private double getScoreBM25(RetrievalModel r) throws IOException {
if (!this.docIteratorHasMatchCache()) {
return 0d;
} else {
Map<Qry, Integer> queryMap = new HashMap<>();
for (Qry arg : this.args) {
int num = queryMap.getOrDefault(arg, 0) + 1;
queryMap.put(arg, num);
}
countingSetup();
double totalScore = 0d;
for (Qry curArg : queryMap.keySet()) {
int curQryFreq = queryMap.get(curArg);
int docid = this.docIteratorGetMatch();
if (!curArg.docIteratorHasMatch(r) || docid != curArg.docIteratorGetMatch()) {
// nothing match , curscore = 0;
continue;
} else {
totalScore += ((QrySop) curArg).getScore(r);
} // end of else
} // end of for loop
return totalScore;
}
}
/**
* Advance the internal document iterator to the specified document, or beyond if it doesn't.
*
* @param docid An internal document id.
*/
public void docIteratorAdvanceTo(int docid) {
for (Qry q_i : this.args) {
q_i.docIteratorAdvanceTo(docid);
}
this.docIteratorClearMatchCache();
}
/**
* An instantiation of docIteratorHasMatch that is true if the query has a document that matches
* the first query argument; some subclasses may choose to use this implementation.
*
* @param r The retrieval model that determines what is a match
* @return True if the query matches, otherwise false.
*/
protected boolean docIteratorHasMatchFirst(RetrievalModel r) {
Qry q_0 = this.args.get(0);
if (q_0.docIteratorHasMatch(r)) {
int docid = q_0.docIteratorGetMatch();
this.docIteratorSetMatchCache(docid);
return true;
} else {
return false;
}
}
/**
* An instantiation of docIteratorHasMatch that is true if the query has a document that matches
* all query arguments; some subclasses may choose to use this implementation.
*
* @param r The retrieval model that determines what is a match
* @return True if the query matches, otherwise false.
*/
protected boolean docIteratorHasMatchAll(RetrievalModel r) {
boolean matchFound = false;
// Keep trying until a match is found or no match is possible.
while (!matchFound) {
// Get the docid of the first query argument.
Qry q_0 = this.args.get(0);
if (!q_0.docIteratorHasMatch(r)) {
return false;
}
int docid_0 = q_0.docIteratorGetMatch();
// Other query arguments must match the docid of the first query
// argument.
matchFound = true;
for (int i = 1; i < this.args.size(); i++) {
Qry q_i = this.args.get(i);
q_i.docIteratorAdvanceTo(docid_0);
if (!q_i.docIteratorHasMatch(r)) { // If any argument is exhausted
return false; // there are no more matches.
}
int docid_i = q_i.docIteratorGetMatch();
if (docid_0 != docid_i) { // docid_0 can't match. Try again.
q_0.docIteratorAdvanceTo(docid_i);
matchFound = false;
break;
}
}
if (matchFound) {
docIteratorSetMatchCache(docid_0);
}
}
return true;
}
/**
* An instantiation of docIteratorHasMatch that is true if the query has a document that matches
* at least one query argument; the match is the smallest docid to match; some subclasses may
* choose to use this implementation.
*
* @param r The retrieval model that determines what is a match
* @return True if the query matches, otherwise false.
*/
protected boolean docIteratorHasMatchMin(RetrievalModel r) {
int minDocid = Qry.INVALID_DOCID;
for (int i = 0; i < this.args.size(); i++) {
Qry q_i = this.args.get(i);
if (q_i.docIteratorHasMatch(r)) {
int q_iDocid = q_i.docIteratorGetMatch();
if ((minDocid > q_iDocid) || (minDocid == Qry.INVALID_DOCID)) {
minDocid = q_iDocid;
}
}
}
if (minDocid != Qry.INVALID_DOCID) {
docIteratorSetMatchCache(minDocid);
return true;
} else {
return false;
}
}
/**
* Initialize the query operator (and its arguments), including any internal iterators. If the
* query operator is of type QryIop, it is fully evaluated, and the results are stored in an
* internal inverted list that may be accessed via the internal iterator.
*
* @param r A retrieval model that guides initialization
* @throws IOException Error accessing the Lucene index.
*/
public void initialize(RetrievalModel r) throws IOException {
for (Qry q_i : this.args) {
q_i.initialize(r);
}
}
/**
* Append an argument to the list of query operator arguments.
*
* @param q The query argument (query operator) to append.
* @throws IllegalArgumentException q is an invalid argument
*/
public void appendArg(Qry q) throws IllegalArgumentException {
// The query parser and query operator type system are too simple
// to detect some kinds of query syntax errors. appendArg does
// additional syntax checking while creating the query tree. It
// also inserts SCORE operators between QrySop operators and QryIop
// arguments, and propagates field information from QryIop
// children to parents. Basically, it creates a well-formed
// query tree.
if (this instanceof QryIopTerm) {
throw new IllegalArgumentException("The TERM operator has no arguments.");
}
// SCORE operators can have only a single argument of type QryIop.
if (this instanceof QrySopScore) {
if (this.args.size() > 0) {
throw new IllegalArgumentException("Score operators can have only one argument");
} else if (!(q instanceof QryIop)) {
throw new IllegalArgumentException(
"The argument to a SCORE operator must be of type QryIop.");
} else {
this.args.add(q);
return;
}
}
// Check whether it is necessary to insert an implied SCORE
// operator between a QrySop operator and a QryIop argument.
if ((this instanceof QrySop) && (q instanceof QryIop)) {
Qry impliedOp = new QrySopScore();
impliedOp.setDisplayName("#SCORE");
impliedOp.appendArg(q);
this.args.add(impliedOp);
return;
}
// QryIop operators must have QryIop arguments in the same field.
if ((this instanceof QryIop) && (q instanceof QryIop)) {
if (this.args.size() == 0) {
((QryIop) this).field = new String(((QryIop) q).getField());
} else {
if (!((QryIop) this).field.equals(((QryIop) q).getField())) {
throw new IllegalArgumentException(
"Arguments to QryIop operators must be in the same field.");
}
}
this.args.add(q);
return;
}
// QrySop operators and their arguments must be of the same type.
if ((this instanceof QrySop) && (q instanceof QrySop)) {
this.args.add(q);
return;
}
throw new IllegalArgumentException(
"Objects of type "
+ q.getClass().getName()
+ " cannot be an argument to a query operator of type "
+ this.getClass().getName());
}
/**
* Initialize the query operator (and its arguments), including any internal iterators. If the
* query operator is of type QryIop, it is fully evaluated, and the results are stored in an
* internal inverted list that may be accessed via the internal iterator.
*
* @param r A retrieval model that guides initialization
* @throws IOException Error accessing the Lucene index.
*/
public void initialize(RetrievalModel r) throws IOException {
Qry q = this.args.get(0);
q.initialize(r);
}
@Override
protected void evaluate() throws IOException {
this.invertedList = new InvList(this.getField());
// no argument
if (args.size() == 0) return;
if (args.size() == 1) {
this.invertedList = ((QryIop) this.args.get(0)).invertedList;
return;
}
// iterate each doc, greedy algorithms
while (true) {
// init doc_id
int doc_id = Qry.INVALID_DOCID;
if (this.docIteratorHasMatchAll(null)) doc_id = this.args.get(0).docIteratorGetMatch();
// fail to find a match all
if (doc_id == Qry.INVALID_DOCID) break;
// the last argument's position in doc
List<Integer> positions = new ArrayList<Integer>();
// iterate each location
while (true) {
boolean success = false;
int position = 0;
// index from 1 to n-1
for (int i = 1; i < this.args.size(); i++) {
// get previous pos
if (!((QryIop) this.args.get(i - 1)).locIteratorHasMatch()) break;
int position1 = ((QryIop) this.args.get(i - 1)).locIteratorGetMatch();
// go beyond pre pos
((QryIop) this.args.get(i)).locIteratorAdvancePast(position1);
if (!((QryIop) this.args.get(i)).locIteratorHasMatch()) break;
int position2 = ((QryIop) this.args.get(i)).locIteratorGetMatch();
// near condition satisfied?
if (position2 > position1 && position2 - position1 <= this.dist) {
position = position2;
if (i == this.args.size() - 1) success = true;
} else {
success = false;
break;
}
}
// find a legal position
if (success) {
positions.add(position);
// move each loc pointer
for (Qry q_i : this.args) {
((QryIop) q_i).locIteratorAdvance();
}
} else {
// move first arg
((QryIop) this.args.get(0)).locIteratorAdvance();
// if no more match position for the first arg
if (!((QryIop) this.args.get(0)).locIteratorHasMatch()) break;
// reset other pointers to beginning
for (int i = 1; i < this.args.size(); i++) {
((QryIop) this.args.get(i)).locIteratorReset();
int preposition = ((QryIop) this.args.get(0)).locIteratorGetMatch();
((QryIop) this.args.get(i)).locIteratorAdvancePast(preposition);
}
}
}
// move doc pointer
for (Qry q_i : this.args) q_i.docIteratorAdvancePast(doc_id);
// success
if (positions.size() > 0) this.invertedList.appendPosting(doc_id, positions);
}
}
