Remember loose objects and fast-track their lookup

Recently created objects are usually what branches point to, and
are usually written out as loose objects.  But due to the high cost
of asking the operating system if a file exists, these are the last
thing that ObjectDirectory examines when looking for an object by
its ObjectId.

Caching recently seen loose objects permits the opening code to
jump directly to the loose object, accelerating lookup for branch
heads that are accessed often.

To avoid exploding the cache its limited to approximately 2048
entries.  When more ids are added, the table is simply cleared
and reset in size.

Change-Id: I18f483217412b102f754ffd496c87061d592e535
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>

org.eclipse.jgit/src/org/eclipse/jgit/storage/file/ObjectDirectory.java

@@ -117,6 +117,8 @@ public class ObjectDirectory extends FileObjectDatabase {
 
 	private final AtomicReference<AlternateHandle[]> alternates;
 
+	private final UnpackedObjectCache unpackedObjectCache;
+
 	/**
 	 * Initialize a reference to an on-disk object directory.
 	 *
@@ -140,6 +142,7 @@ public class ObjectDirectory extends FileObjectDatabase {
 		packDirectory = new File(objects, "pack");
 		alternatesFile = new File(infoDirectory, "alternates");
 		packList = new AtomicReference<PackList>(NO_PACKS);
+		unpackedObjectCache = new UnpackedObjectCache();
 		this.fs = fs;
 
 		alternates = new AtomicReference<AlternateHandle[]>();
@@ -179,6 +182,8 @@ public class ObjectDirectory extends FileObjectDatabase {
 
 	@Override
 	public void close() {
+		unpackedObjectCache.clear();
+
 		final PackList packs = packList.get();
 		packList.set(NO_PACKS);
 		for (final PackFile p : packs.packs)
@@ -255,6 +260,8 @@ public class ObjectDirectory extends FileObjectDatabase {
 	}
 
 	boolean hasObject1(final AnyObjectId objectId) {
+		if (unpackedObjectCache.isUnpacked(objectId))
+			return true;
 		for (final PackFile p : packList.get().packs) {
 			try {
 				if (p.hasObject(objectId)) {
@@ -328,6 +335,14 @@ public class ObjectDirectory extends FileObjectDatabase {
 
 	ObjectLoader openObject1(final WindowCursor curs,
 			final AnyObjectId objectId) throws IOException {
+		if (unpackedObjectCache.isUnpacked(objectId)) {
+			ObjectLoader ldr = openObject2(curs, objectId.name(), objectId);
+			if (ldr != null)
+				return ldr;
+			else
+				unpackedObjectCache.remove(objectId);
+		}
+
 		PackList pList = packList.get();
 		SEARCH: for (;;) {
 			for (final PackFile p : pList.packs) {
@@ -429,15 +444,21 @@ public class ObjectDirectory extends FileObjectDatabase {
 			File path = fileFor(objectName);
 			FileInputStream in = new FileInputStream(path);
 			try {
+				unpackedObjectCache.add(objectId);
 				return UnpackedObject.open(in, path, objectId, curs);
 			} finally {
 				in.close();
 			}
 		} catch (FileNotFoundException noFile) {
+			unpackedObjectCache.remove(objectId);
 			return null;
 		}
 	}
 
+	void addUnpackedObject(ObjectId id) {
+		unpackedObjectCache.add(id);
+	}
+
 	boolean tryAgain1() {
 		final PackList old = packList.get();
 		if (old.tryAgain(packDirectory.lastModified()))

org.eclipse.jgit/src/org/eclipse/jgit/storage/file/ObjectDirectoryInserter.java

@@ -91,16 +91,20 @@ class ObjectDirectoryInserter extends ObjectInserter {
 		}
 
 		final File dst = db.fileFor(id);
-		if (tmp.renameTo(dst))
+		if (tmp.renameTo(dst)) {
+			db.addUnpackedObject(id);
 			return id;
+		}
 
 		// Maybe the directory doesn't exist yet as the object
 		// directories are always lazily created. Note that we
 		// try the rename first as the directory likely does exist.
 		//
 		dst.getParentFile().mkdir();
-		if (tmp.renameTo(dst))
+		if (tmp.renameTo(dst)) {
+			db.addUnpackedObject(id);
 			return id;
+		}
 
 		if (db.has(id)) {
 			tmp.delete();

org.eclipse.jgit/src/org/eclipse/jgit/storage/file/UnpackedObjectCache.java

@@ -0,0 +1,149 @@
+/*
+ * Copyright (C) 2010, Google Inc.
+ * and other copyright owners as documented in the project's IP log.
+ *
+ * This program and the accompanying materials are made available
+ * under the terms of the Eclipse Distribution License v1.0 which
+ * accompanies this distribution, is reproduced below, and is
+ * available at http://www.eclipse.org/org/documents/edl-v10.php
+ *
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright
+ *   notice, this list of conditions and the following disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ *   copyright notice, this list of conditions and the following
+ *   disclaimer in the documentation and/or other materials provided
+ *   with the distribution.
+ *
+ * - Neither the name of the Eclipse Foundation, Inc. nor the
+ *   names of its contributors may be used to endorse or promote
+ *   products derived from this software without specific prior
+ *   written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
+ * CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
+ * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
+ * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+ * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+ * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+ * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+ * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
+ * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+package org.eclipse.jgit.storage.file;
+
+import java.util.concurrent.atomic.AtomicReferenceArray;
+
+import org.eclipse.jgit.lib.AnyObjectId;
+import org.eclipse.jgit.lib.ObjectId;
+
+/** Remembers objects that are currently unpacked. */
+class UnpackedObjectCache {
+	private static final int INITIAL_BITS = 5; // size = 32
+
+	private static final int MAX_BITS = 11; // size = 2048
+
+	private volatile Table table;
+
+	UnpackedObjectCache() {
+		table = new Table(INITIAL_BITS);
+	}
+
+	boolean isUnpacked(AnyObjectId objectId) {
+		return table.contains(objectId);
+	}
+
+	void add(AnyObjectId objectId) {
+		Table t = table;
+		if (t.add(objectId)) {
+			// The object either already exists in the table, or was
+			// successfully added. Either way leave the table alone.
+			//
+		} else {
+			// The object won't fit into the table. Implement a crude
+			// cache removal by just dropping the table away, but double
+			// it in size for the next incarnation.
+			//
+			Table n = new Table(Math.min(t.bits + 1, MAX_BITS));
+			n.add(objectId);
+			table = n;
+		}
+	}
+
+	void remove(AnyObjectId objectId) {
+		if (isUnpacked(objectId))
+			clear();
+	}
+
+	void clear() {
+		table = new Table(INITIAL_BITS);
+	}
+
+	private static class Table {
+		private static final int MAX_CHAIN = 8;
+
+		private final AtomicReferenceArray<ObjectId> ids;
+
+		private final int shift;
+
+		final int bits;
+
+		Table(int bits) {
+			this.ids = new AtomicReferenceArray<ObjectId>(1 << bits);
+			this.shift = 32 - bits;
+			this.bits = bits;
+		}
+
+		boolean contains(AnyObjectId toFind) {
+			int i = index(toFind);
+			for (int n = 0; n < MAX_CHAIN; n++) {
+				ObjectId obj = ids.get(i);
+				if (obj == null)
+					break;
+
+				if (AnyObjectId.equals(obj, toFind))
+					return true;
+
+				if (++i == ids.length())
+					i = 0;
+			}
+			return false;
+		}
+
+		boolean add(AnyObjectId toAdd) {
+			int i = index(toAdd);
+			for (int n = 0; n < MAX_CHAIN;) {
+				ObjectId obj = ids.get(i);
+				if (obj == null) {
+					if (ids.compareAndSet(i, null, toAdd.copy()))
+						return true;
+					else
+						continue;
+				}
+
+				if (AnyObjectId.equals(obj, toAdd))
+					return true;
+
+				if (++i == ids.length())
+					i = 0;
+				n++;
+			}
+			return false;
+		}
+
+		private int index(AnyObjectId id) {
+			return id.hashCode() >>> shift;
+		}
+	}
+}