org.exist.indexing.ngram.NGramIndexWorker Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of exist-index-ngram Show documentation
Show all versions of exist-index-ngram Show documentation
eXist-db NoSQL Database n-gram Index
/*
* eXist-db Open Source Native XML Database
* Copyright (C) 2001 The eXist-db Authors
*
* [email protected]
* http://www.exist-db.org
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
package org.exist.indexing.ngram;
import java.io.IOException;
import java.util.*;
import java.util.concurrent.locks.ReentrantLock;
import javax.xml.stream.XMLStreamConstants;
import javax.xml.stream.XMLStreamException;
import javax.xml.stream.XMLStreamReader;
import org.apache.logging.log4j.LogManager;
import org.apache.logging.log4j.Logger;
import org.exist.collections.Collection;
import org.exist.dom.persistent.AttrImpl;
import org.exist.dom.persistent.AbstractCharacterData;
import org.exist.dom.persistent.DocumentImpl;
import org.exist.dom.persistent.DocumentSet;
import org.exist.dom.persistent.ElementImpl;
import org.exist.dom.persistent.ExtArrayNodeSet;
import org.exist.dom.persistent.IStoredNode;
import org.exist.dom.persistent.Match;
import org.exist.dom.persistent.NodeProxy;
import org.exist.dom.persistent.NodeSet;
import org.exist.dom.QName;
import org.exist.dom.persistent.SymbolTable;
import org.exist.indexing.AbstractMatchListener;
import org.exist.indexing.AbstractStreamListener;
import org.exist.indexing.Index;
import org.exist.indexing.IndexController;
import org.exist.indexing.IndexWorker;
import org.exist.indexing.MatchListener;
import org.exist.indexing.OrderedValuesIndex;
import org.exist.indexing.QNamedKeysIndex;
import org.exist.indexing.StreamListener;
import org.exist.indexing.StreamListener.ReindexMode;
import org.exist.numbering.NodeId;
import org.exist.stax.ExtendedXMLStreamReader;
import org.exist.storage.DBBroker;
import org.exist.storage.ElementValue;
import org.exist.storage.IndexSpec;
import org.exist.storage.NodePath;
import org.exist.storage.OccurrenceList;
import org.exist.storage.btree.BTreeCallback;
import org.exist.storage.btree.BTreeException;
import org.exist.storage.btree.IndexQuery;
import org.exist.storage.btree.Value;
import org.exist.storage.index.BFile;
import org.exist.storage.io.VariableByteInput;
import org.exist.storage.io.VariableByteOutputStream;
import org.exist.storage.lock.LockManager;
import org.exist.storage.lock.ManagedLock;
import org.exist.storage.txn.Txn;
import org.exist.util.*;
import org.exist.util.serializer.AttrList;
import org.exist.xquery.*;
import org.w3c.dom.Element;
import org.w3c.dom.Node;
import org.w3c.dom.NodeList;
import org.xml.sax.SAXException;
import static java.nio.charset.StandardCharsets.UTF_8;
/**
* Each index entry maps a key (collectionId, ngram) to a list of occurrences, which has the
* following structure:
*
* [docId : int, nameType: byte, occurrenceCount: int, entrySize: long, [id: NodeId, offset: int, ...]* ]
*/
public class NGramIndexWorker implements OrderedValuesIndex, QNamedKeysIndex {
private static final Logger LOG = LogManager.getLogger(NGramIndexWorker.class);
private static final String INDEX_ELEMENT = "ngram";
private static final String QNAME_ATTR = "qname";
private static final byte IDX_QNAME = 0;
@SuppressWarnings("unused")
private static final byte IDX_GENERIC = 1;
private final DBBroker broker;
private final LockManager lockManager;
private final org.exist.indexing.ngram.NGramIndex index;
private ReindexMode mode = ReindexMode.STORE;
private char[] buf = new char[1024];
private DocumentImpl currentDoc = null;
private Map config;
private final Deque contentStack = new ArrayDeque<>();
@SuppressWarnings("unused")
private IndexController controller;
private final Map ngrams = new TreeMap<>();
private final VariableByteOutputStream os = new VariableByteOutputStream(128);
private NGramMatchListener matchListener = null;
public NGramIndexWorker(final DBBroker broker, final org.exist.indexing.ngram.NGramIndex index) {
this.broker = broker;
this.lockManager = broker.getBrokerPool().getLockManager();
this.index = index;
Arrays.fill(buf, ' ');
}
@Override
public String getIndexId() {
return org.exist.indexing.ngram.NGramIndex.ID;
}
@Override
public String getIndexName() {
return index.getIndexName();
}
public Index getIndex() {
return index;
}
public int getN() {
return index.getN();
}
@Override
public Object configure(final IndexController controller, final NodeList configNodes, final Map namespaces) throws DatabaseConfigurationException {
this.controller = controller;
// We use a map to store the QNames to be indexed
final Map map = new TreeMap<>();
for (int i = 0; i < configNodes.getLength(); i++) {
final Node node = configNodes.item(i);
if (node.getNodeType() == Node.ELEMENT_NODE &&
INDEX_ELEMENT.equals(node.getLocalName())) {
final String qname = ((Element) node).getAttribute(QNAME_ATTR);
if (qname == null || qname.isEmpty()) {
throw new DatabaseConfigurationException("Configuration error: element " + node.getNodeName() +
" must have an attribute " + QNAME_ATTR);
}
if (LOG.isTraceEnabled()) {
LOG.trace("NGram index defined on {}", qname);
}
final NGramIndexConfig config = new NGramIndexConfig(namespaces, qname);
map.put(config.getQName(), config);
}
}
return map;
}
@Override
public void flush() {
switch (mode) {
case STORE:
saveIndex();
break;
case REMOVE_ALL_NODES:
case REMOVE_SOME_NODES:
dropIndex(mode);
break;
}
}
private void saveIndex() {
if (ngrams.isEmpty()) {
return;
}
final VariableByteOutputStream buf = new VariableByteOutputStream();
for (final Map.Entry entry : ngrams.entrySet()) {
final QNameTerm key = entry.getKey();
final OccurrenceList occurences = entry.getValue();
occurences.sort();
os.clear();
os.writeInt(currentDoc.getDocId());
os.writeByte(key.qname.getNameType());
os.writeInt(occurences.getTermCount());
// write nodeids, freq, and offsets to a `temp` buf
try {
NodeId previous = null;
for (int m = 0; m < occurences.getSize(); ) {
previous = occurences.getNode(m).write(previous, buf);
final int freq = occurences.getOccurrences(m);
buf.writeInt(freq);
for (int n = 0; n < freq; n++) {
buf.writeInt(occurences.getOffset(m + n));
}
m += freq;
}
final byte[] bufData = buf.toByteArray();
// clear the buf for the next iteration
buf.clear();
// Write length of node IDs + frequency + offsets (bytes)
os.writeFixedInt(bufData.length);
// Write the node IDs + frequency + offset
os.write(bufData);
} catch (final IOException e) {
LOG.error("IOException while writing nGram index: {}", e.getMessage(), e);
}
final ByteArray data = os.data();
if (data.size() == 0) {
continue;
}
try (final ManagedLock dbLock = lockManager.acquireBtreeWriteLock(index.db.getLockName())) {
final NGramQNameKey value = new NGramQNameKey(currentDoc.getCollection().getId(), key.qname,
index.getBrokerPool().getSymbols(), key.term);
index.db.append(value, data);
} catch (final LockException e) {
LOG.warn("Failed to acquire lock for file {}", FileUtils.fileName(index.db.getFile()), e);
} catch (final IOException e) {
LOG.warn("IO error for file {}", FileUtils.fileName(index.db.getFile()), e);
} catch (final ReadOnlyException e) {
LOG.warn("Read-only error for file {}", FileUtils.fileName(index.db.getFile()), e);
} finally {
os.clear();
}
}
ngrams.clear();
}
private void dropIndex(final ReindexMode mode) {
if (ngrams.isEmpty()) {
return;
}
final VariableByteOutputStream buf = new VariableByteOutputStream();
for (final Map.Entry entry : ngrams.entrySet()) {
final QNameTerm key = entry.getKey();
final OccurrenceList occurencesList = entry.getValue();
occurencesList.sort();
os.clear();
try (final ManagedLock dbLock = lockManager.acquireBtreeWriteLock(index.db.getLockName())) {
final NGramQNameKey value = new NGramQNameKey(currentDoc.getCollection().getId(), key.qname,
index.getBrokerPool().getSymbols(), key.term);
boolean changed = false;
os.clear();
final VariableByteInput is = index.db.getAsStream(value);
if (is == null) {
continue;
}
while (is.available() > 0) {
final int storedDocId = is.readInt();
final byte nameType = is.readByte();
final int occurrences = is.readInt();
//Read (variable) length of node IDs + frequency + offsets
final int length = is.readFixedInt();
if (storedDocId != currentDoc.getDocId()) {
// data are related to another document:
// copy them to any existing data
os.writeInt(storedDocId);
os.writeByte(nameType);
os.writeInt(occurrences);
os.writeFixedInt(length);
is.copyRaw(os, length);
} else {
// data are related to our document:
if (mode == ReindexMode.REMOVE_ALL_NODES) {
// skip them
is.skipBytes(length);
} else {
// removing nodes: need to filter out the node ids to be removed
// feed the new list with the GIDs
final OccurrenceList newOccurrences = new OccurrenceList();
NodeId previous = null;
for (int m = 0; m < occurrences; m++) {
final NodeId nodeId = index.getBrokerPool().getNodeFactory().createFromStream(previous, is);
previous = nodeId;
final int freq = is.readInt();
// add the node to the new list if it is not
// in the list of removed nodes
if (!occurencesList.contains(nodeId)) {
for (int n = 0; n < freq; n++) {
newOccurrences.add(nodeId, is.readInt());
}
} else {
is.skip(freq);
}
}
// append the data from the new list
if (newOccurrences.getSize() > 0) {
//Don't forget this one
newOccurrences.sort();
os.writeInt(currentDoc.getDocId());
os.writeByte(nameType);
os.writeInt(newOccurrences.getTermCount());
// write nodeids, freq, and offsets to a `temp` buf
previous = null;
for (int m = 0; m < newOccurrences.getSize(); ) {
previous = newOccurrences.getNode(m).write(previous, buf);
final int freq = newOccurrences.getOccurrences(m);
buf.writeInt(freq);
for (int n = 0; n < freq; n++) {
buf.writeInt(newOccurrences.getOffset(m + n));
}
m += freq;
}
final byte[] bufData = buf.toByteArray();
// clear the buf for the next iteration
buf.clear();
// Write length of node IDs + frequency + offsets (bytes)
os.writeFixedInt(bufData.length);
// Write the node IDs + frequency + offset
os.write(bufData);
}
}
changed = true;
}
}
//Store new data, if relevant
if (changed) {
//Well, nothing to store : remove the existing data
if (os.data().size() == 0) {
index.db.remove(value);
} else {
if (index.db.put(value, os.data()) == BFile.UNKNOWN_ADDRESS) {
LOG.error("Could not put index data for token '{}' in '{}'", key.term, FileUtils.fileName(index.db.getFile()));
}
}
}
} catch (final LockException e) {
LOG.warn("Failed to acquire lock for file {}", FileUtils.fileName(index.db.getFile()), e);
} catch (final IOException e) {
LOG.warn("IO error for file {}", FileUtils.fileName(index.db.getFile()), e);
} finally {
os.clear();
}
}
ngrams.clear();
}
@Override
public void removeCollection(final Collection collection, final DBBroker broker, final boolean reindex) {
if (LOG.isDebugEnabled()) {
LOG.debug("Dropping NGram index for collection {}", collection.getURI());
}
try (final ManagedLock dbLock = lockManager.acquireBtreeWriteLock(index.db.getLockName())) {
// remove generic index
final Value value = new NGramQNameKey(collection.getId());
index.db.removeAll(null, new IndexQuery(IndexQuery.TRUNC_RIGHT, value));
} catch (final LockException e) {
LOG.warn("Failed to acquire lock for '{}'", FileUtils.fileName(index.db.getFile()), e);
} catch (final BTreeException | IOException e) {
LOG.error(e.getMessage(), e);
}
}
public NodeSet search(final int contextId, final DocumentSet docs, final List qnames, final String query,
final String ngram, final XQueryContext context, final NodeSet contextSet, final int axis)
throws XPathException {
final List searchQnames;
if (qnames == null || qnames.isEmpty()) {
searchQnames = getDefinedIndexes(context.getBroker(), docs);
} else {
searchQnames = qnames;
}
final NodeSet result = new ExtArrayNodeSet(docs.getDocumentCount(), 250);
for (final Iterator iter = docs.getCollectionIterator(); iter.hasNext(); ) {
final int collectionId = iter.next().getId();
for (final QName qname : searchQnames) {
final NGramQNameKey key = new NGramQNameKey(collectionId, qname, index.getBrokerPool().getSymbols(), query);
try (final ManagedLock dbLock = lockManager.acquireBtreeReadLock(index.db.getLockName())) {
final SearchCallback cb = new SearchCallback(contextId, query, ngram, docs, contextSet, context, result, axis == NodeSet.ANCESTOR);
final int op = query.codePointCount(0, query.length()) < getN() ? IndexQuery.TRUNC_RIGHT : IndexQuery.EQ;
index.db.query(new IndexQuery(op, key), cb);
} catch (final LockException e) {
LOG.warn("Failed to acquire lock for '{}'", FileUtils.fileName(index.db.getFile()), e);
} catch (final IOException | BTreeException e) {
LOG.error("{} in '{}'", e.getMessage(), FileUtils.fileName(index.db.getFile()), e);
}
}
}
result.iterate(); // ensure result is ready to use
return result;
}
/**
* Check index configurations for all collection in the given DocumentSet and return
* a list of QNames, which have indexes defined on them.
*
* @param broker the database broker
* @param docs documents
*/
private List getDefinedIndexes(final DBBroker broker, final DocumentSet docs) {
final List indexes = new ArrayList<>(20);
for (final Iterator i = docs.getCollectionIterator(); i.hasNext(); ) {
final Collection collection = i.next();
final IndexSpec idxConf = collection.getIndexConfiguration(broker);
if (idxConf != null) {
final Map config = (Map) idxConf.getCustomIndexSpec(NGramIndex.ID);
if (config != null) {
for (final Object name : config.keySet()) {
indexes.add((QName) name);
}
}
}
}
return indexes;
}
@Override
public boolean checkIndex(final DBBroker broker) {
return true;
}
@Override
public Occurrences[] scanIndex(final XQueryContext context, final DocumentSet docs, final NodeSet contextSet, final Map hints) {
List qnames = hints == null ? null : (List) hints.get(QNAMES_KEY);
//Expects a StringValue
final Object start = hints == null ? null : hints.get(START_VALUE);
//Expects a StringValue
final Object end = hints == null ? null : hints.get(END_VALUE);
if (qnames == null || qnames.isEmpty()) {
qnames = getDefinedIndexes(context.getBroker(), docs);
}
//TODO : use the IndexWorker.VALUE_COUNT hint, if present, to limit the number of returned entries
final IndexScanCallback cb = new IndexScanCallback(docs, contextSet);
for (final QName qname : qnames) {
for (final Iterator i = docs.getCollectionIterator(); i.hasNext(); ) {
final int collectionId = i.next().getId();
final IndexQuery query;
if (start == null) {
final Value startRef = new NGramQNameKey(collectionId);
query = new IndexQuery(IndexQuery.TRUNC_RIGHT, startRef);
} else if (end == null) {
final Value startRef = new NGramQNameKey(collectionId, qname,
index.getBrokerPool().getSymbols(), start.toString().toLowerCase());
query = new IndexQuery(IndexQuery.TRUNC_RIGHT, startRef);
} else {
final Value startRef = new NGramQNameKey(collectionId, qname,
index.getBrokerPool().getSymbols(), start.toString().toLowerCase());
final Value endRef = new NGramQNameKey(collectionId, qname,
index.getBrokerPool().getSymbols(), end.toString().toLowerCase());
query = new IndexQuery(IndexQuery.BW, startRef, endRef);
}
try (final ManagedLock dbLock = lockManager.acquireBtreeReadLock(index.db.getLockName())) {
index.db.query(query, cb);
} catch (final LockException e) {
LOG.warn("Failed to acquire lock for '{}'", FileUtils.fileName(index.db.getFile()), e);
} catch (final IOException | BTreeException e) {
LOG.error(e.getMessage(), e);
} catch (final TerminatedException e) {
LOG.warn(e.getMessage(), e);
}
}
}
return cb.map.values().toArray(new Occurrences[0]);
}
//This listener is always the same whatever the document and the mode
//It should thus be declared static
private final StreamListener listener = new NGramStreamListener();
@Override
public StreamListener getListener() {
return listener;
}
@Override
public MatchListener getMatchListener(final DBBroker broker, final NodeProxy proxy) {
return getMatchListener(broker, proxy, null);
}
public MatchListener getMatchListener(final DBBroker broker, final NodeProxy proxy, final NGramMatchCallback callback) {
boolean needToFilter = false;
Match nextMatch = proxy.getMatches();
while (nextMatch != null) {
if (nextMatch.getIndexId().equals(org.exist.indexing.ngram.NGramIndex.ID)) {
needToFilter = true;
break;
}
nextMatch = nextMatch.getNextMatch();
}
if (!needToFilter) {
return null;
}
if (matchListener == null) {
matchListener = new NGramMatchListener(broker, proxy);
} else {
matchListener.reset(broker, proxy);
}
matchListener.setMatchCallback(callback);
return matchListener;
}
@Override
public IStoredNode getReindexRoot(final IStoredNode node, final NodePath path,
final boolean insert, final boolean includeSelf) {
if (node.getNodeType() == Node.ATTRIBUTE_NODE) {
return null;
}
final IndexSpec indexConf = node.getOwnerDocument().getCollection().getIndexConfiguration(broker);
if (indexConf != null) {
final Map config = (Map) indexConf.getCustomIndexSpec(NGramIndex.ID);
if (config == null) {
return null;
}
boolean reindexRequired = false;
final int len = node.getNodeType() == Node.ELEMENT_NODE && !includeSelf ? path.length() - 1 : path.length();
for (int i = 0; i < len; i++) {
final QName qn = path.getComponent(i);
if (config.get(qn) != null) {
reindexRequired = true;
break;
}
}
if (reindexRequired) {
IStoredNode topMost = null;
IStoredNode currentNode = node;
while (currentNode != null) {
if (config.get(currentNode.getQName()) != null) {
topMost = currentNode;
}
if (currentNode.getOwnerDocument().getCollection().isTempCollection() && currentNode.getNodeId().getTreeLevel() == 2) {
break;
}
//currentNode = (StoredNode) currentNode.getParentNode();
currentNode = currentNode.getParentStoredNode();
}
return topMost;
}
}
return null;
}
/**
* Split the given text string into ngrams. The size of an ngram is determined
* by counting the codepoints, not the characters. The resulting strings may
* thus be longer than the ngram size.
*
* @param text the text to tokenize
*
* @return the tokenized text
*/
public String[] tokenize(final String text) {
final int len = text.codePointCount(0, text.length());
final int gramSize = index.getN();
final String[] ngrams = new String[len];
int next = 0;
int pos = 0;
final StringBuilder bld = new StringBuilder(gramSize);
for (int i = 0; i < len; i++) {
bld.setLength(0);
int offset = pos;
for (int count = 0; count < gramSize && offset < text.length(); count++) {
final int codepoint = Character.toLowerCase(text.codePointAt(offset));
offset += Character.charCount(codepoint);
if (count == 0) {
pos = offset; // advance pos to next character
}
bld.appendCodePoint(codepoint);
}
ngrams[next++] = bld.toString();
}
return ngrams;
}
private void indexText(final NodeId nodeId, final QName qname, final String text) {
final String[] ngram = tokenize(text);
final int len = text.length();
for (int i = 0, j = 0, cp; i < len; i += Character.charCount(cp), j++) {
cp = text.codePointAt(i);
final QNameTerm key = new QNameTerm(qname, ngram[j]);
OccurrenceList list = ngrams.get(key);
if (list == null) {
list = new OccurrenceList();
list.add(nodeId, i);
ngrams.put(key, list);
} else {
list.add(nodeId, i);
}
}
}
@Override
public void setDocument(final DocumentImpl document) {
setDocument(document, ReindexMode.UNKNOWN);
}
@Override
public void setMode(final ReindexMode newMode) {
// wolf: unnecessary call to setDocument?
// setDocument(currentDoc, newMode);
mode = newMode;
}
@Override
public DocumentImpl getDocument() {
return currentDoc;
}
@Override
public ReindexMode getMode() {
return mode;
}
@Override
public void setDocument(final DocumentImpl document, final ReindexMode newMode) {
currentDoc = document;
//config = null;
while (!contentStack.isEmpty()) {
contentStack.pop().reset();
}
final IndexSpec indexConf = document.getCollection().getIndexConfiguration(broker);
if (indexConf != null) {
config = (Map) indexConf.getCustomIndexSpec(org.exist.indexing.ngram.NGramIndex.ID);
}
mode = newMode;
}
@Override
public QueryRewriter getQueryRewriter(final XQueryContext context) {
return null;
}
private class NGramStreamListener extends AbstractStreamListener {
@Override
public void startElement(final Txn transaction, final ElementImpl element, final NodePath path) {
if (config != null && config.get(element.getQName()) != null) {
final XMLString contentBuf = new XMLString();
contentStack.push(contentBuf);
}
super.startElement(transaction, element, path);
}
@Override
public void attribute(final Txn transaction, final AttrImpl attrib, final NodePath path) {
if (config != null && config.get(attrib.getQName()) != null) {
indexText(attrib.getNodeId(), attrib.getQName(), attrib.getValue());
}
super.attribute(transaction, attrib, path);
}
@Override
public void endElement(final Txn transaction, final ElementImpl element, final NodePath path) {
if (config != null && config.get(element.getQName()) != null) {
final XMLString content = contentStack.pop();
indexText(element.getNodeId(), element.getQName(), content.toString());
content.reset();
}
super.endElement(transaction, element, path);
}
@Override
public void characters(final Txn transaction, final AbstractCharacterData text, final NodePath path) {
if (contentStack != null && !contentStack.isEmpty()) {
for (final Iterator it = contentStack.descendingIterator(); it.hasNext(); ) {
it.next().append(text.getXMLString());
}
}
super.characters(transaction, text, path);
}
@Override
public IndexWorker getWorker() {
return NGramIndexWorker.this;
}
}
private static class NGramMatchListener extends AbstractMatchListener {
private Match match;
private Deque offsetStack = null;
private NGramMatchCallback callback = null;
@SuppressWarnings("unused")
private NodeProxy root;
private NGramMatchListener(final DBBroker broker, final NodeProxy proxy) {
reset(broker, proxy);
}
void setMatchCallback(final NGramMatchCallback cb) {
this.callback = cb;
}
protected void reset(final DBBroker broker, final NodeProxy proxy) {
this.root = proxy;
this.match = proxy.getMatches();
setNextInChain(null);
/* Check if an index is defined on an ancestor of the current node.
* If yes, scan the ancestor to get the offset of the first character
* in the current node. For example, if the indexed node is <a>abc<b>de</b>
* and we query for //a[text:ngram-contains(., 'de')]/b, proxy will be a <b> node, but
* the offsets of the matches are relative to the start of <a>.
*/
NodeSet ancestors = null;
Match nextMatch = this.match;
while (nextMatch != null) {
if (proxy.getNodeId().isDescendantOf(nextMatch.getNodeId())) {
if (ancestors == null) {
ancestors = new ExtArrayNodeSet();
}
ancestors.add(new NodeProxy(proxy.getOwnerDocument(), nextMatch.getNodeId()));
}
nextMatch = nextMatch.getNextMatch();
}
if (ancestors != null && !ancestors.isEmpty()) {
for (final NodeProxy p : ancestors) {
final int thisLevel = p.getNodeId().getTreeLevel();
int startOffset = 0;
try {
final XMLStreamReader reader = broker.getXMLStreamReader(p, false);
while (reader.hasNext()) {
final int ev = reader.next();
final NodeId otherId = (NodeId) reader.getProperty(ExtendedXMLStreamReader.PROPERTY_NODE_ID);
if (otherId.equals(proxy.getNodeId())) {
break;
}
final int otherLevel = otherId.getTreeLevel();
if (ev == XMLStreamConstants.CHARACTERS) {
startOffset += reader.getText().length();
} else if (ev == XMLStreamConstants.END_ELEMENT && otherLevel == thisLevel) {
// finished element...
break; // exit-while
}
}
} catch (final IOException | XMLStreamException e) {
LOG.warn("Problem found while serializing XML: {}", e.getMessage(), e);
}
if (offsetStack == null) {
offsetStack = new ArrayDeque<>();
}
offsetStack.push(new NodeOffset(p.getNodeId(), startOffset));
}
}
}
@Override
public void startElement(final QName qname, final AttrList attribs) throws SAXException {
Match nextMatch = match;
// check if there are any matches in the current element
// if yes, push a NodeOffset object to the stack to track
// the node contents
while (nextMatch != null) {
if (nextMatch.getNodeId().equals(getCurrentNode().getNodeId())) {
if (offsetStack == null) {
offsetStack = new ArrayDeque<>();
}
offsetStack.push(new NodeOffset(nextMatch.getNodeId()));
break;
}
nextMatch = nextMatch.getNextMatch();
}
super.startElement(qname, attribs);
}
@Override
public void endElement(final QName qname) throws SAXException {
Match nextMatch = match;
// check if we need to pop the stack
while (nextMatch != null) {
if (nextMatch.getNodeId().equals(getCurrentNode().getNodeId())) {
offsetStack.pop();
break;
}
nextMatch = nextMatch.getNextMatch();
}
super.endElement(qname);
}
@Override
public void characters(final CharSequence seq) throws SAXException {
List offsets = null; // a list of offsets to process
if (offsetStack != null) {
// walk through the stack to find matches which start in
// the current string of text
for (final Iterator it = offsetStack.descendingIterator(); it.hasNext(); ) {
final NodeOffset no = it.next();
final int end = no.offset + seq.length();
// scan all matches
Match next = match;
while (next != null) {
if (next.getIndexId().equals(NGramIndex.ID) && next.getNodeId().equals(no.nodeId)) {
final int freq = next.getFrequency();
for (int j = 0; j < freq; j++) {
final Match.Offset offset = next.getOffset(j);
if (offset.getOffset() < end &&
offset.getOffset() + offset.getLength() > no.offset) {
// add it to the list to be processed
if (offsets == null) {
offsets = new ArrayList<>(4);
}
// adjust the offset and add it to the list
int start = offset.getOffset() - no.offset;
int len = offset.getLength();
if (start < 0) {
len = len - Math.abs(start);
start = 0;
}
if (start + len > seq.length()) {
len = seq.length() - start;
}
offsets.add(new Match.Offset(start, len));
}
}
}
next = next.getNextMatch();
}
// add the length of the current text to the element content length
no.offset = end;
}
}
// now print out the text, marking all matches with a match element
if (offsets != null) {
FastQSort.sort(offsets, 0, offsets.size() - 1);
final String s = seq.toString();
int pos = 0;
for (final Match.Offset offset : offsets) {
if (offset.getOffset() > pos) {
super.characters(s.substring(pos, pos + (offset.getOffset() - pos)));
}
if (callback == null) {
super.startElement(MATCH_ELEMENT, null);
super.characters(s.substring(offset.getOffset(), offset.getOffset() + offset.getLength()));
super.endElement(MATCH_ELEMENT);
} else {
try {
callback.match(nextListener, s.substring(offset.getOffset(), offset.getOffset() + offset.getLength()),
new NodeProxy(getCurrentNode()));
} catch (final XPathException e) {
throw new SAXException("An error occurred while calling match callback: " + e.getMessage(), e);
}
}
pos = offset.getOffset() + offset.getLength();
}
if (pos < s.length()) {
super.characters(s.substring(pos));
}
} else
super.characters(seq);
}
}
private static class NodeOffset {
private final NodeId nodeId;
private int offset;
private NodeOffset(final NodeId nodeId) {
this(nodeId, 0);
}
private NodeOffset(final NodeId nodeId, final int offset) {
this.nodeId = nodeId;
this.offset = offset;
}
}
private static class QNameTerm implements Comparable {
private final QName qname;
private final String term;
private QNameTerm(final QName qname, final String term) {
this.qname = qname;
this.term = term;
}
@Override
public int compareTo(final QNameTerm other) {
final int cmp = qname.compareTo(other.qname);
if (cmp == 0) {
return term.compareTo(other.term);
}
return cmp;
}
}
private static class NGramQNameKey extends Value {
private static final int COLLECTION_ID_OFFSET = 1;
private static final int NAMETYPE_OFFSET = COLLECTION_ID_OFFSET + Collection.LENGTH_COLLECTION_ID; // 5
private static final int NAMESPACE_OFFSET = NAMETYPE_OFFSET + ElementValue.LENGTH_TYPE; // 6
private static final int LOCALNAME_OFFSET = NAMESPACE_OFFSET + SymbolTable.LENGTH_NS_URI; // 8
private static final int NGRAM_OFFSET = LOCALNAME_OFFSET + SymbolTable.LENGTH_LOCAL_NAME; // 10
NGramQNameKey(final int collectionId) {
len = Collection.LENGTH_COLLECTION_ID + 1;
data = new byte[len];
data[0] = IDX_QNAME;
ByteConversion.intToByte(collectionId, data, COLLECTION_ID_OFFSET);
}
/*
public NGramQNameKey(int collectionId, QName qname, SymbolTable symbols) {
len = NGRAM_OFFSET;
data = new byte[len];
data[0] = IDX_QNAME;
ByteConversion.intToByte(collectionId, data, COLLECTION_ID_OFFSET);
final short namespaceId = symbols.getNSSymbol(qname.getNamespaceURI());
final short localNameId = symbols.getSymbol(qname.getLocalPart());
data[NAMETYPE_OFFSET] = qname.getNameType();
ByteConversion.shortToByte(namespaceId, data, NAMESPACE_OFFSET);
ByteConversion.shortToByte(localNameId, data, LOCALNAME_OFFSET);
}
*/
NGramQNameKey(final int collectionId, final QName qname, final SymbolTable symbols, final String ngram) {
len = UTF8.encoded(ngram) + NGRAM_OFFSET;
data = new byte[len];
data[0] = IDX_QNAME;
ByteConversion.intToByte(collectionId, data, COLLECTION_ID_OFFSET);
final short namespaceId = symbols.getNSSymbol(qname.getNamespaceURI());
final short localNameId = symbols.getSymbol(qname.getLocalPart());
data[NAMETYPE_OFFSET] = qname.getNameType();
ByteConversion.shortToByte(namespaceId, data, NAMESPACE_OFFSET);
ByteConversion.shortToByte(localNameId, data, LOCALNAME_OFFSET);
UTF8.encode(ngram, data, NGRAM_OFFSET);
}
}
private final class SearchCallback implements BTreeCallback {
private final int contextId;
@SuppressWarnings("unused")
private final String query;
private final String ngram;
private final DocumentSet docs;
private final NodeSet contextSet;
private final XQueryContext context;
private final NodeSet resultSet;
private final boolean returnAncestor;
SearchCallback(final int contextId, final String query, final String ngram, final DocumentSet docs,
final NodeSet contextSet, final XQueryContext context, final NodeSet result,
final boolean returnAncestor) {
this.contextId = contextId;
this.query = query;
this.ngram = ngram;
this.docs = docs;
this.context = context;
this.contextSet = contextSet;
this.resultSet = result;
this.returnAncestor = returnAncestor;
}
@Override
public boolean indexInfo(final Value key, final long pointer) throws TerminatedException {
final String ngram = new String(key.getData(), NGramQNameKey.NGRAM_OFFSET, key.getLength() - NGramQNameKey.NGRAM_OFFSET, UTF_8);
try {
final VariableByteInput is = index.db.getAsStream(pointer);
//Does the token already has data in the index ?
if (is == null) {
return true;
}
while (is.available() > 0) {
final int storedDocId = is.readInt();
final byte nameType = is.readByte();
final int occurrences = is.readInt();
//Read (variable) length of node IDs + frequency + offsets
final int length = is.readFixedInt();
final DocumentImpl storedDocument = docs.getDoc(storedDocId);
//Exit if the document is not concerned
if (storedDocument == null) {
is.skipBytes(length);
continue;
}
NodeId previous = null;
for (int m = 0; m < occurrences; m++) {
final NodeId nodeId = index.getBrokerPool().getNodeFactory().createFromStream(previous, is);
previous = nodeId;
final int freq = is.readInt();
final NodeProxy nodeProxy = new NodeProxy(storedDocument, nodeId, nameTypeToNodeType(nameType));
// if a context set is specified, we can directly check if the
// matching node is a descendant of one of the nodes
// in the context set.
if (contextSet != null) {
final int sizeHint = contextSet.getSizeHint(storedDocument);
if (returnAncestor) {
final NodeProxy parentNode = contextSet.parentWithChild(nodeProxy, false, true, NodeProxy.UNKNOWN_NODE_LEVEL);
if (parentNode != null) {
readMatches(ngram, is, nodeId, freq, parentNode);
resultSet.add(parentNode, sizeHint);
} else {
is.skip(freq);
}
} else {
readMatches(ngram, is, nodeId, freq, nodeProxy);
resultSet.add(nodeProxy, sizeHint);
}
// otherwise, we add all text nodes without check
} else {
readMatches(ngram, is, nodeId, freq, nodeProxy);
resultSet.add(nodeProxy, Constants.NO_SIZE_HINT);
}
context.proceed();
}
}
return false;
} catch (final IOException e) {
LOG.error(e.getMessage(), e);
return true;
}
}
private short nameTypeToNodeType(final byte nameType) {
switch (nameType) {
case ElementValue.ELEMENT:
return Node.ELEMENT_NODE;
case ElementValue.ATTRIBUTE:
return Node.ATTRIBUTE_NODE;
case ElementValue.UNKNOWN:
default:
return NodeProxy.UNKNOWN_NODE_TYPE;
}
}
private void readMatches(final String current, final VariableByteInput is, final NodeId nodeId, final int freq,
final NodeProxy parentNode) throws IOException {
int diff = 0;
if (current.length() > ngram.length()) {
diff = current.lastIndexOf(ngram);
}
final Match match = new NGramMatch(contextId, nodeId, ngram, freq);
for (int n = 0; n < freq; n++) {
int offset = is.readInt();
if (diff > 0) {
offset += diff;
}
match.addOffset(offset, ngram.length());
}
parentNode.addMatch(match);
}
}
private final class IndexScanCallback implements BTreeCallback {
private final DocumentSet docs;
private NodeSet contextSet;
private final Map map = new TreeMap<>();
IndexScanCallback(final DocumentSet docs, final NodeSet contextSet) {
this.docs = docs;
this.contextSet = contextSet;
}
@Override
public boolean indexInfo(final Value key, final long pointer) {
final String term = new String(key.getData(), NGramQNameKey.NGRAM_OFFSET, key.getLength() - NGramQNameKey.NGRAM_OFFSET, UTF_8);
final VariableByteInput is;
try {
is = index.db.getAsStream(pointer);
} catch (IOException e) {
LOG.error(e.getMessage(), e);
return true;
}
try {
while (is.available() > 0) {
boolean docAdded = false;
final int storedDocId = is.readInt();
@SuppressWarnings("unused") final byte nameType = is.readByte();
final int occurrences = is.readInt();
//Read (variable) length of node IDs + frequency + offsets
final int length = is.readFixedInt();
final DocumentImpl storedDocument = docs.getDoc(storedDocId);
//Exit if the document is not concerned
if (storedDocument == null) {
is.skipBytes(length);
continue;
}
NodeId previous = null;
for (int m = 0; m < occurrences; m++) {
final NodeId nodeId = index.getBrokerPool().getNodeFactory().createFromStream(previous, is);
previous = nodeId;
final int freq = is.readInt();
is.skip(freq);
boolean include = true;
//TODO : revisit
if (contextSet != null) {
final NodeProxy parentNode = contextSet.parentWithChild(storedDocument, nodeId, false, true);
include = (parentNode != null);
}
if (include) {
Occurrences oc = map.get(term);
if (oc == null) {
oc = new Occurrences(term);
map.put(term, oc);
}
if (!docAdded) {
oc.addDocument(storedDocument);
docAdded = true;
}
oc.addOccurrences(freq);
}
}
}
} catch (final IOException e) {
LOG.error("{} in '{}'", e.getMessage(), FileUtils.fileName(index.db.getFile()), e);
}
return true;
}
}
}