public override bool incrementToken()
{
if (!input.incrementToken())
{
return(false);
}
char[] termBuffer = termAtt.termBuffer();
int len = termAtt.termLength();
// if protected, don't stem. use this to avoid stemming collisions.
if (protWords != null && protWords.contains(termBuffer, 0, len))
{
return(true);
}
stemmer.setCurrent(new string(termBuffer, 0, len));//ugh, wish the Stemmer took a char array
stemmer.stem();
string newstr = stemmer.getCurrent();
termAtt.setTermBuffer(newstr.ToCharArray(), 0, newstr.Length);
return(true);
}
#pragma warning disable 672
public override Token next(Token @in)
{
// check the queue first
if (queuePos < queue.size())
{
return((Token)queue.get(queuePos++));
}
// reset the queue if it had been previously used
if (queuePos != 0)
{
queuePos = 0;
queue.clear();
}
// optimize for the common case: assume there will be
// no subwords (just a simple word)
//
// Would it actually be faster to check for the common form
// of isLetter() isLower()*, and then backtrack if it doesn't match?
int origPosIncrement = 0;
Token t;
while (true)
{
// t is either returned, or a new token is made from it, so it should
// be safe to use the next(Token) method.
#pragma warning disable 612
t = input.next(@in);
#pragma warning restore 612
if (t == null)
{
return(null);
}
char [] termBuffer = t.termBuffer();
int len = t.termLength();
int start = 0;
if (len == 0)
{
continue;
}
int posInc = t.getPositionIncrement();
origPosIncrement += posInc;
//skip protected tokens
if (protWords != null && protWords.contains(termBuffer, 0, len))
{
t.setPositionIncrement(origPosIncrement);
return(t);
}
// Avoid calling charType more than once for each char (basically
// avoid any backtracking).
// makes code slightly more difficult, but faster.
int ch = termBuffer[start];
int type = charType(ch);
int numWords = 0;
while (start < len)
{
// first eat delimiters at the start of this subword
while ((type & SUBWORD_DELIM) != 0 && ++start < len)
{
ch = termBuffer[start];
type = charType(ch);
}
int pos = start;
// save the type of the first char of the subword
// as a way to tell what type of subword token this is (number, word, etc)
int firstType = type;
int lastType = type; // type of the previously read char
while (pos < len)
{
if ((type & lastType) == 0) // no overlap in character type
// check and remove "'s" from the end of a token.
// the pattern to check for is
// ALPHA "'" ("s"|"S") (SUBWORD_DELIM | END)
{
if (stemEnglishPossessive != 0 && ((lastType & ALPHA) != 0))
{
if (ch == '\'' && pos + 1 < len &&
(termBuffer[pos + 1] == 's' || termBuffer[pos + 1] == 'S'))
{
int subWordEnd = pos;
if (pos + 2 >= len)
{
// end of string detected after "'s"
pos += 2;
}
else
{
// make sure that a delimiter follows "'s"
int ch2 = termBuffer[pos + 2];
int type2 = charType(ch2);
if ((type2 & SUBWORD_DELIM) != 0)
{
// if delimiter, move position pointer
// to it (skipping over "'s"
ch = ch2;
type = type2;
pos += 2;
}
}
queue.add(newTok(t, start, subWordEnd));
if ((firstType & ALPHA) != 0)
{
numWords++;
}
break;
}
}
// For case changes, only split on a transition from
// lower to upper case, not vice-versa.
// That will correctly handle the
// case of a word starting with a capital (won't split).
// It will also handle pluralization of
// an uppercase word such as FOOs (won't split).
if (splitOnCaseChange == 0 &&
(lastType & ALPHA) != 0 && (type & ALPHA) != 0)
{
// ALPHA->ALPHA: always ignore if case isn't considered.
}
else if ((lastType & UPPER) != 0 && (type & ALPHA) != 0)
{
// UPPER->letter: Don't split
}
else if (splitOnNumerics == 0 &&
(((lastType & ALPHA) != 0 && (type & DIGIT) != 0) || ((lastType & DIGIT) != 0 && (type & ALPHA) != 0)))
{
// ALPHA->NUMERIC, NUMERIC->ALPHA :Don't split
}
else
{
// NOTE: this code currently assumes that only one flag
// is set for each character now, so we don't have
// to explicitly check for all the classes of transitions
// listed below.
// LOWER->UPPER
// ALPHA->NUMERIC
// NUMERIC->ALPHA
// *->DELIMITER
queue.add(newTok(t, start, pos));
if ((firstType & ALPHA) != 0)
{
numWords++;
}
break;
}
}
if (++pos >= len)
{
if (start == 0)
{
// the subword is the whole original token, so
// return it unchanged.
t.setPositionIncrement(origPosIncrement);
return(t);
}
// optimization... if this is the only token,
// return it immediately.
if (queue.size() == 0 && preserveOriginal == 0)
{
// just adjust the text w/o changing the rest
// of the original token
t.setTermBuffer(termBuffer, start, len - start);
t.setStartOffset(t.startOffset() + start);
t.setPositionIncrement(origPosIncrement);
return(t);
}
Token newtok = newTok(t, start, pos);
queue.add(newtok);
if ((firstType & ALPHA) != 0)
{
numWords++;
}
break;
}
lastType = type;
ch = termBuffer[pos];
type = charType(ch);
}
// start of the next subword is the current position
start = pos;
}
// System.out.println("##########TOKEN=" + s + " ######### WORD DELIMITER QUEUE=" + str(queue));
int numtok = queue.size();
// We reached the end of the current token.
// If the queue is empty, we should continue by reading
// the next token
if (numtok == 0)
{
// the token might have been all delimiters, in which
// case return it if we're meant to preserve it
if (preserveOriginal != 0)
{
return(t);
}
// if this token had a "normal" gap of 1, remove it.
if (posInc == 1)
{
origPosIncrement -= 1;
}
continue;
}
// if number of tokens is 1, there are no catenations to be done.
if (numtok == 1)
{
break;
}
int numNumbers = numtok - numWords;
// check conditions under which the current token
// queue may be used as-is (no catenations needed)
if (catenateAll == 0 && // no "everything" to catenate
(catenateWords == 0 || numWords <= 1) && // no words to catenate
(catenateNumbers == 0 || numNumbers <= 1) && // no numbers to catenate
(generateWordParts != 0 || numWords == 0) && // word generation is on
(generateNumberParts != 0 || numNumbers == 0)) // number generation is on
{
break;
}
// swap queue and the temporary working list, then clear the
// queue in preparation for adding all combinations back to it.
ArrayList /*<Token>*/ tmp = tlist;
tlist = queue;
queue = tmp;
queue.clear();
if (numWords == 0)
{
// all numbers
addCombos(tlist, 0, numtok, generateNumberParts != 0, catenateNumbers != 0 || catenateAll != 0, 1);
if (queue.size() > 0 || preserveOriginal != 0)
{
break;
}
else
{
continue;
}
}
else if (numNumbers == 0)
{
// all words
addCombos(tlist, 0, numtok, generateWordParts != 0, catenateWords != 0 || catenateAll != 0, 1);
if (queue.size() > 0 || preserveOriginal != 0)
{
break;
}
else
{
continue;
}
}
else if (generateNumberParts == 0 && generateWordParts == 0 && catenateNumbers == 0 && catenateWords == 0)
{
// catenate all *only*
// OPT:could be optimized to add to current queue...
addCombos(tlist, 0, numtok, false, catenateAll != 0, 1);
if (queue.size() > 0 || preserveOriginal != 0)
{
break;
}
else
{
continue;
}
}
//
// Find all adjacent tokens of the same type.
//
Token tok = (Token)tlist.get(0);
bool isWord = (tokType(tok) & ALPHA) != 0;
bool wasWord = isWord;
for (int i = 0; i < numtok;)
{
int j;
for (j = i + 1; j < numtok; j++)
{
wasWord = isWord;
tok = (Token)tlist.get(j);
isWord = (tokType(tok) & ALPHA) != 0;
if (isWord != wasWord)
{
break;
}
}
if (wasWord)
{
addCombos(tlist, i, j, generateWordParts != 0, catenateWords != 0, 1);
}
else
{
addCombos(tlist, i, j, generateNumberParts != 0, catenateNumbers != 0, 1);
}
i = j;
}
// take care catenating all subwords
if (catenateAll != 0)
{
addCombos(tlist, 0, numtok, false, true, 0);
}
// NOTE: in certain cases, queue may be empty (for instance, if catenate
// and generate are both set to false). Only exit the loop if the queue
// is not empty.
if (queue.size() > 0 || preserveOriginal != 0)
{
break;
}
}
// System.out.println("##########AFTER COMBINATIONS:"+ str(queue));
if (preserveOriginal != 0)
{
queuePos = 0;
if (queue.size() > 0)
{
// overlap first token with the original
((Token)queue.get(0)).setPositionIncrement(0);
}
return(t); // return the original
}
else
{
queuePos = 1;
Token tok = (Token)queue.get(0);
tok.setPositionIncrement(origPosIncrement);
return(tok);
}
}
