/**
* Returns a pre value.
*
* @param id unique node id
* @return pre value or -1 if id was not found
*/
final int preold(final int id) {
// find pre value in table
for (int p = Math.max(0, id); p < meta.size; ++p) if (id == id(p)) return p;
final int ps = Math.min(meta.size, id);
for (int p = 0; p < ps; ++p) if (id == id(p)) return p;
// id not found
return -1;
}
@Override
public synchronized int costs(final IndexToken it) {
final byte[] tok = it.get();
if (tok.length > data.meta.maxlen) return Integer.MAX_VALUE;
// estimate costs for queries which stretch over multiple index entries
final FTOpt opt = ((FTLexer) it).ftOpt();
if (opt.is(FZ) || opt.is(WC)) return Math.max(1, data.meta.size >> 4);
return entry(tok).size;
}
@Override
public DiskData build() throws IOException {
meta.assign(parser);
meta.dirty = true;
// calculate optimized output buffer sizes to reduce disk fragmentation
final Runtime rt = Runtime.getRuntime();
final long max = Math.min(1 << 22, rt.maxMemory() - rt.freeMemory() >> 2);
int bs = (int) Math.min(meta.filesize, max);
bs = Math.max(IO.BLOCKSIZE, bs - bs % IO.BLOCKSIZE);
// drop old database (if available) and create new one
DropDB.drop(dbname, sopts);
sopts.dbpath(dbname).md();
elemNames = new Names(meta);
attrNames = new Names(meta);
try {
tout = new DataOutput(new TableOutput(meta, DATATBL));
xout = new DataOutput(meta.dbfile(DATATXT), bs);
vout = new DataOutput(meta.dbfile(DATAATV), bs);
sout = new DataOutput(meta.dbfile(DATATMP), bs);
final Performance perf = Prop.debug ? new Performance() : null;
Util.debug(tit() + DOTS);
parse();
if (Prop.debug) Util.errln(" " + perf + " (" + Performance.getMemory() + ')');
} catch (final IOException ex) {
try {
close();
} catch (final IOException ignored) {
}
throw ex;
}
close();
// copy temporary values into database table
try (final DataInput in = new DataInput(meta.dbfile(DATATMP))) {
final TableAccess ta = new TableDiskAccess(meta, true);
for (; spos < ssize; ++spos) ta.write4(in.readNum(), 8, in.readNum());
ta.close();
}
meta.dbfile(DATATMP).delete();
// return database instance
return new DiskData(meta, elemNames, attrNames, path, ns);
}
/**
* Performs a fuzzy search for the specified token with a maximum number of errors.
*
* @param token token to look for
* @param k number of errors allowed
* @return iterator
*/
private synchronized IndexIterator fuzzy(final byte[] token, final int k) {
FTIndexIterator it = FTIndexIterator.FTEMPTY;
final int tokl = token.length, tl = tp.length;
final int e = Math.min(tl - 1, tokl + k);
int s = Math.max(1, tokl - k) - 1;
while (++s <= e) {
int p = tp[s];
if (p == -1) continue;
int t = s + 1, r = -1;
while (t < tl && r == -1) r = tp[t++];
while (p < r) {
if (ls.similar(inY.readBytes(p, s), token, k)) {
it = FTIndexIterator.union(iter(pointer(p, s), size(p, s), inZ, token), it);
}
p += s + ENTRY;
}
}
return it;
}