/**
* Default constructor, called from {@link Open#open}.
*
* @param meta meta data
* @throws IOException I/O Exception
*/
public DiskData(final MetaData meta) throws IOException {
super(meta);
try (final DataInput in = new DataInput(meta.dbfile(DATAINF))) {
meta.read(in);
while (true) {
final String k = string(in.readToken());
if (k.isEmpty()) break;
if (k.equals(DBTAGS)) elemNames = new Names(in, meta);
else if (k.equals(DBATTS)) attrNames = new Names(in, meta);
else if (k.equals(DBPATH)) paths = new PathSummary(this, in);
else if (k.equals(DBNS)) nspaces = new Namespaces(in);
else if (k.equals(DBDOCS)) resources.read(in);
}
}
// open data and indexes
init();
if (meta.updindex) {
idmap = new IdPreMap(meta.dbfile(DATAIDP));
if (meta.textindex) textIndex = new UpdatableDiskValues(this, true);
if (meta.attrindex) attrIndex = new UpdatableDiskValues(this, false);
} else {
if (meta.textindex) textIndex = new DiskValues(this, true);
if (meta.attrindex) attrIndex = new DiskValues(this, false);
}
if (meta.ftxtindex) ftxtIndex = new FTIndex(this);
}
/**
* Creates new node ids and recreates updatable index structures.
*
* @param data data
* @throws IOException I/O Exception during index rebuild
*/
public static void ids(final Data data) throws IOException {
final MetaData md = data.meta;
final int size = md.size;
for (int pre = 0; pre < size; ++pre) data.id(pre, pre);
md.lastid = size - 1;
md.dirty = true;
if (data.meta.updindex) {
data.idmap = new IdPreMap(md.lastid);
if (data.meta.textindex) optimize(IndexType.TEXT, data, true, true, true, null);
if (data.meta.attrindex) optimize(IndexType.ATTRIBUTE, data, true, true, true, null);
}
}
/**
* Updates (renames) an element, attribute or pi name.
*
* @param pre pre value
* @param kind node kind
* @param name new tag, attribute or pi name
* @param uri uri
*/
public final void update(final int pre, final int kind, final byte[] name, final byte[] uri) {
meta.update();
if (kind == PI) {
updateText(pre, trim(concat(name, SPACE, atom(pre))), kind);
} else {
// update/set namespace reference
final int ouri = nspaces.uri(name, pre, this);
final boolean ne = ouri == 0 && uri.length != 0;
final int npre = kind == ATTR ? parent(pre, kind) : pre;
final int nuri =
ne
? nspaces.add(npre, npre, prefix(name), uri, this)
: ouri != 0 && eq(nspaces.uri(ouri), uri) ? ouri : 0;
// write namespace uri reference
table.write1(pre, kind == ELEM ? 3 : 11, nuri);
// write name reference
table.write2(
pre,
1,
(nsFlag(pre) ? 1 << 15 : 0)
| (kind == ELEM ? tagindex : atnindex).index(name, null, false));
// write namespace flag
table.write2(npre, 1, (ne || nsFlag(npre) ? 1 << 15 : 0) | name(npre));
}
}
/**
* Updates (replaces) the value of a single document, text, comment, pi or attribute node.
*
* @param pre pre value to be replaced
* @param kind node kind
* @param value value to be updated (tag name, text, comment, pi)
*/
public final void update(final int pre, final int kind, final byte[] value) {
final byte[] v = kind == PI ? trim(concat(name(pre, kind), SPACE, value)) : value;
if (eq(v, text(pre, kind != ATTR))) return;
meta.update();
updateText(pre, v, kind);
if (kind == DOC) resources.rename(pre, value);
}
/**
* Deletes a node and its descendants.
*
* @param pre pre value of the node to delete
*/
public final void delete(final int pre) {
meta.update();
// size of the subtree to delete
int k = kind(pre);
final int s = size(pre, k);
resources.delete(pre, s);
if (meta.updindex) {
// delete child records from indexes
indexDelete(pre, s);
}
/// explicitly delete text or attribute value
if (k != DOC && k != ELEM) delete(pre, k != ATTR);
// reduce size of ancestors
int par = pre;
// check if we are an attribute (different size counters)
if (k == ATTR) {
par = parent(par, ATTR);
attSize(par, ELEM, attSize(par, ELEM) - 1);
size(par, ELEM, size(par, ELEM) - 1);
k = kind(par);
}
// reduce size of ancestors
while (par > 0 && k != DOC) {
par = parent(par, k);
k = kind(par);
size(par, k, size(par, k) - s);
}
// preserve empty root node
if (kind(pre) == DOC) --meta.ndocs;
if (meta.updindex) {
// delete node and descendants from ID -> PRE map:
idmap.delete(pre, id(pre), -s);
}
// delete node from table structure and reduce document size
table.delete(pre, s);
if (!cache) updateDist(pre, -s);
// propagate PRE value shifts to namespaces
nspaces.delete(pre, s, this);
}
@Override
protected boolean run() {
final Data data = context.data();
final MetaData meta = data.meta;
size = meta.size;
if (!startUpdate()) return false;
boolean ok = true;
try {
// reassign autooptimize flag
final boolean autoopt = options.get(MainOptions.AUTOOPTIMIZE);
if (autoopt != data.meta.autoopt) {
data.meta.autoopt = autoopt;
data.meta.dirty = true;
}
optimize(data, this);
ok = info(DB_OPTIMIZED_X, meta.name, perf);
} catch (final IOException ex) {
ok = error(Util.message(ex));
} finally {
ok &= finishUpdate();
}
return ok;
}
/**
* 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);
}
/**
* Replaces parts of the database with the specified data instance.
*
* @param rpre pre value to be replaced
* @param clip data clip
*/
public final void replace(final int rpre, final DataClip clip) {
meta.update();
final int dsize = clip.size();
final Data data = clip.data;
final int rkind = kind(rpre);
final int rsize = size(rpre, rkind);
final int rpar = parent(rpre, rkind);
final int diff = dsize - rsize;
buffer(dsize);
resources.replace(rpre, rsize, clip);
if (meta.updindex) {
// update index
indexDelete(rpre, rsize);
indexBegin();
}
for (int dpre = clip.start; dpre < clip.end; ++dpre) {
final int dkind = data.kind(dpre);
final int dpar = data.parent(dpre, dkind);
final int pre = rpre + dpre - clip.start;
final int dis = dpar >= 0 ? dpre - dpar : pre - rpar;
switch (dkind) {
case DOC:
// add document
doc(pre, data.size(dpre, dkind), data.text(dpre, true));
meta.ndocs++;
break;
case ELEM:
// add element
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),
false);
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);
attr(
pre,
dis,
atnindex.index(nm, null, false),
data.text(dpre, false),
nspaces.uri(nm, false),
false);
break;
}
}
if (meta.updindex) {
indexEnd();
// update ID -> PRE map:
idmap.delete(rpre, id(rpre), -rsize);
idmap.insert(rpre, meta.lastid - dsize + 1, dsize);
}
// update table:
table.replace(rpre, buffer(), rsize);
buffer(1);
// no distance/size update if the two subtrees are of equal size
if (diff == 0) return;
// increase/decrease size of ancestors, adjust distances of siblings
int p = rpar;
while (p >= 0) {
final int k = kind(p);
size(p, k, size(p, k) + diff);
p = parent(p, k);
}
if (!cache) updateDist(rpre + dsize, diff);
// adjust attribute size of parent if attributes inserted. attribute size
// of parent cannot be reduced via a replace expression.
int dpre = clip.start;
if (data.kind(dpre) == ATTR) {
int d = 0;
while (dpre < clip.end && data.kind(dpre++) == ATTR) d++;
if (d > 1) attSize(rpar, kind(rpar), d + 1);
}
}
/**
* Optimizes the structures of a database.
*
* @param data data
* @param enforceText enforce creation or deletion of text index
* @param enforceAttr enforce creation or deletion of attribute index
* @param enforceToken enforce creation or deletion of token index
* @param enforceFt enforce creation or deletion of full-text index
* @param cmd calling command instance (may be {@code null})
* @throws IOException I/O Exception during index rebuild
*/
public static void optimize(
final Data data,
final boolean enforceText,
final boolean enforceAttr,
final boolean enforceToken,
final boolean enforceFt,
final Optimize cmd)
throws IOException {
// initialize structural indexes
final MetaData md = data.meta;
if (!md.uptodate) {
data.paths.init();
data.elemNames.init();
data.attrNames.init();
md.dirty = true;
final IntList pars = new IntList(), elms = new IntList();
int n = 0;
for (int pre = 0; pre < md.size; ++pre) {
final byte kind = (byte) data.kind(pre);
final int par = data.parent(pre, kind);
while (!pars.isEmpty() && pars.peek() > par) {
pars.pop();
elms.pop();
}
final int level = pars.size();
if (kind == Data.DOC) {
data.paths.put(0, Data.DOC, level);
pars.push(pre);
elms.push(0);
++n;
} else if (kind == Data.ELEM) {
final int id = data.nameId(pre);
data.elemNames.index(data.elemNames.key(id), null, true);
data.paths.put(id, Data.ELEM, level);
pars.push(pre);
elms.push(id);
} else if (kind == Data.ATTR) {
final int id = data.nameId(pre);
final byte[] val = data.text(pre, false);
data.attrNames.index(data.attrNames.key(id), val, true);
data.paths.put(id, Data.ATTR, level, val, md);
} else {
final byte[] val = data.text(pre, true);
if (kind == Data.TEXT && level > 1) data.elemNames.index(elms.peek(), val);
data.paths.put(0, kind, level, val, md);
}
if (cmd != null) cmd.pre = pre;
}
md.ndocs = n;
md.uptodate = true;
}
// rebuild value indexes
optimize(IndexType.TEXT, data, md.createtext, md.textindex, enforceText, cmd);
optimize(IndexType.ATTRIBUTE, data, md.createattr, md.attrindex, enforceAttr, cmd);
optimize(IndexType.TOKEN, data, md.createtoken, md.tokenindex, enforceToken, cmd);
optimize(IndexType.FULLTEXT, data, md.createft, md.ftindex, enforceFt, cmd);
}