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.
412 lines
15 KiB
412 lines
15 KiB
/**
|
|
* RAMIndex
|
|
* Copyright 2008 by Michael Peter Christen; mc@yacy.net, Frankfurt a. M., Germany
|
|
* First released 07.01.2008 at http://yacy.net
|
|
*
|
|
* $LastChangedDate: 2010-06-16 17:11:21 +0200 (Mi, 16 Jun 2010) $
|
|
* $LastChangedRevision: 6922 $
|
|
* $LastChangedBy: orbiter $
|
|
*
|
|
* 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 program in the file lgpl21.txt
|
|
* If not, see <http://www.gnu.org/licenses/>.
|
|
*/
|
|
|
|
package net.yacy.kelondro.index;
|
|
|
|
import java.io.IOException;
|
|
import java.util.ArrayList;
|
|
import java.util.Iterator;
|
|
import java.util.List;
|
|
import java.util.Map;
|
|
import java.util.TreeMap;
|
|
|
|
import net.yacy.kelondro.index.Row.Entry;
|
|
import net.yacy.kelondro.order.CloneableIterator;
|
|
import net.yacy.kelondro.order.MergeIterator;
|
|
import net.yacy.kelondro.order.StackIterator;
|
|
|
|
|
|
public final class RAMIndex implements Index, Iterable<Row.Entry> {
|
|
|
|
private static final TreeMap<String, RAMIndex> objectTracker = new TreeMap<String, RAMIndex>();
|
|
|
|
private final String name;
|
|
private final Row rowdef;
|
|
private RowSet index0;
|
|
private RowSet index1;
|
|
private final Row.EntryComparator entryComparator;
|
|
//private final int spread;
|
|
|
|
public RAMIndex(String name, final Row rowdef, final int expectedspace) {
|
|
this.name = name;
|
|
this.rowdef = rowdef;
|
|
this.entryComparator = new Row.EntryComparator(rowdef.objectOrder);
|
|
//this.spread = Math.max(10, expectedspace / 3000);
|
|
reset();
|
|
objectTracker.put(name, this);
|
|
}
|
|
|
|
private RAMIndex(String name, final Row rowdef, RowSet index0, RowSet index1, Row.EntryComparator entryComparator) {
|
|
this.name = name;
|
|
this.rowdef = rowdef;
|
|
this.index0 = index0;
|
|
this.index1 = index1;
|
|
this.entryComparator = entryComparator;
|
|
objectTracker.put(name, this);
|
|
}
|
|
|
|
public static final Iterator<Map.Entry<String, RAMIndex>> objects() {
|
|
return objectTracker.entrySet().iterator();
|
|
}
|
|
|
|
public RAMIndex clone() {
|
|
return new RAMIndex(this.name + ".clone", this.rowdef, index0.clone(), index1.clone(), entryComparator);
|
|
}
|
|
|
|
public void clear() {
|
|
reset();
|
|
}
|
|
|
|
public void trim() {
|
|
if (this.index0 != null) this.index0.trim();
|
|
if (this.index1 != null) this.index1.trim();
|
|
}
|
|
|
|
public final synchronized void reset() {
|
|
this.index0 = null; // first flush RAM to make room
|
|
this.index0 = new RowSet(rowdef);
|
|
this.index1 = null; // to show that this is the initialization phase
|
|
}
|
|
|
|
public final synchronized void reset(final int initialspace) throws RowSpaceExceededException {
|
|
this.index0 = null; // first flush RAM to make room
|
|
this.index0 = new RowSet(rowdef, initialspace);
|
|
this.index1 = null; // to show that this is the initialization phase
|
|
}
|
|
|
|
public final Row row() {
|
|
return index0.row();
|
|
}
|
|
|
|
protected final void finishInitialization() {
|
|
if (index1 == null) {
|
|
// finish initialization phase
|
|
index0.sort();
|
|
index0.uniq();
|
|
index0.trim();
|
|
index1 = new RowSet(rowdef); //new RowSetArray(rowdef, spread);
|
|
}
|
|
}
|
|
|
|
public final synchronized byte[] smallestKey() {
|
|
final byte[] b0 = index0.smallestKey();
|
|
if (b0 == null) return null;
|
|
if (index1 == null) return b0;
|
|
final byte[] b1 = index0.smallestKey();
|
|
if (b1 == null || rowdef.objectOrder.compare(b1, b0) > 0) return b0;
|
|
return b1;
|
|
}
|
|
|
|
public final synchronized byte[] largestKey() {
|
|
final byte[] b0 = index0.largestKey();
|
|
if (b0 == null) return null;
|
|
if (index1 == null) return b0;
|
|
final byte[] b1 = index0.largestKey();
|
|
if (b1 == null || rowdef.objectOrder.compare(b0, b1) > 0) return b0;
|
|
return b1;
|
|
}
|
|
|
|
public final synchronized Row.Entry get(final byte[] key) {
|
|
assert (key != null);
|
|
finishInitialization();
|
|
assert index0.isSorted();
|
|
final Row.Entry indexentry = index0.get(key);
|
|
if (indexentry != null) return indexentry;
|
|
return index1.get(key);
|
|
}
|
|
|
|
public final synchronized boolean has(final byte[] key) {
|
|
assert (key != null);
|
|
finishInitialization();
|
|
assert index0.isSorted();
|
|
if (index0.has(key)) return true;
|
|
return index1.has(key);
|
|
}
|
|
|
|
public final synchronized Row.Entry replace(final Row.Entry entry) throws RowSpaceExceededException {
|
|
assert (entry != null);
|
|
finishInitialization();
|
|
// if the new entry is within the initialization part, just overwrite it
|
|
assert index0.isSorted();
|
|
final byte[] key = entry.getPrimaryKeyBytes();
|
|
if (index0.has(key)) {
|
|
// replace the entry
|
|
return index0.replace(entry);
|
|
}
|
|
// else place it in the index1
|
|
return index1.replace(entry);
|
|
}
|
|
|
|
/**
|
|
* Adds the row to the index. The row is identified by the primary key of the row.
|
|
* @param row a index row
|
|
* @return true if this set did _not_ already contain the given row.
|
|
* @throws IOException
|
|
* @throws RowSpaceExceededException
|
|
*/
|
|
public final boolean put(final Row.Entry entry) throws RowSpaceExceededException {
|
|
assert (entry != null);
|
|
if (entry == null) return true;
|
|
synchronized (this) {
|
|
finishInitialization();
|
|
// if the new entry is within the initialization part, just overwrite it
|
|
assert index0.isSorted();
|
|
final byte[] key = entry.getPrimaryKeyBytes();
|
|
if (index0.has(key)) {
|
|
// replace the entry
|
|
index0.put(entry);
|
|
return false;
|
|
}
|
|
// else place it in the index1
|
|
return index1.put(entry);
|
|
}
|
|
}
|
|
|
|
public final void addUnique(final Row.Entry entry) throws RowSpaceExceededException {
|
|
assert (entry != null);
|
|
if (entry == null) return;
|
|
synchronized (this) {
|
|
if (index1 == null) {
|
|
// we are in the initialization phase
|
|
index0.addUnique(entry);
|
|
return;
|
|
}
|
|
// initialization is over, add to secondary index
|
|
index1.addUnique(entry);
|
|
}
|
|
}
|
|
|
|
public final void addUnique(final List<Entry> rows) throws RowSpaceExceededException {
|
|
final Iterator<Entry> i = rows.iterator();
|
|
while (i.hasNext()) addUnique(i.next());
|
|
}
|
|
|
|
public final synchronized long inc(final byte[] key, final int col, final long add, final Row.Entry initrow) throws RowSpaceExceededException {
|
|
assert (key != null);
|
|
finishInitialization();
|
|
assert index0.isSorted();
|
|
final long l = index0.inc(key, col, add, null);
|
|
if (l != Long.MIN_VALUE) return l;
|
|
return index1.inc(key, col, add, initrow);
|
|
}
|
|
|
|
public final synchronized ArrayList<RowCollection> removeDoubles() throws RowSpaceExceededException {
|
|
// finish initialization phase explicitely
|
|
index0.sort();
|
|
if (index1 == null) {
|
|
return index0.removeDoubles();
|
|
}
|
|
final ArrayList<RowCollection> d0 = index0.removeDoubles();
|
|
final ArrayList<RowCollection> d1 = index1.removeDoubles();
|
|
d0.addAll(d1);
|
|
return d0;
|
|
}
|
|
|
|
public final synchronized boolean delete(final byte[] key) {
|
|
finishInitialization();
|
|
// if the new entry is within the initialization part, just delete it
|
|
boolean b = index0.delete(key);
|
|
if (b) {
|
|
assert index0.get(key) == null; // check if remove worked
|
|
return true;
|
|
}
|
|
// else remove it from the index1
|
|
b = index1.delete(key);
|
|
assert index1.get(key) == null : "removed " + ((b) ? " true" : " false") + ", and index entry still exists"; // check if remove worked
|
|
return b;
|
|
}
|
|
|
|
public final synchronized Row.Entry remove(final byte[] key) {
|
|
finishInitialization();
|
|
// if the new entry is within the initialization part, just delete it
|
|
int s = index0.size();
|
|
final Row.Entry indexentry = index0.remove(key);
|
|
if (indexentry != null) {
|
|
assert index0.size() < s: "s = " + s + ", index0.size() = " + index0.size();
|
|
assert index0.get(key) == null; // check if remove worked
|
|
return indexentry;
|
|
}
|
|
// else remove it from the index1
|
|
s = index1.size();
|
|
final Row.Entry removed = index1.remove(key);
|
|
assert removed == null || index1.size() < s: "s = " + s + ", index1.size() = " + index1.size();
|
|
assert index1.get(key) == null : "removed " + ((removed == null) ? " is null" : " is not null") + ", and index entry still exists"; // check if remove worked
|
|
return removed;
|
|
}
|
|
|
|
public final synchronized Row.Entry removeOne() {
|
|
if (index1 != null && !index1.isEmpty()) {
|
|
return index1.removeOne();
|
|
}
|
|
if (index0 != null && !index0.isEmpty()) {
|
|
return index0.removeOne();
|
|
}
|
|
return null;
|
|
}
|
|
|
|
public synchronized List<Row.Entry> top(int count) throws IOException {
|
|
List<Row.Entry> list = new ArrayList<Row.Entry>();
|
|
List<Row.Entry> list0 = index1.top(count);
|
|
list.addAll(list0);
|
|
list0 = index0.top(count - list.size());
|
|
list.addAll(list0);
|
|
return list;
|
|
}
|
|
|
|
public long mem() {
|
|
if (index0 != null && index1 == null) {
|
|
return index0.mem();
|
|
}
|
|
if (index0 == null && index1 != null) {
|
|
return index1.mem();
|
|
}
|
|
assert (index0 != null && index1 != null);
|
|
return index0.mem() + index1.mem();
|
|
}
|
|
|
|
public final synchronized int size() {
|
|
if (index0 != null && index1 == null) {
|
|
return index0.size();
|
|
}
|
|
if (index0 == null && index1 != null) {
|
|
return index1.size();
|
|
}
|
|
assert (index0 != null && index1 != null);
|
|
return index0.size() + index1.size();
|
|
}
|
|
|
|
public final synchronized boolean isEmpty() {
|
|
if (index0 != null && index1 == null) {
|
|
return index0.isEmpty();
|
|
}
|
|
if (index0 == null && index1 != null) {
|
|
return index1.isEmpty();
|
|
}
|
|
assert (index0 != null && index1 != null);
|
|
if (!index0.isEmpty()) return false;
|
|
if (!index1.isEmpty()) return false;
|
|
return true;
|
|
}
|
|
|
|
|
|
public final synchronized CloneableIterator<byte[]> keys(final boolean up, final byte[] firstKey) {
|
|
// returns the key-iterator of the underlying kelondroIndex
|
|
if (index1 == null) {
|
|
// finish initialization phase
|
|
index0.sort();
|
|
index0.uniq();
|
|
index1 = new RowSet(rowdef); //new RowSetArray(rowdef, spread);
|
|
return index0.keys(up, firstKey);
|
|
}
|
|
assert (index1 != null);
|
|
if (index0 == null) {
|
|
//assert consistencyAnalysis0() : "consistency problem: " + consistencyAnalysis();
|
|
return index1.keys(up, firstKey);
|
|
}
|
|
// index0 should be sorted
|
|
// sort index1 to enable working of the merge iterator
|
|
//assert consistencyAnalysis0() : "consistency problem: " + consistencyAnalysis();
|
|
final CloneableIterator<byte[]> k0 = index0.keys(up, firstKey);
|
|
final CloneableIterator<byte[]> k1 = index1.keys(up, firstKey);
|
|
if (k0 == null) return k1;
|
|
if (k1 == null) return k0;
|
|
return new MergeIterator<byte[]>(
|
|
k0,
|
|
k1,
|
|
rowdef.objectOrder,
|
|
MergeIterator.simpleMerge,
|
|
true);
|
|
}
|
|
|
|
public final synchronized CloneableIterator<Row.Entry> rows(final boolean up, final byte[] firstKey) {
|
|
// returns the row-iterator of the underlying kelondroIndex
|
|
if (index1 == null) {
|
|
// finish initialization phase
|
|
index0.sort();
|
|
index0.uniq();
|
|
index1 = new RowSet(rowdef); //new RowSetArray(rowdef, spread);
|
|
return index0.rows(up, firstKey);
|
|
}
|
|
assert (index1 != null);
|
|
if (index0 == null) {
|
|
//assert consistencyAnalysis0() : "consistency problem: " + consistencyAnalysis();
|
|
return index1.rows(up, firstKey);
|
|
}
|
|
// index0 should be sorted
|
|
// sort index1 to enable working of the merge iterator
|
|
//index1.sort();
|
|
//assert consistencyAnalysis0() : "consistency problem: " + consistencyAnalysis();
|
|
final CloneableIterator<Row.Entry> k0 = index0.rows(up, firstKey);
|
|
final CloneableIterator<Row.Entry> k1 = index1.rows(up, firstKey);
|
|
if (k0 == null) return k1;
|
|
if (k1 == null) return k0;
|
|
return new MergeIterator<Row.Entry>(
|
|
k0,
|
|
k1,
|
|
entryComparator,
|
|
MergeIterator.simpleMerge,
|
|
true);
|
|
}
|
|
|
|
public final Iterator<Entry> iterator() {
|
|
return rows();
|
|
}
|
|
|
|
public final synchronized CloneableIterator<Row.Entry> rows() {
|
|
// returns the row-iterator of the underlying kelondroIndex
|
|
if (index1 == null) {
|
|
// finish initialization phase
|
|
index0.sort();
|
|
index0.uniq();
|
|
index1 = new RowSet(rowdef); //new RowSetArray(rowdef, spread);
|
|
return index0.rows();
|
|
}
|
|
assert (index1 != null);
|
|
if (index0 == null) {
|
|
//assert consistencyAnalysis0() : "consistency problem: " + consistencyAnalysis();
|
|
return index1.rows();
|
|
}
|
|
// index0 should be sorted
|
|
// sort index1 to enable working of the merge iterator
|
|
//index1.sort();
|
|
//assert consistencyAnalysis0() : "consistency problem: " + consistencyAnalysis();
|
|
return new StackIterator<Row.Entry>(index0.rows(), index1.rows());
|
|
}
|
|
|
|
public final synchronized void close() {
|
|
if (index0 != null) index0.close();
|
|
if (index1 != null) index1.close();
|
|
objectTracker.remove(this.name);
|
|
}
|
|
|
|
public final String filename() {
|
|
return null; // this does not have a file name
|
|
}
|
|
|
|
public final void deleteOnExit() {
|
|
// do nothing, there is no file
|
|
}
|
|
|
|
}
|