src/html/htmlpars.cpp

   1 /////////////////////////////////////////////////////////////////////////////
   2 // Name:        htmlpars.cpp
   3 // Purpose:     wxHtmlParser class (generic parser)
   4 // Author:      Vaclav Slavik
   5 // RCS-ID:      $Id$
   6 // Copyright:   (c) 1999 Vaclav Slavik
   7 // Licence:     wxWindows Licence
   8 /////////////////////////////////////////////////////////////////////////////
   9
  10
  11 #ifdef __GNUG__
  12 #pragma implementation
  13 #endif
  14
  15 #include "wx/wxprec.h"
  16
  17 #include "wx/defs.h"
  18 #if wxUSE_HTML && wxUSE_STREAMS
  19
  20 #ifdef __BORDLANDC__
  21 #pragma hdrstop
  22 #endif
  23
  24 #ifndef WXPRECOMP
  25 #include "wx/wx.h"
  26 #endif
  27
  28 #include "wx/tokenzr.h"
  29 #include "wx/wfstream.h"
  30 #include "wx/url.h"
  31 #include "wx/fontmap.h"
  32 #include "wx/html/htmldefs.h"
  33 #include "wx/html/htmlpars.h"
  34
  35
  36
  37 //-----------------------------------------------------------------------------
  38 // wxHtmlParser
  39 //-----------------------------------------------------------------------------
  40
  41 IMPLEMENT_ABSTRACT_CLASS(wxHtmlParser,wxObject)
  42
  43 wxHtmlParser::wxHtmlParser()
  44     : wxObject(), m_Cache(NULL), m_HandlersHash(wxKEY_STRING),
  45       m_FS(NULL), m_HandlersStack(NULL)
  46 {
  47     m_entitiesParser = new wxHtmlEntitiesParser;
  48 }
  49
  50 wxHtmlParser::~wxHtmlParser()
  51 {
  52     delete m_HandlersStack;
  53     m_HandlersHash.Clear();
  54     m_HandlersList.DeleteContents(TRUE);
  55     m_HandlersList.Clear();
  56     delete m_entitiesParser;
  57 }
  58
  59 wxObject* wxHtmlParser::Parse(const wxString& source)
  60 {
  61     wxObject *result;
  62
  63     InitParser(source);
  64     DoParsing();
  65     result = GetProduct();
  66     DoneParser();
  67     return result;
  68 }
  69
  70 void wxHtmlParser::InitParser(const wxString& source)
  71 {
  72     SetSource(source);
  73 }
  74
  75 void wxHtmlParser::DoneParser()
  76 {
  77     delete m_Cache;
  78     m_Cache = NULL;
  79 }
  80
  81 void wxHtmlParser::SetSource(const wxString& src)
  82 {
  83     m_Source = src;
  84     delete m_Cache;
  85     m_Cache = new wxHtmlTagsCache(m_Source);
  86 }
  87
  88 void wxHtmlParser::DoParsing(int begin_pos, int end_pos)
  89 {
  90     if (end_pos <= begin_pos) return;
  91
  92     char c;
  93     char *temp = new char[end_pos - begin_pos + 1];
  94     int i;
  95     int templen;
  96
  97     templen = 0;
  98     i = begin_pos;
  99
 100     while (i < end_pos)
 101     {
 102         c = m_Source[(unsigned int) i];
 103
 104         // continue building word:
 105         if (c != '<')
 106             {
 107             temp[templen++] = c;
 108             i++;
 109         }
 110
 111         else if (c == '<')
 112             {
 113             wxHtmlTag tag(m_Source, i, end_pos, m_Cache, m_entitiesParser);
 114
 115             if (templen)
 116                 {
 117                 temp[templen] = 0;
 118                 AddText(temp);
 119                 templen = 0;
 120             }
 121             AddTag(tag);
 122             if (tag.HasEnding()) i = tag.GetEndPos2();
 123             else i = tag.GetBeginPos();
 124         }
 125     }
 126
 127     if (templen)
 128     { // last word of block :-(
 129         temp[templen] = 0;
 130         AddText(temp);
 131     }
 132     delete[] temp;
 133 }
 134
 135 void wxHtmlParser::AddTag(const wxHtmlTag& tag)
 136 {
 137     wxHtmlTagHandler *h;
 138     bool inner = FALSE;
 139
 140     h = (wxHtmlTagHandler*) m_HandlersHash.Get(tag.GetName());
 141     if (h)
 142         inner = h->HandleTag(tag);
 143     if (!inner)
 144     {
 145         if (tag.HasEnding())
 146             DoParsing(tag.GetBeginPos(), tag.GetEndPos1());
 147     }
 148 }
 149
 150 void wxHtmlParser::AddTagHandler(wxHtmlTagHandler *handler)
 151 {
 152     wxString s(handler->GetSupportedTags());
 153     wxStringTokenizer tokenizer(s, ", ");
 154
 155     while (tokenizer.HasMoreTokens())
 156         m_HandlersHash.Put(tokenizer.NextToken(), handler);
 157
 158     if (m_HandlersList.IndexOf(handler) == wxNOT_FOUND)
 159         m_HandlersList.Append(handler);
 160
 161     handler->SetParser(this);
 162 }
 163
 164 void wxHtmlParser::PushTagHandler(wxHtmlTagHandler *handler, wxString tags)
 165 {
 166     wxStringTokenizer tokenizer(tags, ", ");
 167     wxString key;
 168
 169     if (m_HandlersStack == NULL)
 170     {
 171         m_HandlersStack = new wxList;
 172         m_HandlersStack->DeleteContents(TRUE);
 173     }
 174
 175     m_HandlersStack->Insert(new wxHashTable(m_HandlersHash));
 176
 177     while (tokenizer.HasMoreTokens())
 178     {
 179         key = tokenizer.NextToken();
 180         m_HandlersHash.Delete(key);
 181         m_HandlersHash.Put(key, handler);
 182     }
 183 }
 184
 185 void wxHtmlParser::PopTagHandler()
 186 {
 187     wxNode *first;
 188
 189     if (m_HandlersStack == NULL ||
 190         (first = m_HandlersStack->GetFirst()) == NULL)
 191     {
 192         wxLogWarning(_("Warning: attempt to remove HTML tag handler from empty stack."));
 193         return;
 194     }
 195     m_HandlersHash = *((wxHashTable*) first->GetData());
 196     m_HandlersStack->DeleteNode(first);
 197 }
 198
 199 //-----------------------------------------------------------------------------
 200 // wxHtmlTagHandler
 201 //-----------------------------------------------------------------------------
 202
 203 IMPLEMENT_ABSTRACT_CLASS(wxHtmlTagHandler,wxObject)
 204
 205
 206 //-----------------------------------------------------------------------------
 207 // wxHtmlEntitiesParser
 208 //-----------------------------------------------------------------------------
 209
 210 IMPLEMENT_DYNAMIC_CLASS(wxHtmlEntitiesParser,wxObject)
 211
 212 wxHtmlEntitiesParser::wxHtmlEntitiesParser()
 213 #if wxUSE_WCHAR_T && !wxUSE_UNICODE
 214     : m_conv(NULL), m_encoding(wxFONTENCODING_SYSTEM)
 215 #endif
 216 {
 217 }
 218
 219 wxHtmlEntitiesParser::~wxHtmlEntitiesParser()
 220 {
 221     delete m_conv;
 222 }
 223
 224 void wxHtmlEntitiesParser::SetEncoding(wxFontEncoding encoding)
 225 {
 226 #if wxUSE_WCHAR_T && !wxUSE_UNICODE
 227     if (encoding == m_encoding) return;
 228     delete m_conv;
 229     m_conv = NULL;
 230     m_encoding = encoding;
 231     if (m_encoding != wxFONTENCODING_SYSTEM)
 232         m_conv = new wxCSConv(wxFontMapper::GetEncodingName(m_encoding));
 233 #endif
 234 }
 235
 236 wxString wxHtmlEntitiesParser::Parse(const wxString& input)
 237 {
 238     const wxChar *c, *last;
 239     const wxChar *in_str = input.c_str();
 240     wxString output;
 241
 242     for (c = in_str, last = in_str; *c != wxT('\0'); c++)
 243     {
 244         if (*c == wxT('&'))
 245         {
 246             if (c - last > 0)
 247                 output.append(last, c - last);
 248             if (++c == wxT('\0')) break;
 249             wxString entity;
 250             const wxChar *ent_s = c;
 251             for (; (*c >= wxT('a') && *c <= wxT('z')) ||
 252                    (*c >= wxT('A') && *c <= wxT('Z')) ||
 253                    (*c >= wxT('0') && *c <= wxT('9')) ||
 254                    *c == wxT('_') || *c == wxT('#'); c++) {}
 255             entity.append(ent_s, c - ent_s);
 256             if (*c == wxT(';')) c++;
 257             output << GetEntityChar(entity);
 258             last = c;
 259         }
 260     }
 261     if (*last != wxT('\0'))
 262         output.append(last);
 263     return output;
 264 }
 265
 266 struct wxHtmlEntityInfo
 267 {
 268     const wxChar *name;
 269     unsigned code;
 270 };
 271
 272 static int compar_entity(const void *key, const void *item)
 273 {
 274     return wxStrcmp((wxChar*)key, ((wxHtmlEntityInfo*)item)->name);
 275 }
 276
 277 wxChar wxHtmlEntitiesParser::GetCharForCode(unsigned code)
 278 {
 279 #if wxUSE_UNICODE
 280     return (wxChar)code;
 281 #elif wxUSE_WCHAR_T
 282     char buf[2];
 283     wchar_t wbuf[2];
 284     wbuf[0] = (wchar_t)code;
 285     wbuf[1] = 0;
 286     wxMBConv *conv = m_conv ? m_conv : &wxConvLocal;
 287     if (conv->WC2MB(buf, wbuf, 1) == (size_t)-1)
 288         return '?';
 289     return buf[0];
 290 #else
 291     return (code < 256) ? (wxChar)code : '?';
 292 #endif
 293 }
 294
 295 wxChar wxHtmlEntitiesParser::GetEntityChar(const wxString& entity)
 296 {
 297     unsigned code = 0;
 298
 299     if (entity[0] == wxT('#'))
 300     {
 301         const wxChar *ent_s = entity.c_str();
 302         const wxChar *format;
 303
 304         if (ent_s[1] == wxT('x') || ent_s[1] == wxT('X'))
 305         {
 306             format = wxT("%x");
 307             ent_s++;
 308         }
 309         else
 310             format = wxT("%u");
 311         ent_s++;
 312
 313         if (wxSscanf(ent_s, format, &code) != 1)
 314             code = 0;
 315     }
 316     else
 317     {
 318         static wxHtmlEntityInfo substitutions[] = {
 319             { wxT("AElig"),198 },
 320             { wxT("Aacute"),193 },
 321             { wxT("Acirc"),194 },
 322             { wxT("Agrave"),192 },
 323             { wxT("Alpha"),913 },
 324             { wxT("Aring"),197 },
 325             { wxT("Atilde"),195 },
 326             { wxT("Auml"),196 },
 327             { wxT("Beta"),914 },
 328             { wxT("Ccedil"),199 },
 329             { wxT("Chi"),935 },
 330             { wxT("Dagger"),8225 },
 331             { wxT("Delta"),916 },
 332             { wxT("ETH"),208 },
 333             { wxT("Eacute"),201 },
 334             { wxT("Ecirc"),202 },
 335             { wxT("Egrave"),200 },
 336             { wxT("Epsilon"),917 },
 337             { wxT("Eta"),919 },
 338             { wxT("Euml"),203 },
 339             { wxT("Gamma"),915 },
 340             { wxT("Iacute"),205 },
 341             { wxT("Icirc"),206 },
 342             { wxT("Igrave"),204 },
 343             { wxT("Iota"),921 },
 344             { wxT("Iuml"),207 },
 345             { wxT("Kappa"),922 },
 346             { wxT("Lambda"),923 },
 347             { wxT("Mu"),924 },
 348             { wxT("Ntilde"),209 },
 349             { wxT("Nu"),925 },
 350             { wxT("OElig"),338 },
 351             { wxT("Oacute"),211 },
 352             { wxT("Ocirc"),212 },
 353             { wxT("Ograve"),210 },
 354             { wxT("Omega"),937 },
 355             { wxT("Omicron"),927 },
 356             { wxT("Oslash"),216 },
 357             { wxT("Otilde"),213 },
 358             { wxT("Ouml"),214 },
 359             { wxT("Phi"),934 },
 360             { wxT("Pi"),928 },
 361             { wxT("Prime"),8243 },
 362             { wxT("Psi"),936 },
 363             { wxT("Rho"),929 },
 364             { wxT("Scaron"),352 },
 365             { wxT("Sigma"),931 },
 366             { wxT("THORN"),222 },
 367             { wxT("Tau"),932 },
 368             { wxT("Theta"),920 },
 369             { wxT("Uacute"),218 },
 370             { wxT("Ucirc"),219 },
 371             { wxT("Ugrave"),217 },
 372             { wxT("Upsilon"),933 },
 373             { wxT("Uuml"),220 },
 374             { wxT("Xi"),926 },
 375             { wxT("Yacute"),221 },
 376             { wxT("Yuml"),376 },
 377             { wxT("Zeta"),918 },
 378             { wxT("aacute"),225 },
 379             { wxT("acirc"),226 },
 380             { wxT("acute"),180 },
 381             { wxT("aelig"),230 },
 382             { wxT("agrave"),224 },
 383             { wxT("alefsym"),8501 },
 384             { wxT("alpha"),945 },
 385             { wxT("amp"),38 },
 386             { wxT("and"),8743 },
 387             { wxT("ang"),8736 },
 388             { wxT("aring"),229 },
 389             { wxT("asymp"),8776 },
 390             { wxT("atilde"),227 },
 391             { wxT("auml"),228 },
 392             { wxT("bdquo"),8222 },
 393             { wxT("beta"),946 },
 394             { wxT("brvbar"),166 },
 395             { wxT("bull"),8226 },
 396             { wxT("cap"),8745 },
 397             { wxT("ccedil"),231 },
 398             { wxT("cedil"),184 },
 399             { wxT("cent"),162 },
 400             { wxT("chi"),967 },
 401             { wxT("circ"),710 },
 402             { wxT("clubs"),9827 },
 403             { wxT("cong"),8773 },
 404             { wxT("copy"),169 },
 405             { wxT("crarr"),8629 },
 406             { wxT("cup"),8746 },
 407             { wxT("curren"),164 },
 408             { wxT("dArr"),8659 },
 409             { wxT("dagger"),8224 },
 410             { wxT("darr"),8595 },
 411             { wxT("deg"),176 },
 412             { wxT("delta"),948 },
 413             { wxT("diams"),9830 },
 414             { wxT("divide"),247 },
 415             { wxT("eacute"),233 },
 416             { wxT("ecirc"),234 },
 417             { wxT("egrave"),232 },
 418             { wxT("empty"),8709 },
 419             { wxT("emsp"),8195 },
 420             { wxT("ensp"),8194 },
 421             { wxT("epsilon"),949 },
 422             { wxT("equiv"),8801 },
 423             { wxT("eta"),951 },
 424             { wxT("eth"),240 },
 425             { wxT("euml"),235 },
 426             { wxT("euro"),8364 },
 427             { wxT("exist"),8707 },
 428             { wxT("fnof"),402 },
 429             { wxT("forall"),8704 },
 430             { wxT("frac12"),189 },
 431             { wxT("frac14"),188 },
 432             { wxT("frac34"),190 },
 433             { wxT("frasl"),8260 },
 434             { wxT("gamma"),947 },
 435             { wxT("ge"),8805 },
 436             { wxT("gt"),62 },
 437             { wxT("hArr"),8660 },
 438             { wxT("harr"),8596 },
 439             { wxT("hearts"),9829 },
 440             { wxT("hellip"),8230 },
 441             { wxT("iacute"),237 },
 442             { wxT("icirc"),238 },
 443             { wxT("iexcl"),161 },
 444             { wxT("igrave"),236 },
 445             { wxT("image"),8465 },
 446             { wxT("infin"),8734 },
 447             { wxT("int"),8747 },
 448             { wxT("iota"),953 },
 449             { wxT("iquest"),191 },
 450             { wxT("isin"),8712 },
 451             { wxT("iuml"),239 },
 452             { wxT("kappa"),954 },
 453             { wxT("lArr"),8656 },
 454             { wxT("lambda"),955 },
 455             { wxT("lang"),9001 },
 456             { wxT("laquo"),171 },
 457             { wxT("larr"),8592 },
 458             { wxT("lceil"),8968 },
 459             { wxT("ldquo"),8220 },
 460             { wxT("le"),8804 },
 461             { wxT("lfloor"),8970 },
 462             { wxT("lowast"),8727 },
 463             { wxT("loz"),9674 },
 464             { wxT("lrm"),8206 },
 465             { wxT("lsaquo"),8249 },
 466             { wxT("lsquo"),8216 },
 467             { wxT("lt"),60 },
 468             { wxT("macr"),175 },
 469             { wxT("mdash"),8212 },
 470             { wxT("micro"),181 },
 471             { wxT("middot"),183 },
 472             { wxT("minus"),8722 },
 473             { wxT("mu"),956 },
 474             { wxT("nabla"),8711 },
 475             { wxT("nbsp"),160 },
 476             { wxT("ndash"),8211 },
 477             { wxT("ne"),8800 },
 478             { wxT("ni"),8715 },
 479             { wxT("not"),172 },
 480             { wxT("notin"),8713 },
 481             { wxT("nsub"),8836 },
 482             { wxT("ntilde"),241 },
 483             { wxT("nu"),957 },
 484             { wxT("oacute"),243 },
 485             { wxT("ocirc"),244 },
 486             { wxT("oelig"),339 },
 487             { wxT("ograve"),242 },
 488             { wxT("oline"),8254 },
 489             { wxT("omega"),969 },
 490             { wxT("omicron"),959 },
 491             { wxT("oplus"),8853 },
 492             { wxT("or"),8744 },
 493             { wxT("ordf"),170 },
 494             { wxT("ordm"),186 },
 495             { wxT("oslash"),248 },
 496             { wxT("otilde"),245 },
 497             { wxT("otimes"),8855 },
 498             { wxT("ouml"),246 },
 499             { wxT("para"),182 },
 500             { wxT("part"),8706 },
 501             { wxT("permil"),8240 },
 502             { wxT("perp"),8869 },
 503             { wxT("phi"),966 },
 504             { wxT("pi"),960 },
 505             { wxT("piv"),982 },
 506             { wxT("plusmn"),177 },
 507             { wxT("pound"),163 },
 508             { wxT("prime"),8242 },
 509             { wxT("prod"),8719 },
 510             { wxT("prop"),8733 },
 511             { wxT("psi"),968 },
 512             { wxT("quot"),34 },
 513             { wxT("rArr"),8658 },
 514             { wxT("radic"),8730 },
 515             { wxT("rang"),9002 },
 516             { wxT("raquo"),187 },
 517             { wxT("rarr"),8594 },
 518             { wxT("rceil"),8969 },
 519             { wxT("rdquo"),8221 },
 520             { wxT("real"),8476 },
 521             { wxT("reg"),174 },
 522             { wxT("rfloor"),8971 },
 523             { wxT("rho"),961 },
 524             { wxT("rlm"),8207 },
 525             { wxT("rsaquo"),8250 },
 526             { wxT("rsquo"),8217 },
 527             { wxT("sbquo"),8218 },
 528             { wxT("scaron"),353 },
 529             { wxT("sdot"),8901 },
 530             { wxT("sect"),167 },
 531             { wxT("shy"),173 },
 532             { wxT("sigma"),963 },
 533             { wxT("sigmaf"),962 },
 534             { wxT("sim"),8764 },
 535             { wxT("spades"),9824 },
 536             { wxT("sub"),8834 },
 537             { wxT("sube"),8838 },
 538             { wxT("sum"),8721 },
 539             { wxT("sup"),8835 },
 540             { wxT("sup1"),185 },
 541             { wxT("sup2"),178 },
 542             { wxT("sup3"),179 },
 543             { wxT("supe"),8839 },
 544             { wxT("szlig"),223 },
 545             { wxT("tau"),964 },
 546             { wxT("there4"),8756 },
 547             { wxT("theta"),952 },
 548             { wxT("thetasym"),977 },
 549             { wxT("thinsp"),8201 },
 550             { wxT("thorn"),254 },
 551             { wxT("tilde"),732 },
 552             { wxT("times"),215 },
 553             { wxT("trade"),8482 },
 554             { wxT("uArr"),8657 },
 555             { wxT("uacute"),250 },
 556             { wxT("uarr"),8593 },
 557             { wxT("ucirc"),251 },
 558             { wxT("ugrave"),249 },
 559             { wxT("uml"),168 },
 560             { wxT("upsih"),978 },
 561             { wxT("upsilon"),965 },
 562             { wxT("uuml"),252 },
 563             { wxT("weierp"),8472 },
 564             { wxT("xi"),958 },
 565             { wxT("yacute"),253 },
 566             { wxT("yen"),165 },
 567             { wxT("yuml"),255 },
 568             { wxT("zeta"),950 },
 569             { wxT("zwj"),8205 },
 570             { wxT("zwnj"),8204 },
 571             {NULL, 0}};
 572         static size_t substitutions_cnt = 0;
 573
 574         if (substitutions_cnt == 0)
 575             while (substitutions[substitutions_cnt].code != 0)
 576                 substitutions_cnt++;
 577
 578         wxHtmlEntityInfo *info;
 579         info = (wxHtmlEntityInfo*) bsearch(entity.c_str(), substitutions,
 580                                            substitutions_cnt,
 581                                            sizeof(wxHtmlEntityInfo),
 582                                            compar_entity);
 583         if (info)
 584             code = info->code;
 585     }
 586
 587     if (code == 0)
 588         return wxT('?');
 589     else
 590         return GetCharForCode(code);
 591 }
 592
 593 #endif