modified assortment data structures

git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@148 6c8d7289-2bf4-0310-a012-ef5d649a1542
pull/1/head
orbiter 20 years ago
parent 377fb07250
commit e26ac60c3e

@ -70,7 +70,7 @@
<property name="javacTarget" value="1.4"/>
<!-- Release Configuration -->
<property name="releaseVersion" value="0.374"/>
<property name="releaseVersion" value="0.375"/>
<property name="releaseFile" value="yacy_dev_v${releaseVersion}_${DSTAMP}.tar.gz"/>
<target name="init">

@ -178,7 +178,7 @@ public class Performance_p {
prop.put("maxURLinWordCache", "" + switchboard.wordIndex.maxURLinWordCache());
prop.put("maxWaitingWordFlush", switchboard.getConfig("maxWaitingWordFlush", "180"));
prop.put("wordCacheMax", switchboard.getConfig("wordCacheMax", "10000"));
prop.put("singletonsSize", switchboard.wordIndex.singletonsSize());
prop.put("singletonsSize", switchboard.wordIndex.assortmentSizes()[0]);
// table thread pool settings
GenericObjectPool.Config crawlerPoolConfig = switchboard.cacheLoader.getPoolConfig();

@ -107,7 +107,7 @@ public class transferRWI {
wordHash = estring.substring(0, p);
wordhashes[i] = wordHash;
entry = new plasmaWordIndexEntry(estring.substring(p));
switchboard.wordIndex.addEntries(plasmaWordIndexEntryContainer.instantContainer(wordHash, entry));
switchboard.wordIndex.addEntries(plasmaWordIndexEntryContainer.instantContainer(wordHash, System.currentTimeMillis(), entry));
urlHash = entry.getUrlHash();
if ((!(unknownURL.contains(urlHash))) &&
(!(switchboard.loadedURL.exists(urlHash)))) {

@ -45,7 +45,7 @@
# Contributions and changes to the program code must be marked as such.
# define variables
version='0.374'
version='0.375'
datestr=`date +%Y%m%d`
#release='yacy_v'$version'_'$datestr
release='yacy_dev_v'$version'_'$datestr

@ -43,6 +43,7 @@ package de.anomic.kelondro;
import java.util.Iterator;
import java.util.Comparator;
import java.util.Set;
public class kelondroMergeIterator implements Iterator {
@ -105,4 +106,20 @@ public class kelondroMergeIterator implements Iterator {
public void remove() {
throw new java.lang.UnsupportedOperationException("merge does not support remove");
}
public static Iterator cascade(Set /*of*/ iterators, boolean up) {
// this extends the ability to combine two iterators
// to the abiliy of combining a set of iterators
if (iterators == null) return null;
if (iterators.size() == 0) return null;
return cascade(iterators.iterator(), up);
}
private static Iterator cascade(Iterator /*of*/ iiterators, boolean up) {
if (iiterators == null) return null;
if (!(iiterators.hasNext())) return null;
Iterator one = (Iterator) iiterators.next();
if (!(iiterators.hasNext())) return one;
return new kelondroMergeIterator(one, cascade(iiterators, up), up);
}
}

@ -112,7 +112,7 @@ public final class plasmaSearch {
wordHash = plasmaWordIndexEntry.word2hash(word);
entry = new plasmaWordIndexEntry(urlHash, count, p++, 0, 0,
age, quality, language, doctype, true);
wordIndex.addEntries(plasmaWordIndexEntryContainer.instantContainer(wordHash, entry));
wordIndex.addEntries(plasmaWordIndexEntryContainer.instantContainer(wordHash, System.currentTimeMillis(), entry));
}
//System.out.println("DEBUG: plasmaSearch.addPageIndex: added " + condenser.getWords().size() + " words, flushed " + c + " entries");
return condenser.getWords().size();

@ -76,8 +76,8 @@ public final class plasmaWordIndex {
return ramCache.wordCacheRAMSize();
}
public int singletonsSize() {
return ramCache.singletonsSize();
public int[] assortmentSizes() {
return ramCache.assortmentsSizes();
}
public void setMaxWords(int maxWords) {

@ -47,7 +47,7 @@
For each 'x' there is an assortment database, where 1<=x<=max
If a word appears on more than 'max' web pages, the corresponing url-list
is stored to some kind of back-end database which we consider as the
'slowes' option to save data.
'slowest' option to save data. This here is the fastest file-based.
*/
package de.anomic.plasma;
@ -124,34 +124,21 @@ public final class plasmaWordIndexAssortment {
}
}
public record newRecord(plasmaWordIndexEntry entry, long creationTime) {
return new record(new plasmaWordIndexEntry[]{entry}, creationTime);
}
public record newRecord(plasmaWordIndexEntry[] entries, long creationTime) {
return new record(entries, creationTime);
}
public class record {
public plasmaWordIndexEntry[] entries;
public long creationTime;
public record(plasmaWordIndexEntry[] entries, long creationTime) {
this.entries = entries;
this.creationTime = creationTime;
}
}
public void store(String wordHash, record newRecord) {
public void store(String wordHash, plasmaWordIndexEntryContainer newContainer) {
// stores a word index to assortment database
// this throws an exception if the word hash already existed
//log.logDebug("storeAssortment: wordHash=" + wordHash + ", urlHash=" + entry.getUrlHash() + ", time=" + creationTime);
if (newContainer.size() != assortmentCapacity) throw new RuntimeException("plasmaWordIndexAssortment.store: wrong container size");
byte[][] row = new byte[this.bufferStructureLength][];
row[0] = wordHash.getBytes();
row[1] = kelondroRecords.long2bytes(1, 4);
row[2] = kelondroRecords.long2bytes(newRecord.creationTime, 8);
for (int i = 0; i < assortmentCapacity; i++) {
row[3 + 2 * i] = newRecord.entries[i].getUrlHash().getBytes();
row[4 + 2 * i] = newRecord.entries[i].toEncodedForm(true).getBytes();
row[2] = kelondroRecords.long2bytes(newContainer.updated(), 8);
Iterator entries = newContainer.entries();
plasmaWordIndexEntry entry;
for (int i = 0; i < assortmentCapacity; i++) {
entry = (plasmaWordIndexEntry) entries.next();
row[3 + 2 * i] = entry.getUrlHash().getBytes();
row[4 + 2 * i] = entry.toEncodedForm(true).getBytes();
}
byte[][] oldrow = null;
try {
@ -168,50 +155,7 @@ public final class plasmaWordIndexAssortment {
if (oldrow != null) throw new RuntimeException("Store to assortment ambiguous");
}
/*
public record read(String wordHash) {
// returns a single word index from assortment database; returns null if index does not exist
//log.logDebug("readAssortment: wordHash=" + wordHash);
byte[][] row = null;
try {
row = assortments.get(wordHash.getBytes());
} catch (IOException e) {
log.logFailure("readAssortment/IO-error: " + e.getMessage() + " - reset assortment-DB");
e.printStackTrace();
resetDatabase();
} catch (kelondroException e) {
log.logFailure("readAssortment/kelondro-error: " + e.getMessage() + " - reset assortment-DB");
e.printStackTrace();
resetDatabase();
}
if (row == null) return null;
long creationTime = kelondroRecords.bytes2long(row[2]);
plasmaWordIndexEntry[] wordEntries = new plasmaWordIndexEntry[this.bufferStructureLength];
for (int i = 0; i < assortmentCapacity; i++) {
wordEntries[i] = new plasmaWordIndexEntry(new String(row[3 + 2 * i]), new String(row[4 + 2 * i]));
}
return new record(wordEntries, creationTime);
}
public void remove(String wordHash) {
// deletes a word index from assortment database
//log.logDebug("removeAssortment: wordHash=" + wordHash);
byte[][] row = null;
try {
row = assortments.remove(wordHash.getBytes());
} catch (IOException e) {
log.logFailure("removeAssortment/IO-error: " + e.getMessage() + " - reset assortment-DB");
e.printStackTrace();
resetDatabase();
} catch (kelondroException e) {
log.logFailure("removeAssortment/kelondro-error: " + e.getMessage() + " - reset assortment-DB");
e.printStackTrace();
resetDatabase();
}
}
*/
public record remove(String wordHash) {
public plasmaWordIndexEntryContainer remove(String wordHash) {
// deletes a word index from assortment database
// and returns the content record
byte[][] row = null;
@ -229,12 +173,13 @@ public final class plasmaWordIndexAssortment {
return null;
}
if (row == null) return null;
long creationTime = kelondroRecords.bytes2long(row[2]);
long updateTime = kelondroRecords.bytes2long(row[2]);
plasmaWordIndexEntry[] wordEntries = new plasmaWordIndexEntry[this.bufferStructureLength];
plasmaWordIndexEntryContainer container = new plasmaWordIndexEntryContainer(wordHash);
for (int i = 0; i < assortmentCapacity; i++) {
wordEntries[i] = new plasmaWordIndexEntry(new String(row[3 + 2 * i]), new String(row[4 + 2 * i]));
container.add(new plasmaWordIndexEntry[]{new plasmaWordIndexEntry(new String(row[3 + 2 * i]), new String(row[4 + 2 * i]))}, updateTime);
}
return new record(wordEntries, creationTime);
return container;
}
private void resetDatabase() {

@ -0,0 +1,127 @@
// plasmaWordIndexAssortmentCluster.java
// -------------------------------------
// part of YACY
// (C) by Michael Peter Christen; mc@anomic.de
// first published on http://www.anomic.de
// Frankfurt, Germany, 2005
// last major change: 20.5.2005
//
// This program is free software; you can redistribute it and/or modify
// it under the terms of the GNU General Public License as published by
// the Free Software Foundation; either version 2 of the License, or
// (at your option) any later version.
//
// This program 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 General Public License for more details.
//
// You should have received a copy of the GNU General Public License
// along with this program; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
// Using this software in any meaning (reading, learning, copying, compiling,
// running) means that you agree that the Author(s) is (are) not responsible
// for cost, loss of data or any harm that may be caused directly or indirectly
// by usage of this softare or this documentation. The usage of this software
// is on your own risk. The installation and usage (starting/running) of this
// software may allow other people or application to access your computer and
// any attached devices and is highly dependent on the configuration of the
// software which must be done by the user of the software; the author(s) is
// (are) also not responsible for proper configuration and usage of the
// software, even if provoked by documentation provided together with
// the software.
//
// Any changes to this file according to the GPL as documented in the file
// gpl.txt aside this file in the shipment you received can be done to the
// lines that follows this copyright notice here, but changes must not be
// done inside the copyright notive above. A re-distribution must contain
// the intact and unchanged copyright notice.
// Contributions and changes to the program code must be marked as such.
/*
An assortment-cluster is a set of assortments.
Each one carries a different number of URL's
*/
package de.anomic.plasma;
import java.io.File;
import java.util.*;
import de.anomic.kelondro.*;
import de.anomic.server.serverLog;
public final class plasmaWordIndexAssortmentCluster {
// class variables
private File assortmentsPath;
private int clusterCapacity;
private serverLog log;
private plasmaWordIndexAssortment[] assortments;
private long completeBufferSize;
public plasmaWordIndexAssortmentCluster(File assortmentsPath, int clusterCapacity, int bufferkb, serverLog log) {
// set class variables
if (!(assortmentsPath.exists())) assortmentsPath.mkdirs();
this.clusterCapacity = clusterCapacity;
this.completeBufferSize = bufferkb * 1024;
this.log = log;
this.assortments = new plasmaWordIndexAssortment[clusterCapacity];
// initialize cluster
for (int i = 0; i < clusterCapacity; i++) {
assortments[i] = new plasmaWordIndexAssortment(assortmentsPath, i + 1, (int) completeBufferSize / clusterCapacity, log);
}
}
public plasmaWordIndexEntryContainer storeTry(String wordHash, plasmaWordIndexEntryContainer newContainer) {
// this tries to store the record. If the record does not fit, or a same hash already
// exists and would not fit together with the new record, then the record is deleted from
// the assortmen(s) and returned together with the newRecord.
// if storage was successful, NULL is returned.
if (newContainer.size() > clusterCapacity) return newContainer; // it will not fit
plasmaWordIndexEntryContainer buffer;
for (int i = 0; i < clusterCapacity; i++) {
buffer = assortments[i].remove(wordHash);
if (buffer != null) newContainer.add(buffer);
if (newContainer.size() > clusterCapacity) return newContainer; // it will not fit
}
// we collected all records and the result will fit somewhere..
assortments[newContainer.size() - 1].store(wordHash, newContainer);
// return null to show that we have stored the new Record successfully
return null;
}
public plasmaWordIndexEntryContainer removeFromAll(String wordHash) {
// collect all records from all the assortments and return them
plasmaWordIndexEntryContainer buffer, record = new plasmaWordIndexEntryContainer(wordHash);
for (int i = 0; i < clusterCapacity; i++) {
buffer = assortments[i].remove(wordHash);
if (buffer != null) record.add(buffer);
}
return record;
}
public Iterator hashConjunction(String startWordHash, boolean up) {
HashSet iterators = new HashSet();
for (int i = 0; i < clusterCapacity; i++) iterators.add(assortments[i].hashes(startWordHash, up, true));
return kelondroMergeIterator.cascade(iterators, up);
}
public int sizeTotal() {
int total = 0;
for (int i = 0; i < clusterCapacity; i++) total += assortments[i].size();
return total;
}
public int[] sizes() {
int[] sizes = new int[clusterCapacity];
for (int i = 0; i < clusterCapacity; i++) sizes[i] = assortments[i].size();
return sizes;
}
public void close() {
for (int i = 0; i < clusterCapacity; i++) assortments[i].close();
}
}

@ -55,6 +55,8 @@ public final class plasmaWordIndexCache implements plasmaWordIndexInterface {
private static final String indexDumpFileName = "indexDump0.stack";
private static final String oldSingletonFileName = "indexSingletons0.db";
private static final String newSingletonFileName = "indexAssortment001.db";
private static final int assortmentLimit = 1;
// class variables
private File databaseRoot;
@ -64,7 +66,7 @@ public final class plasmaWordIndexCache implements plasmaWordIndexInterface {
private HashMap hashDate;
private int maxWords;
private serverLog log;
private plasmaWordIndexAssortment singletons;
private plasmaWordIndexAssortmentCluster assortmentCluster;
private int singletonBufferSize; //kb
// calculated constants
@ -92,7 +94,7 @@ public final class plasmaWordIndexCache implements plasmaWordIndexInterface {
this.maxWords = 10000;
this.backend = backend;
this.log = log;
this.singletons = new plasmaWordIndexAssortment(databaseRoot, 1, singletonBufferSize, log);
this.assortmentCluster = new plasmaWordIndexAssortmentCluster(databaseRoot, assortmentLimit, singletonBufferSize, log);
// read in dump of last session
try {
@ -212,12 +214,12 @@ public final class plasmaWordIndexCache implements plasmaWordIndexInterface {
this.maxWords = maxWords;
}
public int singletonsSize() {
return singletons.size();
public int[] assortmentsSizes() {
return assortmentCluster.sizes();
}
public int size() {
return java.lang.Math.max(singletons.size(), java.lang.Math.max(backend.size(), cache.size()));
return java.lang.Math.max(assortmentCluster.sizeTotal(), java.lang.Math.max(backend.size(), cache.size()));
}
public Iterator wordHashes(String startWordHash, boolean up) {
@ -229,7 +231,7 @@ public final class plasmaWordIndexCache implements plasmaWordIndexInterface {
return new kelondroMergeIterator(
new kelondroMergeIterator(
cache.keySet().iterator(),
singletons.hashes(startWordHash, true, false),
assortmentCluster.hashConjunction(startWordHash, true),
true),
backend.wordHashes(startWordHash, true),
true);
@ -258,28 +260,29 @@ public final class plasmaWordIndexCache implements plasmaWordIndexInterface {
}
// now decide where to flush that container
plasmaWordIndexAssortment.record singleton = singletons.remove(key);
if (singleton == null) {
// not found in singletons
if (container.size() == 1) {
// it is a singleton: store to singleton
singletons.store(key, singletons.newRecord(container.getOne(), time));
return 1;
plasmaWordIndexEntryContainer flushedFromAssortment = assortmentCluster.removeFromAll(key);
if (flushedFromAssortment == null) {
// not found in assortments
if (container.size() <= assortmentLimit) {
// this fits into the assortments
plasmaWordIndexEntryContainer feedback = assortmentCluster.storeTry(key, container);
if (feedback == null) {
return container.size();
} else {
// *** should care about another option here ***
return backend.addEntries(feedback, time);
}
} else {
// store to back-end; this should be a rare case
return backend.addEntries(container, time);
}
} else {
// we have a singleton and need to integrate this in the flush
plasmaWordIndexEntry oldEntry = singleton.entries[0];
long oldTime = singleton.creationTime;
// put new entries to the container
if (!(container.contains(oldEntry.getUrlHash()))) container.add(oldEntry);
// we have some records and must integrate them into the flush
container.add(flushedFromAssortment);
// possibly reintegrate
if (reintegrate) {
// put singleton together with container back to ram
// put assortmentRecord together with container back to ram
synchronized (cache) {
cache.put(key, container);
hashScore.setScore(key, container.size());
@ -288,22 +291,20 @@ public final class plasmaWordIndexCache implements plasmaWordIndexInterface {
return -1;
} else {
// add this to the backend
return backend.addEntries(container, java.lang.Math.max(time, oldTime));
return backend.addEntries(container, java.lang.Math.max(time, flushedFromAssortment.updated()));
}
}
}
private boolean flushFromSingleton(String key) {
// this should only be called if the singleton shall be deleted or returned in an index entity
plasmaWordIndexAssortment.record singleton = singletons.remove(key);
if (singleton == null) {
plasmaWordIndexEntryContainer container = assortmentCluster.removeFromAll(key);
if (container == null) {
return false;
} else {
// we have a singleton
plasmaWordIndexEntry entry = (plasmaWordIndexEntry) singleton.entries[0];
long time = singleton.creationTime;
// we have a non-empty entry-container
// integrate it to the backend
return backend.addEntries(plasmaWordIndexEntryContainer.instantContainer(key, entry), time) > 0;
return backend.addEntries(container, container.updated()) > 0;
}
}
@ -395,7 +396,7 @@ public final class plasmaWordIndexCache implements plasmaWordIndexInterface {
hashScore.deleteScore(wordHash);
hashDate.remove(wordHash);
}
singletons.remove(wordHash);
assortmentCluster.removeFromAll(wordHash);
backend.deleteIndex(wordHash);
}
@ -405,7 +406,7 @@ public final class plasmaWordIndexCache implements plasmaWordIndexInterface {
return backend.removeEntries(wordHash, urlHashes, deleteComplete);
}
public synchronized int addEntries(plasmaWordIndexEntryContainer container, long creationTime) {
public synchronized int addEntries(plasmaWordIndexEntryContainer container, long updateTime) {
//serverLog.logDebug("PLASMA INDEXING", "addEntryToIndexMem: cache.size=" + cache.size() + "; hashScore.size=" + hashScore.size());
if (cache.size() >= this.maxWords) flushFromMemToLimit();
//if (flushc > 0) serverLog.logDebug("PLASMA INDEXING", "addEntryToIndexMem - flushed " + flushc + " entries");
@ -420,25 +421,25 @@ public final class plasmaWordIndexCache implements plasmaWordIndexInterface {
if (added > 0) {
cache.put(wordHash, entries);
hashScore.addScore(wordHash, added);
hashDate.put(wordHash, new Long(creationTime));
hashDate.put(wordHash, new Long(updateTime));
}
}
//System.out.println("DEBUG: cache = " + cache.toString());
return added;
}
private void addEntry(String wordHash, plasmaWordIndexEntry newEntry, long creationTime) {
private void addEntry(String wordHash, plasmaWordIndexEntry newEntry, long updateTime) {
plasmaWordIndexEntryContainer entries = (plasmaWordIndexEntryContainer) cache.get(wordHash);
if (entries == null) entries = new plasmaWordIndexEntryContainer(wordHash);
if (entries.add(newEntry)) {
if (entries.add(new plasmaWordIndexEntry[]{newEntry}, updateTime) > 0) {
cache.put(wordHash, entries);
hashScore.incScore(wordHash);
hashDate.put(wordHash, new Long(creationTime));
hashDate.put(wordHash, new Long(updateTime));
}
}
public void close(int waitingSeconds) {
singletons.close();
assortmentCluster.close();
try {
dump(waitingSeconds);
} catch (IOException e){

@ -39,6 +39,17 @@
// the intact and unchanged copyright notice.
// Contributions and changes to the program code must be marked as such.
/*
an indexContainer is a bag of indexEntries for a single word
such an container represents a RWI snipplet:
it collects a new RWI until it is so big that it should be flushed to either
- an indexAssortment: collection of indexContainers of same size or
- the backend storage
the creationTime is necessary to organize caching of containers
*/
package de.anomic.plasma;
import java.util.HashMap;
@ -48,10 +59,12 @@ import de.anomic.server.serverCodings;
public class plasmaWordIndexEntryContainer implements Comparable {
private String wordHash;
private HashMap container;
private HashMap container; // urlHash/plasmaWordIndexEntry - Mapping
private long updateTime;
public plasmaWordIndexEntryContainer(String wordHash) {
this.wordHash = wordHash;
this.updateTime = 0;
container = new HashMap(); // a urlhash/plasmaWordIndexEntry - relation
}
@ -59,16 +72,19 @@ public class plasmaWordIndexEntryContainer implements Comparable {
return container.size();
}
public long updated() {
return updateTime;
}
public String wordHash() {
return wordHash;
}
public boolean add(plasmaWordIndexEntry entry) {
// returns true if the new entry was added, false if it already existet
String urlHash = entry.getUrlHash();
if (container.containsKey(urlHash)) return false;
container.put(urlHash, entry);
return true;
public int add(plasmaWordIndexEntry[] entries, long updateTime) {
int c = 0;
for (int i = 0; i < entries.length; i++) if (add(entries[i])) c++;
this.updateTime = java.lang.Math.max(this.updateTime, updateTime);
return c;
}
public int add(plasmaWordIndexEntryContainer c) {
@ -78,28 +94,38 @@ public class plasmaWordIndexEntryContainer implements Comparable {
while (i.hasNext()) {
if (add((plasmaWordIndexEntry) i.next())) x++;
}
this.updateTime = java.lang.Math.max(this.updateTime, c.updateTime);
return x;
}
private boolean add(plasmaWordIndexEntry entry) {
// returns true if the new entry was added, false if it already existet
String urlHash = entry.getUrlHash();
if (container.containsKey(urlHash)) return false;
container.put(urlHash, entry);
return true;
}
public boolean contains(String urlHash) {
return container.containsKey(urlHash);
}
public plasmaWordIndexEntry getOne() {
return (plasmaWordIndexEntry) container.values().toArray()[0];
public plasmaWordIndexEntry[] getEntryArray() {
return (plasmaWordIndexEntry[]) container.values().toArray();
}
public Iterator entries() {
// returns an iterator of plasmaWordIndexEntry objects
return container.values().iterator();
}
public static plasmaWordIndexEntryContainer instantContainer(String wordHash, plasmaWordIndexEntry entry) {
public static plasmaWordIndexEntryContainer instantContainer(String wordHash, long creationTime, plasmaWordIndexEntry entry) {
plasmaWordIndexEntryContainer c = new plasmaWordIndexEntryContainer(wordHash);
c.add(entry);
c.updateTime = creationTime;
return c;
}
public String toString() {
return "C[" + wordHash + "] has " + container.size() + " entries";
}

@ -298,7 +298,9 @@ public class yacyClient {
plasmaWordIndexEntry entry = new plasmaWordIndexEntry(link.hash(), link.wordCount(), 0, 0, 0,
plasmaSearch.calcVirtualAge(link.moddate()), link.quality(),
link.language(), link.doctype(), false);
for (int m = 0; m < words; m++) container[m].add(entry);
for (int m = 0; m < words; m++) {
container[m].add(new plasmaWordIndexEntry[]{entry}, System.currentTimeMillis());
}
}
// finally insert the containers to the index

Loading…
Cancel
Save