src/main/java/jp/sourceforge/jindolf/parser/HtmlParser.java

   1 /*
   2  * XHTML parser
   3  *
   4  * License : The MIT License
   5  * Copyright(c) 2009 olyutorskii
   6  */
   7
   8 package jp.sourceforge.jindolf.parser;
   9
  10 import java.util.regex.Pattern;
  11 import jp.sourceforge.jindolf.corelib.PeriodType;
  12 import jp.sourceforge.jindolf.corelib.VillageState;
  13
  14 /**
  15  * 人狼BBS各種XHTML文字列のパースを行いハンドラに通知する。
  16  */
  17 public class HtmlParser extends AbstractParser{
  18
  19     private static final String SP = "\u0020";
  20
  21
  22     private BasicHandler basicHandler;
  23     private final TalkParser     talkParser     = new TalkParser(this);
  24     private final SysEventParser sysEventParser = new SysEventParser(this);
  25
  26     private final SeqRange rangepool_1 = new SeqRange();
  27     private final SeqRange rangepool_2 = new SeqRange();
  28
  29     /**
  30      * コンストラクタ。
  31      */
  32     public HtmlParser(){
  33         super();
  34         return;
  35     }
  36
  37     /**
  38      * {@link BasicHandler}ハンドラを登録する。
  39      * @param basicHandler ハンドラ
  40      */
  41     public void setBasicHandler(BasicHandler basicHandler){
  42         this.basicHandler = basicHandler;
  43         return;
  44     }
  45
  46     /**
  47      * {@link TalkHandler}ハンドラを登録する。
  48      * @param talkHandler ハンドラ
  49      */
  50     public void setTalkHandler(TalkHandler talkHandler){
  51         this.talkParser.setTalkHandler(talkHandler);
  52         return;
  53     }
  54
  55     /**
  56      * {@link SysEventHandler}ハンドラを登録する。
  57      * @param handler ハンドラ
  58      */
  59     public void setSysEventHandler(SysEventHandler handler){
  60         this.sysEventParser.setSysEventHandler(handler);
  61         return;
  62     }
  63
  64     private static final Pattern XMLDECL_PATTERN =
  65             compile("<\\?xml\u0020");
  66     private static final Pattern O_HTML_PATTERN =
  67             compile("<html\u0020");
  68     private static final Pattern TITLE_PATTERN =
  69             compile("<title>([^<]*)</title>");
  70     private static final Pattern O_BODY_PATTERN =
  71             compile("<body>");
  72     private static final Pattern O_DIVMAIN_PATTERN =
  73             compile("<div\u0020class=\"main\">");
  74
  75     /**
  76      * XHTML先頭部分のパース。
  77      * @throws HtmlParseException パースエラー
  78      */
  79     private void parseHead() throws HtmlParseException{
  80         setContextErrorMessage("lost head part");
  81
  82         SeqRange titleRange = this.rangepool_1;
  83
  84         lookingAtAffirm(XMLDECL_PATTERN);
  85         shrinkRegion();
  86
  87         findAffirm(O_HTML_PATTERN);
  88         shrinkRegion();
  89
  90         findAffirm(TITLE_PATTERN);
  91         titleRange.setLastMatchedGroupRange(getMatcher(), 1);
  92         shrinkRegion();
  93
  94         this.basicHandler.pageTitle(getContent(), titleRange);
  95
  96         findAffirm(O_BODY_PATTERN);
  97         shrinkRegion();
  98
  99         findAffirm(O_DIVMAIN_PATTERN);
 100         shrinkRegion();
 101
 102         return;
 103     }
 104
 105     private static final Pattern LOGINFORM_PATTERN =
 106             compile(
 107                   "("
 108                     +"<form"
 109                     +SP + "action=\"index\\.rb\""
 110                     +SP + "method=\"post\""
 111                     +SP + "class=\"login_form\""
 112                     +">"
 113                 + ")|("
 114                     +"<div"
 115                     +SP + "class=\"login_form\""
 116                     +">"
 117                 + ")"
 118             );
 119     private static final Pattern C_EDIV_PATTERN =
 120             compile(
 121                   SP_I
 122                 + "<a\u0020href=\"[^\"]*\">[^<]*</a>"
 123                 + SP_I
 124                 + "</div>"
 125             );
 126     private static final Pattern USERID_PATTERN =
 127             compile(
 128                   "name=\"user_id\""
 129                 + SP
 130                 + "value=\"([^\"]*)\""
 131             );
 132     private static final Pattern C_FORM_PATTERN =
 133             compile("</form>");
 134
 135     /**
 136      * ログインフォームのパース。
 137      * ログイン名までの認識を確認したのはF国のみ。
 138      * @throws HtmlParseException パースエラー
 139      */
 140     private void parseLoginForm() throws HtmlParseException{
 141         setContextErrorMessage("lost login form");
 142
 143         SeqRange accountRange = this.rangepool_1;
 144
 145         boolean isLand_E_Form;
 146         findAffirm(LOGINFORM_PATTERN);
 147         if(isGroupMatched(1)){
 148             isLand_E_Form = false;
 149         }else{                         // E国ログインフォーム検出
 150             isLand_E_Form = true;
 151         }
 152         shrinkRegion();
 153
 154         if(isLand_E_Form){
 155             lookingAtAffirm(C_EDIV_PATTERN);
 156             shrinkRegion();
 157             return;
 158         }else{
 159             findAffirm(USERID_PATTERN);
 160             accountRange.setLastMatchedGroupRange(getMatcher(), 1);
 161             shrinkRegion();
 162
 163             if(accountRange.length() > 0){
 164                 this.basicHandler
 165                     .loginName(getContent(), accountRange);
 166             }
 167
 168             findAffirm(C_FORM_PATTERN);
 169             shrinkRegion();
 170         }
 171
 172         return;
 173     }
 174
 175     private static final Pattern VILLAGEINFO_PATTERN =
 176             compile(
 177                  "([^<]+?)" +SP_I          // 最短一致数量子
 178                 +"<strong>"
 179                     +"\uff08"
 180                     +"([0-9]+)"                       // 月
 181                     +"/"
 182                     +"([0-9]+)"                       // 日
 183                     +SP
 184                     +"(?:(?:(午前)|(午後))\u0020)?"  // AMPM
 185                     +"([0-9]+)"                       // 時
 186                     +"(?:時\u0020|\\:)"
 187                     +"([0-9]+)"                       // 分
 188                     +"分?\u0020に更新"
 189                     +"\uff09"
 190                 +"</strong>"
 191             );
 192
 193     /**
 194      * 村に関する各種情報をパース。
 195      * @throws HtmlParseException パースエラー
 196      */
 197     private void parseVillageInfo() throws HtmlParseException{
 198         setContextErrorMessage("lose village information");
 199
 200         SeqRange villageRange = this.rangepool_1;
 201
 202         sweepSpace();
 203
 204         lookingAtAffirm(VILLAGEINFO_PATTERN);
 205         villageRange.setLastMatchedGroupRange(getMatcher(), 1);
 206
 207         int month  = parseGroupedInt(2);
 208         int day    = parseGroupedInt(3);
 209         int hour   = parseGroupedInt(6);
 210         int minute = parseGroupedInt(7);
 211         if(isGroupMatched(5)){  // 午後指定
 212             hour = (hour + 12) % 24;
 213         }
 214         shrinkRegion();
 215
 216         this.basicHandler.villageName(getContent(), villageRange);
 217         this.basicHandler.commitTime(month, day, hour, minute);
 218
 219         return;
 220     }
 221
 222     private static final Pattern O_PARAG_PATTERN = compile("<p>");
 223     private static final Pattern PERIODLINK_PATTERN =
 224             compile(
 225             "("
 226                 + "<span\u0020class=\"time\">"
 227             +")|(?:"
 228                 + "<a\u0020href=\"([^\"]*)\">"
 229             +")|("
 230                 + "</p>"
 231             +")"
 232             );
 233     private static final Pattern PERIOD_PATTERN =
 234             compile(
 235                 "(プロローグ)" +
 236             "|"+
 237                 "(エピローグ)" +
 238             "|"+
 239                 "(終了)" +
 240             "|"+
 241                 "([0-9]+)日目"
 242             );
 243     private static final Pattern C_SPAN_PATTERN   = compile("</span>");
 244     private static final Pattern C_ANCHOR_PATTERN = compile("</a>");
 245
 246     /**
 247      * Period間リンクをパース。
 248      * @throws HtmlParseException パースエラー
 249      */
 250     private void parsePeriodLink() throws HtmlParseException{
 251         setContextErrorMessage("lost period link");
 252
 253         SeqRange anchorRange = this.rangepool_1;
 254
 255         findAffirm(O_PARAG_PATTERN);
 256         shrinkRegion();
 257
 258         for(;;){
 259             Pattern closePattern;
 260             anchorRange.setInvalid();
 261
 262             sweepSpace();
 263             lookingAtAffirm(PERIODLINK_PATTERN);
 264             if(isGroupMatched(1)){
 265                 closePattern = C_SPAN_PATTERN;
 266             }else if(isGroupMatched(2)){
 267                 closePattern = C_ANCHOR_PATTERN;
 268                 anchorRange.setLastMatchedGroupRange(getMatcher(), 2);
 269             }else if(isGroupMatched(3)){
 270                 shrinkRegion();
 271                 break;
 272             }else{
 273                 assert false;
 274                 throw buildParseException();
 275             }
 276             shrinkRegion();
 277
 278             int day = -1;
 279             PeriodType periodType = null;
 280             lookingAtAffirm(PERIOD_PATTERN);
 281             if(isGroupMatched(1)){
 282                 periodType = PeriodType.PROLOGUE;
 283             }else if(isGroupMatched(2)){
 284                 periodType = PeriodType.EPILOGUE;
 285             }else if(isGroupMatched(3)){
 286                 periodType = null;
 287             }else if(isGroupMatched(4)){
 288                 periodType = PeriodType.PROGRESS;
 289                 day = parseGroupedInt(4);
 290             }else{
 291                 assert false;
 292                 throw buildParseException();
 293             }
 294             shrinkRegion();
 295
 296             lookingAtAffirm(closePattern);
 297             shrinkRegion();
 298
 299             this.basicHandler.periodLink(getContent(),
 300                                          anchorRange,
 301                                          periodType, day );
 302         }
 303
 304         return;
 305     }
 306
 307     private static final Pattern O_MESSAGE_PATTERN =
 308             compile("<div\u0020class=\"message(?:\u0020ch[0-9]+)?\">");
 309     private static final Pattern O_RELOAD_PATTERN =
 310             compile("<div\u0020id=\"reload\">");
 311     private static final Pattern O_MSGKIND_PATTERN =
 312             compile(
 313              "(?:"
 314                 +"<div\u0020class=\"(?:(announce)|(order)|(extra))\">"
 315             +")|(?:"
 316                 +"(?:"
 317                 +"(?:<a name=\"[^\"]*\">)?"
 318                 +SP_I
 319                 +"<span\u0020class=\"mes_no\">"
 320                     +"([0-9]+)\\."
 321                 +"</span>)?"
 322                 +SP_I
 323                 +"(?:</a>)?"
 324                 +SP_I
 325                 +"<a\u0020name=\"([^\"]*)\"(?:\u0020class=\"ch_name\")?>"
 326             +")"
 327             );
 328     private static final Pattern C_DIV_PATTERN = compile("</div>");
 329
 330     /**
 331      * 各種メッセージをパース。
 332      * @throws HtmlParseException パースエラー
 333      */
 334     private void parseMessage() throws HtmlParseException{
 335         setContextErrorMessage("lost message");
 336
 337         boolean skipGarbage = true;
 338
 339         for(;;){
 340             sweepSpace();
 341
 342             boolean matched;
 343             if(skipGarbage){
 344                 skipGarbage = false;
 345                 matched = findProbe(O_MESSAGE_PATTERN); // 最初の1回のみ
 346             }else{
 347                 matched = lookingAtProbe(O_MESSAGE_PATTERN);
 348             }
 349             if( ! matched ){
 350                 matched = lookingAtProbe(O_RELOAD_PATTERN);
 351                 if(matched){
 352                     shrinkRegion();
 353                     findAffirm(C_DIV_PATTERN);
 354                     shrinkRegion();
 355                     continue;
 356                 }
 357                 break;
 358             }
 359             shrinkRegion();
 360
 361             dispatchFamily();
 362
 363             lookingAtAffirm(C_DIV_PATTERN);
 364             shrinkRegion();
 365         }
 366
 367         return;
 368     }
 369
 370     /**
 371      * イベント種別によって処理を振り分ける。
 372      * @throws HtmlParseException パースエラー
 373      */
 374     private void dispatchFamily() throws HtmlParseException{
 375         sweepSpace();
 376
 377         SeqRange nameRange = this.rangepool_1;
 378
 379         lookingAtAffirm(O_MSGKIND_PATTERN);
 380         if(isGroupMatched(1)){
 381             shrinkRegion();
 382             this.sysEventParser.parseAnnounce();
 383         }else if(isGroupMatched(2)){
 384             shrinkRegion();
 385             this.sysEventParser.parseOrder();
 386         }else if(isGroupMatched(3)){
 387             shrinkRegion();
 388             this.sysEventParser.parseExtra();
 389         }else if(isGroupMatched(5)){
 390             nameRange.setLastMatchedGroupRange(getMatcher(), 5);
 391             int talkNo = -1;
 392             if(isGroupMatched(4)){
 393                 talkNo = parseGroupedInt(4);
 394             }
 395             shrinkRegion();
 396             this.talkParser.parseTalk(talkNo, nameRange);
 397         }else{
 398             assert false;
 399             throw buildParseException();
 400         }
 401
 402         return;
 403     }
 404
 405     private static final Pattern O_LISTTABLE_PATTERN =
 406             compile("<table\u0020class=\"list\">");
 407     private static final Pattern ACTIVEVILLAGE =
 408             compile(
 409              "("
 410                 +"</table>"
 411             +")|(?:"
 412                 +"<tr><td>"
 413                 +"<a\u0020href=\"([^\"]*)\">([^<]*)</a>"
 414                 +"\u0020<strong>\uff08"
 415                     +"(?:(?:(午前)|(午後))\u0020)?"  // AMPM
 416                     +"([0-9]+)"                       // 時
 417                     +"(?:時\u0020|\\:)"
 418                     +"([0-9]+)"                       // 分
 419                     +"分?\u0020更新"
 420                 +"\uff09</strong>"
 421                 +"</td><td>(?:"
 422                 +"[^<]*"
 423                     + "(参加者募集中です。)"
 424                     +"|(開始待ちです。)"
 425                     +"|(進行中です。)"
 426                     +"|(勝敗が決定しました。)"
 427                     +"|(終了・ログ公開中。)"
 428                 +")</td></tr>"
 429             +")"
 430             );
 431
 432     /**
 433      * トップページの村一覧表のパース。
 434      * @throws HtmlParseException パースエラー
 435      */
 436     private void parseTopList() throws HtmlParseException{
 437         setContextErrorMessage("lost village list");
 438
 439         SeqRange anchorRange  = this.rangepool_1;
 440         SeqRange villageRange = this.rangepool_2;
 441
 442         if( ! findProbe(O_LISTTABLE_PATTERN) ) return;
 443         shrinkRegion();
 444
 445         for(;;){
 446             lookingAtAffirm(ACTIVEVILLAGE);
 447             if(isGroupMatched(1)) break;
 448             anchorRange .setLastMatchedGroupRange(getMatcher(), 2);
 449             villageRange.setLastMatchedGroupRange(getMatcher(), 3);
 450             int hour = parseGroupedInt(6);
 451             if(isGroupMatched(5)){
 452                 hour = (hour + 12) % 24;
 453             }
 454             int minute = parseGroupedInt(7);
 455
 456             VillageState state;
 457             if(isGroupMatched(8)){
 458                 state = VillageState.PROLOGUE;
 459             }else if(isGroupMatched(9)){
 460                 state = VillageState.PROLOGUE;
 461             }else if(isGroupMatched(10)){
 462                 state = VillageState.PROGRESS;
 463             }else if(isGroupMatched(11)){
 464                 state = VillageState.EPILOGUE;
 465             }else if(isGroupMatched(12)){
 466                 state = VillageState.GAMEOVER;
 467             }else{
 468                 assert false;
 469                 throw buildParseException();
 470             }
 471
 472             shrinkRegion();
 473
 474             sweepSpace();
 475
 476             this.basicHandler.villageRecord(getContent(),
 477                                             anchorRange,
 478                                             villageRange,
 479                                             hour, minute,
 480                                             state );
 481         }
 482
 483         return;
 484     }
 485
 486     private static final Pattern O_LISTLOG_PATTERN =
 487             compile(
 488             "<a\u0020href=\"(index[^\"]*(?:ready_0|000_ready))\">"
 489             +"([^<]*)"
 490             +"</a><br\u0020/>"
 491             );
 492
 493     /**
 494      * 村一覧ページのパース。
 495      * @throws HtmlParseException パースエラー
 496      */
 497     private void parseLogList() throws HtmlParseException{
 498         setContextErrorMessage("lost village list");
 499
 500         SeqRange anchorRange  = this.rangepool_1;
 501         SeqRange villageRange = this.rangepool_2;
 502
 503         boolean is1st = true;
 504         for(;;){
 505             boolean matched;
 506             if(is1st){
 507                 matched = findProbe(O_LISTLOG_PATTERN);
 508                 is1st = false;
 509             }else{
 510                 matched = lookingAtProbe(O_LISTLOG_PATTERN);
 511             }
 512             if( ! matched ) break;
 513
 514             anchorRange .setLastMatchedGroupRange(getMatcher(), 1);
 515             villageRange.setLastMatchedGroupRange(getMatcher(), 2);
 516
 517             shrinkRegion();
 518
 519             this.basicHandler.villageRecord(getContent(),
 520                                             anchorRange,
 521                                             villageRange,
 522                                             -1, -1,
 523                                             VillageState.GAMEOVER );
 524         }
 525
 526         return;
 527     }
 528
 529     private static final Pattern C_BODY_PATTERN =
 530             compile("</body>");
 531     private static final Pattern C_HTML_PATTERN =
 532             compile(SP_I+ "</html>" +SP_I);
 533
 534     /**
 535      * XHTML末尾のパース。
 536      * @throws HtmlParseException パースエラー
 537      */
 538     private void parseTail() throws HtmlParseException{
 539         setContextErrorMessage("lost last part");
 540
 541         findAffirm(C_BODY_PATTERN);
 542         shrinkRegion();
 543
 544         matchesAffirm(C_HTML_PATTERN);
 545         shrinkRegion();
 546
 547         return;
 548     }
 549
 550     private static final Pattern LISTTITLE_PATTERN =
 551             compile("終了した村の記録");
 552
 553     /**
 554      * 人狼BBSのページ種別を自動認識しつつパースする。
 555      * @param content パース対象の文字列
 556      * @throws HtmlParseException パースエラー
 557      */
 558     public void parseAutomatic(DecodedContent content)
 559             throws HtmlParseException{
 560         setContent(content);
 561
 562         this.basicHandler.startParse(getContent());
 563
 564         parseHead();
 565
 566         sweepSpace();
 567
 568         if(lookingAtProbe(LISTTITLE_PATTERN)){
 569             shrinkRegion();
 570             this.basicHandler.pageType(PageType.VILLAGELIST_PAGE);
 571             parseLogList();
 572         }else{
 573             parseLoginForm();
 574             sweepSpace();
 575             if(lookingAtProbe(O_PARAG_PATTERN)){
 576                 shrinkRegion();
 577                 this.basicHandler.pageType(PageType.TOP_PAGE);
 578                 parseTopList();
 579             }else{
 580                 this.basicHandler.pageType(PageType.PERIOD_PAGE);
 581                 parseVillageInfo();
 582                 parsePeriodLink();
 583                 parseMessage();
 584             }
 585         }
 586
 587         parseTail();
 588
 589         this.basicHandler.endParse();
 590
 591         reset();
 592
 593         return;
 594     }
 595
 596 }