re.java

linux下建立JAVA虚拟机的源码KAFFE

      // ALTERNATION OPERATOR
      //  \| or | (if RE_NO_BK_VBAR) or newline (if RE_NEWLINE_ALT)
      //  not available if RE_LIMITED_OPS is set

      // TODO: the '\n' literal here should be a test against REToken.newline,
      // which unfortunately may be more than a single character.
      if ( ( (unit.ch == '|' && (syntax.get(RESyntax.RE_NO_BK_VBAR) ^ (unit.bk || quot)))
	     || (syntax.get(RESyntax.RE_NEWLINE_ALT) && (unit.ch == '\n') && !(unit.bk || quot)) )
	   && !syntax.get(RESyntax.RE_LIMITED_OPS)) {
	// make everything up to here be a branch. create vector if nec.
	addToken(currentToken);
	RE theBranch = new RE(firstToken, lastToken, numSubs, subIndex, minimumLength, maximumLength);
	minimumLength = 0;
	maximumLength = 0;
	if (branches == null) {
	    branches = new Vector();
	}
	branches.addElement(theBranch);
	firstToken = lastToken = currentToken = null;
      }
      
      // INTERVAL OPERATOR:
      //  {x} | {x,} | {x,y}  (RE_INTERVALS && RE_NO_BK_BRACES)
      //  \{x\} | \{x,\} | \{x,y\} (RE_INTERVALS && !RE_NO_BK_BRACES)
      //
      // OPEN QUESTION: 
      //  what is proper interpretation of '{' at start of string?
      //
      // This method used to check "repeat.empty.token" to avoid such regexp
      // as "(a*){2,}", but now "repeat.empty.token" is allowed.

      else if ((unit.ch == '{') && syntax.get(RESyntax.RE_INTERVALS) && (syntax.get(RESyntax.RE_NO_BK_BRACES) ^ (unit.bk || quot))) {
	int newIndex = getMinMax(pattern,index,minMax,syntax);
        if (newIndex > index) {
          if (minMax.first > minMax.second)
            throw new REException(getLocalizedMessage("interval.order"),REException.REG_BADRPT,newIndex);
          if (currentToken == null)
            throw new REException(getLocalizedMessage("repeat.no.token"),REException.REG_BADRPT,newIndex);
          if (currentToken instanceof RETokenRepeated) 
            throw new REException(getLocalizedMessage("repeat.chained"),REException.REG_BADRPT,newIndex);
          if (currentToken instanceof RETokenWordBoundary || currentToken instanceof RETokenWordBoundary)
            throw new REException(getLocalizedMessage("repeat.assertion"),REException.REG_BADRPT,newIndex);
          index = newIndex;
          currentToken = setRepeated(currentToken,minMax.first,minMax.second,index); 
        }
        else {
          addToken(currentToken);
          currentToken = new RETokenChar(subIndex,unit.ch,insens);
        } 
      }
      
      // LIST OPERATOR:
      //  [...] | [^...]

      else if ((unit.ch == '[') && !(unit.bk || quot)) {
	// Create a new RETokenOneOf
	ParseCharClassResult result = parseCharClass(
		subIndex, pattern, index, pLength, cflags, syntax, 0);
	addToken(currentToken);
	currentToken = result.token;
	index = result.index;
      }

      // SUBEXPRESSIONS
      //  (...) | \(...\) depending on RE_NO_BK_PARENS

      else if ((unit.ch == '(') && (syntax.get(RESyntax.RE_NO_BK_PARENS) ^ (unit.bk || quot))) {
	boolean pure = false;
	boolean comment = false;
        boolean lookAhead = false;
        boolean lookBehind = false;
        boolean independent = false;
        boolean negativelh = false;
        boolean negativelb = false;
	if ((index+1 < pLength) && (pattern[index] == '?')) {
	  switch (pattern[index+1]) {
          case '!':
            if (syntax.get(RESyntax.RE_LOOKAHEAD)) {
              pure = true;
              negativelh = true;
              lookAhead = true;
              index += 2;
            }
            break;
          case '=':
            if (syntax.get(RESyntax.RE_LOOKAHEAD)) {
              pure = true;
              lookAhead = true;
              index += 2;
            }
            break;
	  case '<':
	    // We assume that if the syntax supports look-ahead,
	    // it also supports look-behind.
	    if (syntax.get(RESyntax.RE_LOOKAHEAD)) {
		index++;
		switch (pattern[index +1]) {
		case '!':
		  pure = true;
		  negativelb = true;
		  lookBehind = true;
		  index += 2;
		  break;
		case '=':
		  pure = true;
		  lookBehind = true;
		  index += 2;
		}
	    }
	    break;
	  case '>':
	    // We assume that if the syntax supports look-ahead,
	    // it also supports independent group.
            if (syntax.get(RESyntax.RE_LOOKAHEAD)) {
              pure = true;
              independent = true;
              index += 2;
            }
            break;
	  case 'i':
	  case 'd':
	  case 'm':
	  case 's':
	  // case 'u':  not supported
	  // case 'x':  not supported
	  case '-':
            if (!syntax.get(RESyntax.RE_EMBEDDED_FLAGS)) break;
	    // Set or reset syntax flags.
	    int flagIndex = index + 1;
	    int endFlag = -1;
	    RESyntax newSyntax = new RESyntax(syntax);
	    int newCflags = cflags;
	    boolean negate = false;
	    while (flagIndex < pLength && endFlag < 0) {
	        switch(pattern[flagIndex]) {
	  	case 'i':
		  if (negate)
		    newCflags &= ~REG_ICASE;
		  else
		    newCflags |= REG_ICASE;
		  flagIndex++;
		  break;
	  	case 'd':
		  if (negate)
		    newSyntax.setLineSeparator(RESyntax.DEFAULT_LINE_SEPARATOR);
		  else
		    newSyntax.setLineSeparator("\n");
		  flagIndex++;
		  break;
	  	case 'm':
		  if (negate)
		    newCflags &= ~REG_MULTILINE;
		  else
		    newCflags |= REG_MULTILINE;
		  flagIndex++;
		  break;
	  	case 's':
		  if (negate)
		    newCflags &= ~REG_DOT_NEWLINE;
		  else
		    newCflags |= REG_DOT_NEWLINE;
		  flagIndex++;
		  break;
	  	// case 'u': not supported
	  	// case 'x': not supported
	  	case '-':
		  negate = true;
		  flagIndex++;
		  break;
		case ':':
		case ')':
		  endFlag = pattern[flagIndex];
		  break;
		default:
            	  throw new REException(getLocalizedMessage("repeat.no.token"), REException.REG_BADRPT, index);
		}
	    }
	    if (endFlag == ')') {
		syntax = newSyntax;
		cflags = newCflags;
		insens = ((cflags & REG_ICASE) > 0);
		// This can be treated as though it were a comment.
		comment = true;
		index = flagIndex - 1;
		break;
	    }
	    if (endFlag == ':') {
		savedSyntax = syntax;
		savedCflags = cflags;
		flagsSaved = true;
		syntax = newSyntax;
		cflags = newCflags;
		insens = ((cflags & REG_ICASE) > 0);
		index = flagIndex -1;
		// Fall through to the next case.
	    }
	    else {
	        throw new REException(getLocalizedMessage("unmatched.paren"), REException.REG_ESUBREG,index);
	    }
	  case ':':
	    if (syntax.get(RESyntax.RE_PURE_GROUPING)) {
	      pure = true;
	      index += 2;
	    }
	    break;
	  case '#':
	    if (syntax.get(RESyntax.RE_COMMENTS)) {
	      comment = true;
	    }
	    break;
          default:
            throw new REException(getLocalizedMessage("repeat.no.token"), REException.REG_BADRPT, index);
	  }
	}

	if (index >= pLength) {
	    throw new REException(getLocalizedMessage("unmatched.paren"), REException.REG_ESUBREG,index);
	}

	// find end of subexpression
	int endIndex = index;
	int nextIndex = index;
	int nested = 0;

	while ( ((nextIndex = getCharUnit(pattern,endIndex,unit,false)) > 0)
		&& !(nested == 0 && (unit.ch == ')') && (syntax.get(RESyntax.RE_NO_BK_PARENS) ^ (unit.bk || quot))) ) {
	  if ((endIndex = nextIndex) >= pLength)
	    throw new REException(getLocalizedMessage("subexpr.no.end"),REException.REG_ESUBREG,nextIndex);
	  else if ((unit.ch == '[') && !(unit.bk || quot)) {
	    // I hate to do something similar to the LIST OPERATOR matters
	    // above, but ...
	    int listIndex = nextIndex;
	    if (listIndex < pLength && pattern[listIndex] == '^') listIndex++;
	    if (listIndex < pLength && pattern[listIndex] == ']') listIndex++;
	    int listEndIndex = -1;
	    int listNest = 0;
	    while (listIndex < pLength && listEndIndex < 0) {
	      switch(pattern[listIndex++]) {
		case '\\':
		  listIndex++;
		  break;
		case '[':
		  // Sun's API document says that regexp like "[a-d[m-p]]"
		  // is legal. Even something like "[[[^]]]]" is accepted.
		  listNest++;
		  if (listIndex < pLength && pattern[listIndex] == '^') listIndex++;
		  if (listIndex < pLength && pattern[listIndex] == ']') listIndex++;
		  break;
		case ']':
		  if (listNest == 0)
		    listEndIndex = listIndex;
		  listNest--;
		  break;
	      }
	    }
	    if (listEndIndex >= 0) {
	      nextIndex = listEndIndex;
	      if ((endIndex = nextIndex) >= pLength)
	        throw new REException(getLocalizedMessage("subexpr.no.end"),REException.REG_ESUBREG,nextIndex);
	      else
	        continue;
	    }
	    throw new REException(getLocalizedMessage("subexpr.no.end"),REException.REG_ESUBREG,nextIndex);
	  }
	  else if (unit.ch == '(' && (syntax.get(RESyntax.RE_NO_BK_PARENS) ^ (unit.bk || quot)))
	    nested++;
	  else if (unit.ch == ')' && (syntax.get(RESyntax.RE_NO_BK_PARENS) ^ (unit.bk || quot)))
	    nested--;
	}

	// endIndex is now position at a ')','\)' 
	// nextIndex is end of string or position after ')' or '\)'

	if (comment) index = nextIndex;
	else { // not a comment
	  // create RE subexpression as token.
	  addToken(currentToken);
	  if (!pure) {
	    numSubs++;
	  }

	  int useIndex = (pure || lookAhead || lookBehind || independent) ?
			 0 : nextSub + numSubs;
	  currentToken = new RE(String.valueOf(pattern,index,endIndex-index).toCharArray(),cflags,syntax,useIndex,nextSub + numSubs);
	  numSubs += ((RE) currentToken).getNumSubs();

          if (lookAhead) {
	      currentToken = new RETokenLookAhead(currentToken,negativelh);
	  }
          else if (lookBehind) {
	      currentToken = new RETokenLookBehind(currentToken,negativelb);
	  }
          else if (independent) {
	      currentToken = new RETokenIndependent(currentToken);
	  }

	  index = nextIndex;
	  if (flagsSaved) {
	      syntax = savedSyntax;
	      cflags = savedCflags;
	      insens = ((cflags & REG_ICASE) > 0);
	      flagsSaved = false;
	  }
	} // not a comment
      } // subexpression
    
      // UNMATCHED RIGHT PAREN
      // ) or \) throw exception if
      // !syntax.get(RESyntax.RE_UNMATCHED_RIGHT_PAREN_ORD)
      else if (!syntax.get(RESyntax.RE_UNMATCHED_RIGHT_PAREN_ORD) && ((unit.ch == ')') && (syntax.get(RESyntax.RE_NO_BK_PARENS) ^ (unit.bk || quot)))) {
	throw new REException(getLocalizedMessage("unmatched.paren"),REException.REG_EPAREN,index);
      }

      // START OF LINE OPERATOR
      //  ^

      else if ((unit.ch == '^') && !(unit.bk || quot)) {
	addToken(currentToken);
	currentToken = null;
	addToken(new RETokenStart(subIndex,((cflags & REG_MULTILINE) > 0) ? syntax.getLineSeparator() : null));
      }

      // END OF LINE OPERATOR
      //  $

      else if ((unit.ch == '$') && !(unit.bk || quot)) {
	addToken(currentToken);
	currentToken = null;
	addToken(new RETokenEnd(subIndex,((cflags & REG_MULTILINE) > 0) ? syntax.getLineSeparator() : null));
      }

      // MATCH-ANY-CHARACTER OPERATOR (except possibly newline and null)
      //  .

      else if ((unit.ch == '.') && !(unit.bk || quot)) {
	addToken(currentToken);
	currentToken = new RETokenAny(subIndex,syntax.get(RESyntax.RE_DOT_NEWLINE) || ((cflags & REG_DOT_NEWLINE) > 0),syntax.get(RESyntax.RE_DOT_NOT_NULL));
      }

      // ZERO-OR-MORE REPEAT OPERATOR
      //  *
      //
      // This method used to check "repeat.empty.token" to avoid such regexp
      // as "(a*)*", but now "repeat.empty.token" is allowed.

      else if ((unit.ch == '*') && !(unit.bk || quot)) {
	if (currentToken == null)
          throw new REException(getLocalizedMessage("repeat.no.token"),REException.REG_BADRPT,index);
	if (currentToken instanceof RETokenRepeated)
          throw new REException(getLocalizedMessage("repeat.chained"),REException.REG_BADRPT,index);
	if (currentToken instanceof RETokenWordBoundary || currentToken instanceof RETokenWordBoundary)
	  throw new REException(getLocalizedMessage("repeat.assertion"),REException.REG_BADRPT,index);
	currentToken = setRepeated(currentToken,0,Integer.MAX_VALUE,index);
      }

      // ONE-OR-MORE REPEAT OPERATOR / POSSESSIVE MATCHING OPERATOR
      //  + | \+ depending on RE_BK_PLUS_QM
      //  not available if RE_LIMITED_OPS is set
      //
      // This method used to check "repeat.empty.token" to avoid such regexp
      // as "(a*)+", but now "repeat.empty.token" is allowed.

      else if ((unit.ch == '+') && !syntax.get(RESyntax.RE_LIMITED_OPS) && (!syntax.get(RESyntax.RE_BK_PLUS_QM) ^ (unit.bk || quot))) {
	if (currentToken == null)
          throw new REException(getLocalizedMessage("repeat.no.token"),REException.REG_BADRPT,index);
	
	// Check for possessive matching on RETokenRepeated
	if (currentToken instanceof RETokenRepeated) {
	  RETokenRepeated tokenRep = (RETokenRepeated)currentToken;
	  if (syntax.get(RESyntax.RE_POSSESSIVE_OPS) && !tokenRep.isPossessive() && !tokenRep.isStingy())
	    tokenRep.makePossessive();
	  else
	    throw new REException(getLocalizedMessage("repeat.chained"),REException.REG_BADRPT,index);