Tighten up local packed object representation during packing

Rather than making a loader, and then using that to fill the object
representation, parse the header and set up our data directly.
This saves some time, as we don't waste cycles on information we
won't use right now.

The weight computed for a representation is now its actual stored
size in the pack file, rather than its inflated size.  This accounts
for changes made when the compression level is modified on the
repository.  It is however more costly to determine the weight of
the object, since we have to find its length in the pack.  To try and
recover that cost we now cache the length as part of our ObjectToPack
record, so it doesn't have to be found during the output phase.

A LocalObjectToPack now costs us (assuming 32 bit pointers):

                   (32 bit)     (64 bit)
  vm header:         8 bytes      8 bytes
  ObjectId:         20 bytes     20 bytes
  PackedObjectInfo: 12 bytes     12 bytes
  ObjectToPack:      8 bytes     12 bytes
  LocalOTP:         20 bytes     24 bytes
                 -----------    ---------
                    68 bytes     74 bytes

Change-Id: I923d2736186eb2ac8ab498d3eb137e17930fcb50
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>

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

@@ -49,35 +49,70 @@ import org.eclipse.jgit.lib.ObjectId;
 import org.eclipse.jgit.lib.StoredObjectRepresentation;
 
 class LocalObjectRepresentation extends StoredObjectRepresentation {
-	final PackedObjectLoader ldr;
+	static LocalObjectRepresentation newWhole(PackFile f, long p, long length) {
+		LocalObjectRepresentation r = new LocalObjectRepresentation() {
+			@Override
+			public int getFormat() {
+				return PACK_WHOLE;
+			}
+		};
+		r.pack = f;
+		r.offset = p;
+		r.length = length;
+		return r;
+	}
 
-	LocalObjectRepresentation(PackedObjectLoader ldr) {
+	static LocalObjectRepresentation newDelta(PackFile f, long p, long n,
-		this.ldr = ldr;
+			ObjectId base) {
+		LocalObjectRepresentation r = new Delta();
+		r.pack = f;
+		r.offset = p;
+		r.length = n;
+		r.baseId = base;
+		return r;
 	}
 
-	@Override
+	static LocalObjectRepresentation newDelta(PackFile f, long p, long n,
-	public int getFormat() {
+			long base) {
-		if (ldr instanceof DeltaPackedObjectLoader)
+		LocalObjectRepresentation r = new Delta();
-			return PACK_DELTA;
+		r.pack = f;
-		if (ldr instanceof WholePackedObjectLoader)
+		r.offset = p;
-			return PACK_WHOLE;
+		r.length = n;
-		return FORMAT_OTHER;
+		r.baseOffset = base;
+		return r;
 	}
 
+	PackFile pack;
+
+	long offset;
+
+	long length;
+
+	private long baseOffset;
+
+	private ObjectId baseId;
+
 	@Override
 	public int getWeight() {
-		long sz = ldr.getRawSize();
+		return (int) Math.min(length, Integer.MAX_VALUE);
-		if (Integer.MAX_VALUE < sz)
-			return WEIGHT_UNKNOWN;
-		return (int) sz;
 	}
 
 	@Override
 	public ObjectId getDeltaBase() {
-		try {
+		if (baseId == null && getFormat() == PACK_DELTA) {
-			return ldr.getDeltaBase();
+			try {
-		} catch (IOException e) {
+				baseId = pack.findObjectForOffset(baseOffset);
-			return null;
+			} catch (IOException error) {
+				return null;
+			}
+		}
+		return baseId;
+	}
+
+	private static final class Delta extends LocalObjectRepresentation {
+		@Override
+		public int getFormat() {
+			return PACK_DELTA;
 		}
 	}
 }

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

@@ -50,10 +50,13 @@ import org.eclipse.jgit.revwalk.RevObject;
 /** {@link ObjectToPack} for {@link ObjectDirectory}. */
 class LocalObjectToPack extends ObjectToPack {
 	/** Pack to reuse compressed data from, otherwise null. */
-	PackFile copyFromPack;
+	PackFile pack;
 
-	/** Offset of the object's header in {@link #copyFromPack}. */
+	/** Offset of the object's header in {@link #pack}. */
-	long copyOffset;
+	long offset;
+
+	/** Length of the data section of the object. */
+	long length;
 
 	LocalObjectToPack(RevObject obj) {
 		super(obj);
@@ -61,8 +64,9 @@ class LocalObjectToPack extends ObjectToPack {
 
 	@Override
 	public void select(StoredObjectRepresentation ref) {
-		LocalObjectRepresentation ptr = (LocalObjectRepresentation)ref;
+		LocalObjectRepresentation ptr = (LocalObjectRepresentation) ref;
-		this.copyFromPack = ptr.ldr.pack;
+		this.pack = ptr.pack;
-		this.copyOffset = ptr.ldr.objectOffset;
+		this.offset = ptr.offset;
+		this.length = ptr.length;
 	}
 }

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

@@ -299,9 +299,9 @@ public class ObjectDirectory extends FileObjectDatabase {
 		SEARCH: for (;;) {
 			for (final PackFile p : pList.packs) {
 				try {
-					PackedObjectLoader ldr = p.get(curs, otp);
+					LocalObjectRepresentation rep = p.representation(curs, otp);
-					if (ldr != null)
+					if (rep != null)
-						packer.select(otp, new LocalObjectRepresentation(ldr));
+						packer.select(otp, rep);
 				} catch (PackMismatchException e) {
 					// Pack was modified; refresh the entire pack list.
 					//

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

@@ -309,7 +309,7 @@ public class PackFile implements Iterable<PackIndex.MutableEntry> {
 
 		// Rip apart the header so we can discover the size.
 		//
-		readFully(src.copyOffset, buf, 0, 20, curs);
+		readFully(src.offset, buf, 0, 20, curs);
 		int c = buf[0] & 0xff;
 		final int typeCode = (c >> 4) & 7;
 		long inflatedLength = c & 15;
@@ -331,7 +331,7 @@ public class PackFile implements Iterable<PackIndex.MutableEntry> {
 			crc1.update(buf, 0, headerCnt);
 			crc2.update(buf, 0, headerCnt);
 
-			readFully(src.copyOffset + headerCnt, buf, 0, 20, curs);
+			readFully(src.offset + headerCnt, buf, 0, 20, curs);
 			crc1.update(buf, 0, 20);
 			crc2.update(buf, 0, headerCnt);
 			headerCnt += 20;
@@ -340,8 +340,8 @@ public class PackFile implements Iterable<PackIndex.MutableEntry> {
 			crc2.update(buf, 0, headerCnt);
 		}
 
-		final long dataOffset = src.copyOffset + headerCnt;
+		final long dataOffset = src.offset + headerCnt;
-		final long dataLength;
+		final long dataLength = src.length;
 		final long expectedCRC;
 		final ByteArrayWindow quickCopy;
 
@@ -349,7 +349,6 @@ public class PackFile implements Iterable<PackIndex.MutableEntry> {
 		// we report it missing instead.
 		//
 		try {
-			dataLength = findEndOffset(src.copyOffset) - dataOffset;
 			quickCopy = curs.quickCopy(this, dataOffset, dataLength);
 
 			if (idx().hasCRC32Support()) {
@@ -370,10 +369,10 @@ public class PackFile implements Iterable<PackIndex.MutableEntry> {
 					}
 				}
 				if (crc1.getValue() != expectedCRC) {
-					setCorrupt(src.copyOffset);
+					setCorrupt(src.offset);
 					throw new CorruptObjectException(MessageFormat.format(
 							JGitText.get().objectAtHasBadZlibStream,
-							src.copyOffset, getPackFile()));
+							src.offset, getPackFile()));
 				}
 			} else {
 				// We don't have a CRC32 code in the index, so compute it
@@ -399,20 +398,20 @@ public class PackFile implements Iterable<PackIndex.MutableEntry> {
 					}
 				}
 				if (!inf.finished() || inf.getBytesRead() != dataLength) {
-					setCorrupt(src.copyOffset);
+					setCorrupt(src.offset);
 					throw new EOFException(MessageFormat.format(
 							JGitText.get().shortCompressedStreamAt,
-							src.copyOffset));
+							src.offset));
 				}
 				expectedCRC = crc1.getValue();
 			}
 		} catch (DataFormatException dataFormat) {
-			setCorrupt(src.copyOffset);
+			setCorrupt(src.offset);
 
 			CorruptObjectException corruptObject = new CorruptObjectException(
 					MessageFormat.format(
 							JGitText.get().objectAtHasBadZlibStream,
-							src.copyOffset, getPackFile()));
+							src.offset, getPackFile()));
 			corruptObject.initCause(dataFormat);
 
 			StoredObjectRepresentationNotAvailableException gone;
@@ -458,7 +457,7 @@ public class PackFile implements Iterable<PackIndex.MutableEntry> {
 			}
 			if (crc2.getValue() != expectedCRC) {
 				throw new CorruptObjectException(MessageFormat.format(JGitText
-						.get().objectAtHasBadZlibStream, src.copyOffset,
+						.get().objectAtHasBadZlibStream, src.offset,
 						getPackFile()));
 			}
 		}
@@ -661,6 +660,55 @@ public class PackFile implements Iterable<PackIndex.MutableEntry> {
 		}
 	}
 
+	LocalObjectRepresentation representation(final WindowCursor curs,
+			final AnyObjectId objectId) throws IOException {
+		final long pos = idx().findOffset(objectId);
+		if (pos < 0)
+			return null;
+
+		final byte[] ib = curs.tempId;
+		readFully(pos, ib, 0, 20, curs);
+		int c = ib[0] & 0xff;
+		int p = 1;
+		final int typeCode = (c >> 4) & 7;
+		while ((c & 0x80) != 0)
+			c = ib[p++] & 0xff;
+
+		long len = (findEndOffset(pos) - pos);
+		switch (typeCode) {
+		case Constants.OBJ_COMMIT:
+		case Constants.OBJ_TREE:
+		case Constants.OBJ_BLOB:
+		case Constants.OBJ_TAG:
+			return LocalObjectRepresentation.newWhole(this, pos, len - p);
+
+		case Constants.OBJ_OFS_DELTA: {
+			c = ib[p++] & 0xff;
+			long ofs = c & 127;
+			while ((c & 128) != 0) {
+				ofs += 1;
+				c = ib[p++] & 0xff;
+				ofs <<= 7;
+				ofs += (c & 127);
+			}
+			ofs = pos - ofs;
+			return LocalObjectRepresentation.newDelta(this, pos, len - p, ofs);
+		}
+
+		case Constants.OBJ_REF_DELTA: {
+			len -= p;
+			len -= Constants.OBJECT_ID_LENGTH;
+			readFully(pos + p, ib, 0, 20, curs);
+			ObjectId id = ObjectId.fromRaw(ib);
+			return LocalObjectRepresentation.newDelta(this, pos, len, id);
+		}
+
+		default:
+			throw new IOException(MessageFormat.format(
+					JGitText.get().unknownObjectType, typeCode));
+		}
+	}
+
 	private long findEndOffset(final long startOffset)
 			throws IOException, CorruptObjectException {
 		final long maxOffset = length - 20;

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

@@ -103,7 +103,7 @@ final class WindowCursor extends ObjectReader implements ObjectReuseAsIs {
 	public void copyObjectAsIs(PackOutputStream out, ObjectToPack otp)
 			throws IOException, StoredObjectRepresentationNotAvailableException {
 		LocalObjectToPack src = (LocalObjectToPack) otp;
-		src.copyFromPack.copyAsIs(out, src, this);
+		src.pack.copyAsIs(out, src, this);
 	}
 
 	/**