@Override
public boolean indexAccessible(final IndexInfo ii) {
/* If the following conditions yield true, the index is accessed:
* - all query terms are statically available
* - no FTTimes option is specified
* - explicitly set case, diacritics and stemming match options do not
* conflict with index options. */
data = ii.ic.data;
final MetaData md = data.meta;
final FTOpt fto = ftt.opt;
/* Index will be applied if no explicit match options have been set
* that conflict with the index options. As a consequence, though, index-
* based querying might yield other results than sequential scanning. */
if (occ != null
|| fto.cs != null && md.casesens == (fto.cs == FTCase.INSENSITIVE)
|| fto.isSet(DC) && md.diacritics != fto.is(DC)
|| fto.isSet(ST) && md.stemming != fto.is(ST)
|| fto.ln != null && !fto.ln.equals(md.language)) return false;
// adopt database options to tokenizer
fto.copy(md);
// estimate costs if text is not known at compile time
if (tokens == null) {
ii.costs = Math.max(2, data.meta.size / 30);
return true;
}
// summarize number of hits; break loop if no hits are expected
final FTLexer ft = new FTLexer(fto);
ii.costs = 0;
for (byte[] t : tokens) {
ft.init(t);
while (ft.hasNext()) {
final byte[] tok = ft.nextToken();
if (fto.sw != null && fto.sw.contains(tok)) continue;
if (fto.is(WC)) {
// don't use index if one of the terms starts with a wildcard
t = ft.get();
if (t[0] == '.') return false;
// don't use index if certain characters or more than 1 dot are found
int d = 0;
for (final byte w : t) {
if (w == '{' || w == '\\' || w == '.' && ++d > 1) return false;
}
}
// favor full-text index requests over exact queries
final int costs = data.costs(ft);
if (costs != 0) ii.costs += Math.max(2, costs / 100);
}
}
return true;
}
@Override
public synchronized IndexIterator iter(final IndexToken it) {
final byte[] tok = it.get();
// wildcard search
final FTLexer lexer = (FTLexer) it;
final FTOpt opt = lexer.ftOpt();
if (opt.is(WC)) return wc(tok);
// fuzzy search
if (opt.is(FZ)) return fuzzy(tok, lexer.lserror(tok));
// return cached or new result
final IndexEntry e = entry(tok);
return e.size > 0 ? iter(e.offset, e.size, inZ, tok) : FTIndexIterator.FTEMPTY;
}
/**
* Caches and returns all unique tokens specified in a query.
*
* @param list token list
* @return token set
*/
private TokenSet unique(final TokenList list) {
// cache all query tokens in a set (duplicates are removed)
final TokenSet ts = new TokenSet();
switch (mode) {
case ALL:
case ANY:
for (final byte[] t : list) ts.add(t);
break;
case ALL_WORDS:
case ANY_WORD:
final FTLexer l = new FTLexer(ftt.opt);
for (final byte[] t : list) {
l.init(t);
while (l.hasNext()) ts.add(l.nextToken());
}
break;
case PHRASE:
final TokenBuilder tb = new TokenBuilder();
for (final byte[] t : list) tb.add(t).add(' ');
ts.add(tb.trim().finish());
}
return ts;
}
/**
* Returns a scan-based index iterator.
*
* @param lex lexer, including the queried value
* @return node iterator
* @throws QueryException query exception
*/
private FTIndexIterator scan(final FTLexer lex) throws QueryException {
final FTLexer input = new FTLexer(ftt.opt);
final FTTokens fttokens = ftt.cache(lex.get());
return new FTIndexIterator() {
final int sz = data.meta.size;
int pre = -1, ps;
@Override
public int pre() {
return pre;
}
@Override
public boolean more() {
while (++pre < sz) {
if (data.kind(pre) != Data.TEXT) continue;
input.init(data.text(pre, true));
matches.reset(ps);
try {
if (ftt.contains(fttokens, input) != 0) return true;
} catch (final QueryException ignore) {
// ignore exceptions
}
}
return false;
}
@Override
public FTMatches matches() {
return matches;
}
@Override
public void pos(final int p) {
ps = p;
}
@Override
public int size() {
// worst case
return Math.max(1, sz >>> 1);
}
};
}