/**
* 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));
}
}
@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);
}
/**
* Returns the distance of the specified node.
*
* @param pre pre value
* @param kind node kind
* @return distance
*/
public int dist(final int pre, final int kind) {
switch (kind) {
case ELEM:
return table.read4(pre, 4);
case TEXT:
case COMM:
case PI:
return table.read4(pre, 8);
case ATTR:
int d = table.read1(pre, 0) >> 3 & IO.MAXATTS;
// skip additional attributes, if value is larger than maximum range
if (d >= IO.MAXATTS) while (d < pre && kind(pre - d) == ATTR) d++;
return d;
default:
return pre + 1;
}
}
/**
* 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);
}
/**
* Sets the attribute size.
*
* @param pre pre value
* @param kind node kind
* @param value value
*/
private void attSize(final int pre, final int kind, final int value) {
if (kind == ELEM) table.write1(pre, 0, value << 3 | ELEM);
}
/**
* Sets the distance.
*
* @param pre pre value
* @param kind node kind
* @param value value
*/
public void dist(final int pre, final int kind, final int value) {
if (kind == ATTR) table.write1(pre, 0, value << 3 | ATTR);
else if (kind != DOC) table.write4(pre, kind == ELEM ? 4 : 8, value);
}
/**
* Sets the disk offset of a text/attribute value.
*
* @param pre pre value
* @param off offset
*/
final void textOff(final int pre, final long off) {
table.write5(pre, 3, off);
}
/**
* Sets the size value.
*
* @param pre pre reference
* @param kind node kind
* @param value value to be stored
*/
public final void size(final int pre, final int kind, final int value) {
if (kind == ELEM || kind == DOC) table.write4(pre, 8, value);
}
/**
* 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);
}
/**
* Returns the disk offset of a text (text, comment, pi) or attribute value.
*
* @param pre pre value
* @return disk offset
*/
final long textOff(final int pre) {
return table.read5(pre, 3);
}
/**
* Returns a unique node id.
*
* @param pre pre value
* @return node id
*/
public final int id(final int pre) {
return table.read4(pre, 12);
}
/**
* Returns a reference to the namespace of the addressed element or attribute.
*
* @param pre pre value
* @param kind node kind
* @return namespace URI
*/
public final int uri(final int pre, final int kind) {
return kind == ELEM || kind == ATTR ? table.read1(pre, kind == ELEM ? 3 : 11) & 0xFF : 0;
}
/**
* Returns the namespace flag of the addressed element.
*
* @param pre pre value
* @return namespace flag
*/
public final boolean nsFlag(final int pre) {
return (table.read1(pre, 1) & 0x80) != 0;
}
/**
* Returns a reference to the tag or attribute name id.
*
* @param pre pre value
* @return token reference
*/
public final int name(final int pre) {
return table.read2(pre, 1) & 0x7FFF;
}
/**
* Returns a number of attributes.
*
* @param pre pre value
* @param kind node kind
* @return number of attributes
*/
public final int attSize(final int pre, final int kind) {
int s = kind == ELEM ? table.read1(pre, 0) >> 3 & IO.MAXATTS : 1;
// skip additional attributes, if value is larger than maximum range
if (s >= IO.MAXATTS) while (s < meta.size - pre && kind(pre + s) == ATTR) s++;
return s;
}
/**
* Returns a size value (number of descendant table entries).
*
* @param pre pre value
* @param kind node kind
* @return size value
*/
public final int size(final int pre, final int kind) {
return kind == ELEM || kind == DOC ? table.read4(pre, 8) : 1;
}
/**
* Sets the namespace flag. Should be only called for element nodes.
*
* @param pre pre value
* @param ne namespace flag
*/
public final void nsFlag(final int pre, final boolean ne) {
table.write1(pre, 1, table.read1(pre, 1) & 0x7F | (ne ? 0x80 : 0));
}
/**
* 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);
}
}
/**
* Inserts the internal buffer to the storage without updating the table structure.
*
* @param pre insert position
*/
public final void insert(final int pre) {
table.insert(pre, buffer());
}
/**
* Returns the node kind, which may be {@link #DOC}, {@link #ELEM}, {@link #TEXT}, {@link #ATTR},
* {@link #COMM} or {@link #PI}.
*
* @param pre pre value
* @return node kind
*/
public final int kind(final int pre) {
return table.read1(pre, 0) & 0x07;
}
