/**
* 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 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);
}
/**
* Adds an attribute entry to the internal update buffer.
*
* @param pre pre value
* @param dist parent distance
* @param name attribute name
* @param value attribute value
* @param uri namespace uri reference
* @param ne namespace flag
*/
public final void attr(
final int pre,
final int dist,
final int name,
final byte[] value,
final int uri,
final boolean ne) {
// add attribute to text storage
final int i = newID();
final long v = index(pre, i, value, ATTR);
final int n = ne ? 1 << 7 : 0;
s(Math.min(IO.MAXATTS, dist) << 3 | ATTR);
s(n | (byte) (name >> 8));
s(name);
s(v >> 32);
s(v >> 24);
s(v >> 16);
s(v >> 8);
s(v);
s(0);
s(0);
s(0);
s(uri);
s(i >> 24);
s(i >> 16);
s(i >> 8);
s(i);
}
/**
* Adds an element entry to the internal update buffer.
*
* @param dist parent distance
* @param name tag name index
* @param asize number of attributes
* @param size node size
* @param uri namespace uri reference
* @param ne namespace flag
*/
public final void elem(
final int dist,
final int name,
final int asize,
final int size,
final int uri,
final boolean ne) {
// build and insert new entry
final int i = newID();
final int n = ne ? 1 << 7 : 0;
s(Math.min(IO.MAXATTS, asize) << 3 | ELEM);
s(n | (byte) (name >> 8));
s(name);
s(uri);
s(dist >> 24);
s(dist >> 16);
s(dist >> 8);
s(dist);
s(size >> 24);
s(size >> 16);
s(size >> 8);
s(size);
s(i >> 24);
s(i >> 16);
s(i >> 8);
s(i);
}
/**
* Performs a wildcard search for the specified token.
*
* @param token token to look for
* @return iterator
*/
private synchronized IndexIterator wc(final byte[] token) {
final FTIndexIterator it = FTIndexIterator.FTEMPTY;
final FTWildcard wc = new FTWildcard(token);
if (!wc.parse()) return it;
final IntList pr = new IntList();
final IntList ps = new IntList();
final byte[] pref = wc.prefix();
final int pl = pref.length, tl = tp.length;
final int l = Math.min(tl - 1, wc.max());
for (int ti = pl; ti <= l; ti++) {
int i = tp[ti];
if (i == -1) continue;
int c = ti + 1;
int e = -1;
while (c < tl && e == -1) e = tp[c++];
i = find(pref, i, e, ti);
while (i < e) {
final byte[] t = inY.readBytes(i, ti);
if (!startsWith(t, pref)) break;
if (wc.match(t)) {
inZ.cursor(pointer(i, ti));
final int s = size(i, ti);
for (int d = 0; d < s; d++) {
pr.add(inZ.readNum());
ps.add(inZ.readNum());
}
}
i += ti + ENTRY;
}
}
return iter(new FTCache(pr, ps), token);
}
/**
* 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;
}
@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;
}
/**
* Inserts a data instance at the specified pre value. Note that the specified data instance must
* differ from this instance.
*
* @param ipre value at which to insert new data
* @param ipar parent pre value of node
* @param clip data clip
*/
public final void insert(final int ipre, final int ipar, final DataClip clip) {
meta.update();
// update value and document indexes
if (meta.updindex) indexBegin();
resources.insert(ipre, clip);
final int dsize = clip.size();
final int buf = Math.min(dsize, IO.BLOCKSIZE >> IO.NODEPOWER);
// resize buffer to cache more entries
buffer(buf);
// find all namespaces in scope to avoid duplicate declarations
final TokenMap nsScope = nspaces.scope(ipar, this);
// loop through all entries
final IntList preStack = new IntList();
final NSNode nsRoot = nspaces.current();
final HashSet<NSNode> newNodes = new HashSet<NSNode>();
final IntList flagPres = new IntList();
// indicates if database only contains a dummy node
final Data data = clip.data;
int c = 0;
for (int dpre = clip.start; dpre < clip.end; ++dpre, ++c) {
if (c != 0 && c % buf == 0) insert(ipre + c - buf);
final int pre = ipre + c;
final int dkind = data.kind(dpre);
final int dpar = data.parent(dpre, dkind);
// ipar < 0 if document nodes on top level are added
final int dis = dpar >= 0 ? dpre - dpar : ipar >= 0 ? pre - ipar : 0;
final int par = dis == 0 ? -1 : pre - dis;
if (c == 0) nspaces.root(par, this);
while (!preStack.isEmpty() && preStack.peek() > par) nspaces.close(preStack.pop());
switch (dkind) {
case DOC:
// add document
nspaces.prepare();
final int s = data.size(dpre, dkind);
doc(pre, s, data.text(dpre, true));
meta.ndocs++;
preStack.push(pre);
break;
case ELEM:
// add element
nspaces.prepare();
boolean ne = false;
if (data.nsFlag(dpre)) {
final Atts at = data.ns(dpre);
for (int a = 0; a < at.size(); ++a) {
// see if prefix has been declared/ is part of current ns scope
final byte[] old = nsScope.get(at.name(a));
if (old == null || !eq(old, at.value(a))) {
// we have to keep track of all new NSNodes that are added
// to the Namespace structure, as their pre values must not
// be updated. I.e. if an NSNode N with pre value 3 existed
// prior to inserting and two new nodes are inserted at
// location pre == 3 we have to make sure N and only N gets
// updated.
newNodes.add(nspaces.add(at.name(a), at.value(a), pre));
ne = true;
}
}
}
byte[] nm = data.name(dpre, dkind);
elem(
dis,
tagindex.index(nm, null, false),
data.attSize(dpre, dkind),
data.size(dpre, dkind),
nspaces.uri(nm, true),
ne);
preStack.push(pre);
break;
case TEXT:
case COMM:
case PI:
// add text
text(pre, dis, data.text(dpre, true), dkind);
break;
case ATTR:
// add attribute
nm = data.name(dpre, dkind);
// check if prefix already in nsScope or not
final byte[] attPref = prefix(nm);
// check if prefix of attribute has already been declared, otherwise
// add declaration to parent node
if (data.nsFlag(dpre) && nsScope.get(attPref) == null) {
nspaces.add(
par,
preStack.isEmpty() ? -1 : preStack.peek(),
attPref,
data.nspaces.uri(data.uri(dpre, dkind)),
this);
// save pre value to set ns flag later for this node. can't be done
// here as direct table access would interfere with the buffer
flagPres.add(par);
}
attr(
pre,
dis,
atnindex.index(nm, null, false),
data.text(dpre, false),
nspaces.uri(nm, false),
false);
break;
}
}
// finalize and update namespace structure
while (!preStack.isEmpty()) nspaces.close(preStack.pop());
nspaces.root(nsRoot);
if (bp != 0) insert(ipre + c - 1 - (c - 1) % buf);
// reset buffer to old size
buffer(1);
// set ns flags
for (int f = 0; f < flagPres.size(); f++) {
final int fl = flagPres.get(f);
table.write2(fl, 1, name(fl) | 1 << 15);
}
// increase size of ancestors
int p = ipar;
while (p >= 0) {
final int k = kind(p);
size(p, k, size(p, k) + dsize);
p = parent(p, k);
}
if (meta.updindex) {
// add the entries to the ID -> PRE mapping:
idmap.insert(ipre, id(ipre), dsize);
indexEnd();
}
if (!cache) updateDist(ipre + dsize, dsize);
// propagate PRE value shifts to namespaces
if (ipar != -1) nspaces.insert(ipre, dsize, newNodes);
}
