105                 setSkipWhitespace(true);
 106         }
 107
 108         protected void stopBold() {
 109                 --fBold;
 110                 if (fBold == 0) {
 111                         if (fTextPresentation != null) {
 112                                 fTextPresentation.addStyleRange(new StyleRange(fStartOffset,
 113                                                 fCounter - fStartOffset, null, null, SWT.BOLD));
 114                         }
 115                         fStartOffset = -1;
 116                 }
 117         }
 118
 119         /**
 120          * @see SubstitutionTextReader#computeSubstitution(char)
 121          */
 122         protected String computeSubstitution(int c) throws IOException {
 123
 124                 if (c == '<')
 125                         return processHTMLTag();
 126                 else if (c == '&')
 127                         return processEntity();
 128                 else if (fIsPreformattedText)
 129                         return processPreformattedText(c);
 130
 131                 return null;
 132         }
 133
 134         private String html2Text(String html) {
 135
 136                 String tag = html;
 137                 if ('/' == tag.charAt(0))
 138                         tag = tag.substring(1);
 139
 140                 if (!fgTags.contains(tag))
 141                         return EMPTY_STRING;
 142
 143                 if ("pre".equals(html)) { //$NON-NLS-1$
 144                         startPreformattedText();
 145                         return EMPTY_STRING;
 146                 }
 147
 148                 if ("/pre".equals(html)) { //$NON-NLS-1$
 149                         stopPreformattedText();
 150                         return EMPTY_STRING;
 151                 }
 152
 153                 if (fIsPreformattedText)
 154                         return EMPTY_STRING;
 155
 156                 if ("b".equals(html)) { //$NON-NLS-1$
 157                         startBold();
 158                         return EMPTY_STRING;
 159                 }
 160
 161                 if ("h5".equals(html) || "dt".equals(html)) { //$NON-NLS-1$ //$NON-NLS-2$
 162                         startBold();
 163                         return EMPTY_STRING;
 164                 }
 165
 166                 if ("dl".equals(html)) //$NON-NLS-1$
 167                         return LINE_DELIM;
 168
 169                 if ("dd".equals(html)) //$NON-NLS-1$
 170                         return "\t"; //$NON-NLS-1$
 171
 172                 if ("li".equals(html)) //$NON-NLS-1$
 173                         return LINE_DELIM
 174                                         + "\t" + PHPUIMessages.getString("HTML2TextReader.dash"); //$NON-NLS-1$ //$NON-NLS-2$
 175
 176                 if ("/b".equals(html)) { //$NON-NLS-1$
 177                         stopBold();
 178                         return EMPTY_STRING;
 179                 }
 180
 181                 if ("p".equals(html)) { //$NON-NLS-1$
 182                         fInParagraph = true;
 183                         return LINE_DELIM;
 184                 }
 185
 186                 if ("br".equals(html)) //$NON-NLS-1$
 187                         return LINE_DELIM;
 188
 189                 if ("/p".equals(html)) { //$NON-NLS-1$
 190                         boolean inParagraph = fInParagraph;
 191                         fInParagraph = false;
 192                         return inParagraph ? EMPTY_STRING : LINE_DELIM;
 193                 }
 194
 195                 if ("/h5".equals(html) || "/dt".equals(html)) { //$NON-NLS-1$ //$NON-NLS-2$
 196                         stopBold();
 197                         return LINE_DELIM;
 198                 }
 199
 200                 if ("/dd".equals(html)) //$NON-NLS-1$
 201                         return LINE_DELIM;
 202
 203                 return EMPTY_STRING;
 204         }
 205
 206         /*
 207          * A '<' has been read. Process a html tag
 208          */
 209         private String processHTMLTag() throws IOException {
 210
 211                 StringBuffer buf = new StringBuffer();
 212                 int ch;
 213                 do {
 214
 215                         ch = nextChar();
 216
 217                         while (ch != -1 && ch != '>') {
 218                                 buf.append(Character.toLowerCase((char) ch));
 219                                 ch = nextChar();
 220                                 if (ch == '"') {
 221                                         buf.append(Character.toLowerCase((char) ch));
 222                                         ch = nextChar();
 223                                         while (ch != -1 && ch != '"') {
 224                                                 buf.append(Character.toLowerCase((char) ch));
 225                                                 ch = nextChar();
 226                                         }
 227                                 }
 228                                 if (ch == '<') {
 229                                         unread(ch);
 230                                         return '<' + buf.toString();
 231                                 }
 232                         }
 233
 234                         if (ch == -1)
 235                                 return null;
 236
 237                         int tagLen = buf.length();
 238                         // needs special treatment for comments
 239                         if ((tagLen >= 3 && "!--".equals(buf.substring(0, 3))) //$NON-NLS-1$
 240                                         && !(tagLen >= 5 && "--!".equals(buf.substring(tagLen - 3)))) { //$NON-NLS-1$
 241                                 // unfinished comment
 242                                 buf.append(ch);
 243                         } else {
 244                                 break;
 245                         }
 246                 } while (true);
 247
 248                 return html2Text(buf.toString());
 249         }
 250
 251         private String processPreformattedText(int c) {
 252                 if (c == '\r' || c == '\n')
 253                         fCounter++;
 254                 return null;
 255         }
 256
 257         private void unread(int ch) throws IOException {
 258                 ((PushbackReader) getReader()).unread(ch);
 259         }
 260
 261         protected String entity2Text(String symbol) {
 262                 if (symbol.length() > 1 && symbol.charAt(0) == '#') {
 263                         int ch;
 264                         try {
 265                                 if (symbol.charAt(1) == 'x') {
 266                                         ch = Integer.parseInt(symbol.substring(2), 16);
 267                                 } else {
 268                                         ch = Integer.parseInt(symbol.substring(1), 10);
 269                                 }
 270                                 return EMPTY_STRING + (char) ch;
 271                         } catch (NumberFormatException e) {
 272                         }
 273                 } else {
 274                         String str = (String) fgEntityLookup.get(symbol);
 275                         if (str != null) {
 276                                 return str;
 277                         }
 278                 }
 279                 return "&" + symbol; // not found //$NON-NLS-1$
 280         }
 281
 282         /*
 283          * A '&' has been read. Process a entity
 284          */
 285         private String processEntity() throws IOException {
 286                 StringBuffer buf = new StringBuffer();
 287                 int ch = nextChar();
 288                 while (Character.isLetterOrDigit((char) ch) || ch == '#') {
 289                         buf.append((char) ch);
 290                         ch = nextChar();
 291                 }
 292
 293                 if (ch == ';')
 294                         return entity2Text(buf.toString());
 295
 296                 buf.insert(0, '&');
 297                 if (ch != -1)
 298                         buf.append((char) ch);
 299                 return buf.toString();
 300         }
 301 }