C++ Words::getWordLen方法代码示例

//.........这里部分代码省略.........
		for ( int32_t i = 0 ; i < oldn && n + 3 < MAX_TIT_CANDIDATES ; i++ ) {
			// stop if no root title segments
			if ( nr <= 0 ) break;
			// get the word info
			Words *w = cptrs[i];
			int32_t   a = as[i];
			int32_t   b = bs[i];
			// init
			int32_t lasta = a;
			char prev  = false;
			// char length in bytes
			//int32_t charlen = 1;
			// see how many we add
			int32_t added = 0;
			char *skipTo = NULL;
			bool qualified = true;
			// . scan the words looking for a token
			// . sometimes the candidates end in ": " so put in "k < b-1"
			// . made this from k<b-1 to k<b to fix
			//   "Hot Tub Time Machine (2010) - IMDb" to strip IMDb
			for ( int32_t k = a ; k < b && n + 3 < MAX_TIT_CANDIDATES; k++){
				// get word
				char *wp = w->getWord(k);
				// skip if not alnum
				if ( ! w->isAlnum(k) ) {
					// in order for next alnum word to
					// qualify for "clipping" if it matches
					// the root title, there has to be more
					// than just spaces here, some punct.
					// otherwise title
					// "T. D. Jakes: Biography from Answers.com"
					// becomes
					// "T. D. Jakes: Biography from"
					qualified=isWordQualified(wp,w->getWordLen(k));
					continue;
				}
				// gotta be qualified!
				if ( ! qualified ) continue;
				// skip if in root title
				if ( skipTo && wp < skipTo ) continue;
				// does this match any root page title segments?
				int32_t j;
				for ( j = 0 ; j < nr ; j++ ) {
					// . compare to root title
					// . break out if we matched!
					if ( ! strncmp( wp, rootTitles[j], rootTitleLens[j] ) ) {
						break;
					}
				}

				// if we did not match a root title segment,
				// keep on chugging
				if ( j >= nr ) continue;
				// . we got a root title match!
				// . skip over
				skipTo = wp + rootTitleLens[j];
				// must land on qualified punct then!!
				int32_t e = k+1;
				for ( ; e<b && w->getWord(e)<skipTo ; e++ );
				// ok, word #e must be a qualified punct
				if ( e<b &&
				     ! isWordQualified(w->getWord(e),w->getWordLen(e)))
					// assume no match then!!
					continue;
				// if we had a previous guy, reset the end of the
				// previous candidate

//.........这里部分代码省略.........

	// . the match at the center of the window is match #"mm", so that
	//   matches->m_matches[mm] is the Match class
	// . set "mi" to it and back up "mi" as int32_t as >= a
	for ( mi = mm ; mi > 0 && ms[mi-1].m_wordNum >=a ; mi-- )
		;

	// now get the score of this excerpt. Also mark all the represented 
	// query words. Mark the represented query words in the array that
	// comes to us. also mark how many times the same word is repeated in
	// this summary.
	int64_t score = 0LL;

	// is a url contained in the summary, that looks bad! punish!
	bool hasUrl = false;

	// the word count we did above was just an approximate. count it right
	wordCount = 0;

	// for debug
	//char buf[5000];
	//char *xp = buf;
	SafeBuf xp;

	// wtf?
	if ( b > nw ) {
		b = nw;
	}

	// first score from the starting match down to a, including match
	for ( int32_t i = a ; i < b ; i++ ) {
		// debug print out
		if ( g_conf.m_logDebugSummary ) {
			int32_t len = words->getWordLen(i);
			char cs;
			for (int32_t k=0;k<len; k+=cs ) {
				const char *c = words->getWord(i)+k;
				cs = getUtf8CharSize(c);
				if ( is_binary_utf8 ( c ) ) {
					continue;
				}
				xp.safeMemcpy ( c , cs );
				xp.nullTerm();
			}
		}

		// skip if in bad section, marquee, select, script, style
		if ( sp && (sp[i]->m_flags & badFlags) ) {
			continue;
		}

		// don't count just numeric words
		if ( words->isNum(i) ) {
			continue;
		}

		// check if there is a url. best way to check for '://'
		if ( wids && !wids[i] ) {
			const char *wrd = words->getWord(i);
			int32_t  wrdLen = words->getWordLen(i);
			if ( wrdLen == 3 && wrd[0] == ':' && wrd[1] == '/' &&  wrd[2] == '/' ) {
				hasUrl = true;
			}
		}

		// skip if not wid