You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
yacy_search_server/source/de/anomic/index/indexRAMRI.java

520 lines
22 KiB

// indexRAMRI.java
// (C) 2005, 2006 by Michael Peter Christen; mc@yacy.net, Frankfurt a. M., Germany
// first published 2005 on http://yacy.net
//
// This is a part of YaCy, a peer-to-peer based web search engine
//
// $LastChangedDate: 2006-04-02 22:40:07 +0200 (So, 02 Apr 2006) $
// $LastChangedRevision: 1986 $
// $LastChangedBy: orbiter $
//
// LICENSE
//
// 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
package de.anomic.index;
import java.io.File;
import java.io.IOException;
import java.util.Collections;
import java.util.Iterator;
import java.util.Map;
import java.util.Set;
import java.util.SortedMap;
import java.util.TreeMap;
import de.anomic.kelondro.kelondroBase64Order;
import de.anomic.kelondro.kelondroBufferedRA;
import de.anomic.kelondro.kelondroCloneableIterator;
import de.anomic.kelondro.kelondroException;
import de.anomic.kelondro.kelondroFixedWidthArray;
import de.anomic.kelondro.kelondroMScoreCluster;
import de.anomic.kelondro.kelondroNaturalOrder;
import de.anomic.kelondro.kelondroRow;
import de.anomic.server.serverByteBuffer;
import de.anomic.server.serverFileUtils;
import de.anomic.server.logging.serverLog;
import de.anomic.yacy.yacySeedDB;
public final class indexRAMRI implements indexRI {
// class variables
private final File databaseRoot;
protected final SortedMap cache; // wordhash-container
private final kelondroMScoreCluster hashScore;
private final kelondroMScoreCluster hashDate;
private long initTime;
private int cacheMaxCount;
public int cacheReferenceCountLimit;
public long cacheReferenceAgeLimit;
private final serverLog log;
private String indexArrayFileName;
private kelondroRow payloadrow;
private kelondroRow bufferStructureBasis;
public indexRAMRI(File databaseRoot, kelondroRow payloadrow, int wCacheReferenceCountLimitInit, long wCacheReferenceAgeLimitInit, String dumpname, serverLog log) {
// creates a new index cache
// the cache has a back-end where indexes that do not fit in the cache are flushed
this.databaseRoot = databaseRoot;
this.cache = Collections.synchronizedSortedMap(new TreeMap());
this.hashScore = new kelondroMScoreCluster();
this.hashDate = new kelondroMScoreCluster();
this.initTime = System.currentTimeMillis();
this.cacheMaxCount = 10000;
this.cacheReferenceCountLimit = wCacheReferenceCountLimitInit;
this.cacheReferenceAgeLimit = wCacheReferenceAgeLimitInit;
this.log = log;
this.indexArrayFileName = dumpname;
this.payloadrow = payloadrow;
this.bufferStructureBasis = new kelondroRow(
"byte[] wordhash-" + yacySeedDB.commonHashLength + ", " +
"Cardinal occ-4 {b256}, " +
"Cardinal time-8 {b256}, " +
"byte[] urlprops-" + payloadrow.objectsize(),
kelondroBase64Order.enhancedCoder, 0);
// read in dump of last session
try {
restore();
} catch (IOException e){
log.logSevere("unable to restore cache dump: " + e.getMessage(), e);
}
}
public int minMem() {
// there is no specific large array that needs to be maintained
// this value is just a guess of the possible overhead
return 100 * 1024; // 100 kb
}
public synchronized long getUpdateTime(String wordHash) {
indexContainer entries = getContainer(wordHash, null);
if (entries == null) return 0;
return entries.updated();
}
private void dump() throws IOException {
log.logConfig("creating dump for index cache '" + indexArrayFileName + "', " + cache.size() + " words (and much more urls)");
File indexDumpFile = new File(databaseRoot, indexArrayFileName);
if (indexDumpFile.exists()) indexDumpFile.delete();
kelondroFixedWidthArray dumpArray = null;
kelondroBufferedRA writeBuffer = null;
if (false /*serverMemory.available() > 50 * bufferStructureBasis.objectsize() * cache.size()*/) {
writeBuffer = new kelondroBufferedRA();
dumpArray = new kelondroFixedWidthArray(writeBuffer, indexDumpFile.getCanonicalPath(), bufferStructureBasis, 0);
log.logInfo("started dump of ram cache: " + cache.size() + " words; memory-enhanced write");
} else {
dumpArray = new kelondroFixedWidthArray(indexDumpFile, bufferStructureBasis, 0);
log.logInfo("started dump of ram cache: " + cache.size() + " words; low-memory write");
}
long startTime = System.currentTimeMillis();
long messageTime = System.currentTimeMillis() + 5000;
long wordsPerSecond = 0, wordcount = 0, urlcount = 0;
Map.Entry entry;
String wordHash;
indexContainer container;
long updateTime;
indexRWIEntry iEntry;
kelondroRow.Entry row = dumpArray.row().newEntry();
byte[] occ, time;
// write wCache
synchronized (cache) {
Iterator i = cache.entrySet().iterator();
while (i.hasNext()) {
// get entries
entry = (Map.Entry) i.next();
wordHash = (String) entry.getKey();
updateTime = getUpdateTime(wordHash);
container = (indexContainer) entry.getValue();
// put entries on stack
if (container != null) {
Iterator ci = container.entries();
occ = kelondroNaturalOrder.encodeLong(container.size(), 4);
time = kelondroNaturalOrder.encodeLong(updateTime, 8);
while (ci.hasNext()) {
iEntry = (indexRWIEntry) ci.next();
row.setCol(0, wordHash.getBytes());
row.setCol(1, occ);
row.setCol(2, time);
row.setCol(3, iEntry.toKelondroEntry().bytes());
dumpArray.set((int) urlcount++, row);
}
}
wordcount++;
i.remove(); // free some mem
// write a log
if (System.currentTimeMillis() > messageTime) {
// System.gc(); // for better statistic
wordsPerSecond = wordcount * 1000
/ (1 + System.currentTimeMillis() - startTime);
log.logInfo("dump status: " + wordcount
+ " words done, "
+ (cache.size() / (wordsPerSecond + 1))
+ " seconds remaining, free mem = "
+ (Runtime.getRuntime().freeMemory() / 1024 / 1024)
+ "MB");
messageTime = System.currentTimeMillis() + 5000;
}
}
}
if (writeBuffer != null) {
serverByteBuffer bb = writeBuffer.getBuffer();
//System.out.println("*** byteBuffer size = " + bb.length());
serverFileUtils.write(bb.getBytes(), indexDumpFile);
writeBuffer.close();
}
dumpArray.close();
dumpArray = null;
log.logInfo("finished dump of ram cache: " + urlcount + " word/URL relations in " + ((System.currentTimeMillis() - startTime) / 1000) + " seconds");
}
private long restore() throws IOException {
File indexDumpFile = new File(databaseRoot, indexArrayFileName);
if (!(indexDumpFile.exists())) return 0;
kelondroFixedWidthArray dumpArray;
kelondroBufferedRA readBuffer = null;
if (false /*serverMemory.available() > indexDumpFile.length() * 2*/) {
readBuffer = new kelondroBufferedRA(new serverByteBuffer(serverFileUtils.read(indexDumpFile)));
dumpArray = new kelondroFixedWidthArray(readBuffer, indexDumpFile.getCanonicalPath(), bufferStructureBasis, 0);
log.logInfo("started restore of ram cache '" + indexArrayFileName + "', " + dumpArray.size() + " word/URL relations; memory-enhanced read");
} else {
dumpArray = new kelondroFixedWidthArray(indexDumpFile, bufferStructureBasis, 0);
log.logInfo("started restore of ram cache '" + indexArrayFileName + "', " + dumpArray.size() + " word/URL relations; low-memory read");
}
long startTime = System.currentTimeMillis();
long messageTime = System.currentTimeMillis() + 5000;
long urlCount = 0, urlsPerSecond = 0;
try {
synchronized (cache) {
Iterator i = dumpArray.contentRows(-1);
String wordHash;
//long creationTime;
indexRWIEntry wordEntry;
kelondroRow.EntryIndex row;
//Runtime rt = Runtime.getRuntime();
while (i.hasNext()) {
// get out one entry
row = (kelondroRow.EntryIndex) i.next();
if ((row == null) || (row.empty(0)) || (row.empty(3))) continue;
wordHash = row.getColString(0, "UTF-8");
//creationTime = kelondroRecords.bytes2long(row[2]);
wordEntry = new indexRWIEntry(row.getColBytes(3));
// store to cache
addEntry(wordHash, wordEntry, startTime, false);
urlCount++;
// protect against memory shortage
//while (rt.freeMemory() < 1000000) {flushFromMem(); java.lang.System.gc();}
// write a log
if (System.currentTimeMillis() > messageTime) {
System.gc(); // for better statistic
urlsPerSecond = 1 + urlCount * 1000 / (1 + System.currentTimeMillis() - startTime);
log.logInfo("restoring status: " + urlCount + " urls done, " + ((dumpArray.size() - urlCount) / urlsPerSecond) + " seconds remaining, free mem = " + (Runtime.getRuntime().freeMemory() / 1024 / 1024) + "MB");
messageTime = System.currentTimeMillis() + 5000;
}
}
}
if (readBuffer != null) readBuffer.close();
dumpArray.close();
dumpArray = null;
log.logConfig("finished restore: " + cache.size() + " words in " + ((System.currentTimeMillis() - startTime) / 1000) + " seconds");
} catch (kelondroException e) {
// restore failed
log.logSevere("failed restore of indexCache array dump: " + e.getMessage(), e);
} finally {
if (dumpArray != null) try {dumpArray.close();}catch(Exception e){}
}
return urlCount;
}
// cache settings
public int maxURLinCache() {
if (hashScore.size() == 0) return 0;
return hashScore.getMaxScore();
}
public long minAgeOfCache() {
if (hashDate.size() == 0) return 0;
return System.currentTimeMillis() - longEmit(hashDate.getMaxScore());
}
public long maxAgeOfCache() {
if (hashDate.size() == 0) return 0;
return System.currentTimeMillis() - longEmit(hashDate.getMinScore());
}
public void setMaxWordCount(int maxWords) {
this.cacheMaxCount = maxWords;
}
public int getMaxWordCount() {
return this.cacheMaxCount;
}
public int size() {
return cache.size();
}
public synchronized int indexSize(String wordHash) {
indexContainer cacheIndex = (indexContainer) cache.get(wordHash);
if (cacheIndex == null) return 0;
return cacheIndex.size();
}
public synchronized kelondroCloneableIterator wordContainers(String startWordHash, boolean rot) {
// we return an iterator object that creates top-level-clones of the indexContainers
// in the cache, so that manipulations of the iterated objects do not change
// objects in the cache.
return new wordContainerIterator(startWordHash, rot);
}
public class wordContainerIterator implements kelondroCloneableIterator {
// this class exists, because the wCache cannot be iterated with rotation
// and because every indeContainer Object that is iterated must be returned as top-level-clone
// so this class simulates wCache.tailMap(startWordHash).values().iterator()
// plus the mentioned features
private boolean rot;
private Iterator iterator;
public wordContainerIterator(String startWordHash, boolean rot) {
this.rot = rot;
this.iterator = (startWordHash == null) ? cache.values().iterator() : cache.tailMap(startWordHash).values().iterator();
// The collection's iterator will return the values in the order that their corresponding keys appear in the tree.
}
public Object clone(Object secondWordHash) {
return new wordContainerIterator((String) secondWordHash, rot);
}
public boolean hasNext() {
if (rot) return true;
return iterator.hasNext();
}
public Object next() {
if (iterator.hasNext()) {
return ((indexContainer) iterator.next()).topLevelClone();
} else {
// rotation iteration
if (rot) {
iterator = cache.values().iterator();
return ((indexContainer) iterator.next()).topLevelClone();
} else {
return null;
}
}
}
public void remove() {
iterator.remove();
}
}
public synchronized String maxScoreWordHash() {
if (cache.size() == 0) return null;
try {
return (String) hashScore.getMaxObject();
} catch (Exception e) {
log.logSevere("flushFromMem: " + e.getMessage(), e);
}
return null;
}
public synchronized String bestFlushWordHash() {
// select appropriate hash
// we have 2 different methods to find a good hash:
// - the oldest entry in the cache
// - the entry with maximum count
if (cache.size() == 0) return null;
try {
String hash = null;
int count = hashScore.getMaxScore();
if ((count >= cacheReferenceCountLimit) &&
((hash = (String) hashScore.getMaxObject()) != null)) {
// we MUST flush high-score entries, because a loop deletes entries in cache until this condition fails
// in this cache we MUST NOT check wCacheMinAge
return hash;
}
long oldestTime = longEmit(hashDate.getMinScore());
if (((System.currentTimeMillis() - oldestTime) > cacheReferenceAgeLimit) &&
((hash = (String) hashDate.getMinObject()) != null)) {
// flush out-dated entries
return hash;
}
// cases with respect to memory situation
if (Runtime.getRuntime().freeMemory() < 100000) {
// urgent low-memory case
hash = (String) hashScore.getMaxObject(); // flush high-score entries (saves RAM)
} else {
// not-efficient-so-far case. cleans up unnecessary cache slots
hash = (String) hashDate.getMinObject(); // flush oldest entries
}
return hash;
} catch (Exception e) {
log.logSevere("flushFromMem: " + e.getMessage(), e);
}
return null;
}
private int intTime(long longTime) {
return (int) Math.max(0, ((longTime - initTime) / 1000));
}
private long longEmit(int intTime) {
return (((long) intTime) * (long) 1000) + initTime;
}
public boolean hasContainer(String wordHash) {
return cache.containsKey(wordHash);
}
public int sizeContainer(String wordHash) {
return ((indexContainer) cache.get(wordHash)).size();
}
public synchronized indexContainer getContainer(String wordHash, Set urlselection) {
// retrieve container
indexContainer container = (indexContainer) cache.get(wordHash);
// We must not use the container from cache to store everything we find,
// as that container remains linked to in the cache and might be changed later
// while the returned container is still in use.
// create a clone from the container
if (container != null) container = container.topLevelClone();
// select the urlselection
if ((urlselection != null) && (container != null)) container.select(urlselection);
return container;
}
public synchronized indexContainer deleteContainer(String wordHash) {
// returns the index that had been deleted
indexContainer container = (indexContainer) cache.remove(wordHash);
hashScore.deleteScore(wordHash);
hashDate.deleteScore(wordHash);
return container;
}
public synchronized boolean removeEntry(String wordHash, String urlHash) {
indexContainer c = (indexContainer) cache.get(wordHash);
if ((c != null) && (c.remove(urlHash) != null)) {
// removal successful
if (c.size() == 0) {
deleteContainer(wordHash);
} else {
cache.put(wordHash, c);
hashScore.decScore(wordHash);
hashDate.setScore(wordHash, intTime(System.currentTimeMillis()));
}
return true;
}
return false;
}
public synchronized int removeEntries(String wordHash, Set urlHashes) {
if (urlHashes.size() == 0) return 0;
indexContainer c = (indexContainer) cache.get(wordHash);
int count;
if ((c != null) && ((count = c.removeEntries(urlHashes)) > 0)) {
// removal successful
if (c.size() == 0) {
deleteContainer(wordHash);
} else {
cache.put(wordHash, c);
hashScore.setScore(wordHash, c.size());
hashDate.setScore(wordHash, intTime(System.currentTimeMillis()));
}
return count;
}
return 0;
}
public synchronized int tryRemoveURLs(String urlHash) {
// this tries to delete an index from the cache that has this
// urlHash assigned. This can only work if the entry is really fresh
// Such entries must be searched in the latest entries
int delCount = 0;
Iterator i = cache.entrySet().iterator();
Map.Entry entry;
String wordhash;
indexContainer c;
while (i.hasNext()) {
entry = (Map.Entry) i.next();
wordhash = (String) entry.getKey();
// get container
c = (indexContainer) entry.getValue();
if (c.remove(urlHash) != null) {
if (c.size() == 0) {
i.remove();
} else {
cache.put(wordhash, c); // superfluous?
}
delCount++;
}
}
return delCount;
}
public synchronized void addEntries(indexContainer container, long updateTime, boolean dhtCase) {
// this puts the entries into the cache, not into the assortment directly
int added = 0;
if ((container == null) || (container.size() == 0)) return;
// put new words into cache
String wordHash = container.getWordHash();
indexContainer entries = (indexContainer) cache.get(wordHash); // null pointer exception? wordhash != null! must be cache==null
if (entries == null) {
entries = container.topLevelClone();
added = entries.size();
} else {
added = entries.putAllRecent(container);
}
if (added > 0) {
cache.put(wordHash, entries);
hashScore.addScore(wordHash, added);
hashDate.setScore(wordHash, intTime(updateTime));
}
entries = null;
}
public synchronized void addEntry(String wordHash, indexRWIEntry newEntry, long updateTime, boolean dhtCase) {
indexContainer container = (indexContainer) cache.get(wordHash);
if (container == null) container = new indexContainer(wordHash, this.payloadrow, 1);
container.put(newEntry);
cache.put(wordHash, container);
hashScore.incScore(wordHash);
hashDate.setScore(wordHash, intTime(updateTime));
}
public synchronized void close() {
// dump cache
try {
dump();
} catch (IOException e){
log.logSevere("unable to dump cache: " + e.getMessage(), e);
}
}
}