By doing the release at the higher level class, we can ensure
the release occurs if the inserter was allocated, even if the
implementation forgets to do this. Since the higher level class
is what allocated it, it makes sense to have it also do the release.
Change-Id: Id617b2db864c3208ed68cba4eda80e51612359ad
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Everyone else does. This must have been a spot I missed during
some sort of squash while developing the series.
Change-Id: I62eae50b618f47ee33ad7cf71fc05b724f603201
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Similar to what we did with the file code, move the pack writer
into its own package so the related classes and their package
private methods are hidden from the rest of the library.
Change-Id: Ic1b5c7c8c8d266e90c910d8d68dfc8e93586854f
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
We no longer need an ObjectLoader to be lazy and try to delay
the materialization of the object content. That was done only
to support PackWriter searching for a good reuse candidate.
Instead, simplify the code base by doing the materialization
immediately when the loader asks for it, because any caller
asking for the loader is going to need the content.
Change-Id: Id867b1004529744f234ab8f9cfab3d2c52ca3bd0
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
These were only used by PackWriter to help it filter object
representations. Their only user disappeared when we rewrote the
object selection code path to use the new representation type.
Change-Id: I9ed676bfe4f87fcf94aa21e53bda43115912e145
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
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>
This move isolates all of the local file specific implementation code
into a single package, where their package-private methods and support
classes are properly hidden away from the rest of the core library.
Because of the sheer number of files impacted, I have limited this
change to only the renames and the updated imports.
Change-Id: Icca4884e1a418f83f8b617d0c4c78b73d8a4bd17
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Objects that fall completely within a single window can be worked
with in a zero-copy fashion, provided that the window is backed by
a normal byte[] and not by a ByteBuffer.
This works for a surprising number of objects. The default window
size is 8 KiB, but most deltas are quite a bit smaller than that.
Objects smaller than 1/2 of the window size have a very good chance
of falling completely within a window's array, which means we can
work with them without copying their data around.
Larger objects, or objects which are unlucky enough to span over a
window boundary, get copied through the temporary buffer. We pay
a tiny penalty to realize we can't use the zero-copy code path,
but its easier than trying to keep track of two adjacent windows.
With this change (as well as everything preceeding it), packing
is actually a bit faster. Some crude benchmarks based on cloning
linux-2.6.git (~324 MiB, 1,624,785 objects) over localhost using
C git client and JGit daemon shows we get better throughput, and
slightly better times:
Total Time | Throughput
(old) (now) | (old) (now)
--------------+---------------------------
2m45s 2m37s | 12.49 MiB/s 21.17 MiB/s
2m42s 2m36s | 16.29 MiB/s 22.63 MiB/s
2m37s 2m31s | 16.07 MiB/s 21.92 MiB/s
Change-Id: I48b2c8d37f08d7bf5e76c5a8020cde4a16ae3396
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Output of selected reuses is refactored to use a new ObjectReuseAsIs
interface that extends the ObjectReader. This interface allows the
reader to control how it performs the reuse into the output stream,
but also allows it to throw an exception to request the writer to
find a different candidate representation.
The PackFile reuse code was overhauled, cleaning up the APIs so they
aren't exposed in the object loader, but instead are now a single
method on the PackFile itself. The reuse algorithm was changed to do
a data verification pass, followed by the copy pass to the output.
This permits us to work around a corrupt object in a pack file by
seeking another copy of that object when this one is bad.
The reuse code was also optimized for the common case, where the
in-pack representation is under 16 KiB. In these smaller cases
data is sent to the pack writer more directly, avoiding some copying.
Change-Id: I6350c2b444118305e8446ce1dfd049259832bcca
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
The new selection implementation uses a public API on the
ObjectReader, allowing the storage library to enumerate its
candidates and select the best one for this packer without
needing to build a temporary list of the candidates first.
Change-Id: Ie01496434f7d3581d6d3bbb9e33c8f9fa649b6cd
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Make the lower bits available for flags that PackWriter can use to
keep track of facts about the object. We shouldn't need more than
2^24 delta depths, unpacking that chain is unfathomable anyway.
This change gets us 4 bits that are unused in the lower end of the
word, which are typically easier to load from Java and most machine
instruction sets. We can use these in later changes.
Change-Id: Ib9e11221b5bca17c8a531e4ed130ba14c0e3744f
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
The ObjectReader class is dual-purposed into being a factory for the
ObjectToPack, permitting specific ObjectDatabase implementations
to override the method and offer their own custom subclass of the
generic ObjectToPack class. By allowing them to directly extend the
type, each implementation can add custom fields to support tracking
where an object is stored, without incurring any additional penalties
like a parallel Map<ObjectId,Object> would cost.
The reader was chosen to act as a factory rather than the database,
as the reader will eventually be tied more tightly with the
ObjectWalk and TreeWalk. During object enumeration the reader
would have had to load the object for the RevWalk, and may chose
to cache object position data internally so it can later be reused
and fed into the ObjectToPack instance supplied to the PackWriter.
Since a reader is not thread-safe, and is scoped to this PackWriter
and its internal ObjectWalk, its a great place for the database to
perform caching, if any.
Right now this change goes a bit backwards by changing what should
be generic ObjectToPack references inside of PackWriter to the very
PackFile specific LocalObjectToPack subclass. We will correct these
in a later commit as we start to refine what the ObjectToPack API
will eventually look like in order to better support the PackWriter.
Change-Id: I9f047d26b97e46dee3bc0ccb4060bbebedbe8ea9
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
This shortens the implementation within PackWriter, and starts
to open the door for some other refactorings based on changing
the ObjectToPack to be a public part of the API.
Change-Id: Id849cbffc4de20b903e844a2de7737eeb8b7a3ff
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Only the ObjectDirectory type of database knows where to find the
objects directory on the local filesystem, so defer to it whenever
we need to know where the objects reside. Since this is the type
returned by FileRepository's getObjectDatabase() method, we mostly
don't have to do much other than use a slightly longer invocation.
Change-Id: Ie5f58132a6411b56c3acad73646ad169d78a0654
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Some types of repositories might not be stored on local disk. For
these, they will most likely return null for getDirectory() as the
java.io.File type cannot describe where their storage is, its not
in the host's filesystem.
Document that getDirectory() can return null now, and update all
current non-test callers in JGit that might run into problems on
such repositories. For the most part, just act like its bare.
Change-Id: I061236a691372a267fd7d41f0550650e165d2066
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Replace the old crude event listener system with a much more generic
implementation, patterned after the event dispatch techniques used
in Google Web Toolkit 1.5 and later.
Each event delivers to an interface that defines a single method,
and the event itself is what performs the delivery in a type-safe
way through its own dispatch method.
Listeners are registered in a generic listener list, indexed by
the interface they implement and wish to receive an event for.
Delivery of events is performed by looping through all listeners
implementing the event's corresponding listener interface, and using
the event's own dispatch method to deliver the event. This is the
classical "double dispatch" pattern for event delivery.
Listeners can be unregistered by invoking remove() on their
registration handle. This change therefore requires application
code to track the handle if it wishes to remove the listener at a
later point in time.
Event delivery is now exposed as a generic public method on the
Repository class, making it easier for any type of message to
be sent out to any type of listener that has registered, without
needing to pre-arrange for type-safe fireFoo() methods.
New event types can be added in the future simply by defining a
new RepositoryEvent subclass and a corresponding RepositoryListener
interface that it dispatches to. By always adding new events through
a new interface, we never need to worry about defining an Adapter
to provide default no-op implementations of new event methods.
Change-Id: I651417b3098b9afc93d91085e9f0b2265df8fc81
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
This better matches with the name used in the environment
(GIT_WORK_TREE), in the configuration file (core.worktree),
and in our builder object.
Since we are already breaking a good chunk of other code
related to repository access, and this fairly easy to fix
in an application's code base, I'm not going to offer the
wrapper getWorkDir() method.
Change-Id: Ib698ba4bbc213c48114f342378cecfe377e37bb7
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
The new FileRepositoryBuilder class helps applications to construct
a properly configured FileRepository, with properties assumed based
upon the standard Git rules for the local filesystem.
To better support simple command line applications, environment
variable handling and repository searching was moved into this
builder class.
The change gets rid of the ever-growing FileRepository constructor
variants, and the multitude of java.io.File typed parameters,
by using simple named setter methods.
Change-Id: I17e8e0392ad1dbf6a90a7eb49a6d809388d27e4c
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Not every type of Repository will be able to map an ObjectId into
a local file system path that stores that object's file contents.
Heck, its not even true for the FileRepository, as an object can
be stored in a pack file and not in its loose format.
Remove this from our public API, it was a mistake to publish it.
Change-Id: I20d1b8c39104023936e6d46a5b0d7ef39ff118e8
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Rather than relying on the repository's ability to give us the
local file path for a loose object, just pass its inflated form to
the ObjectInserter for the repository. We have to recompress it,
which may slow down fetches, but this is the slow dumb protocol.
The extra cost to do the compression locally isn't going to be a
major bottleneck.
This nicely removes the nasty part about computing the object
identity by hand, allowing us to instead rely upon the inserter's
internal computation. Unfortunately it means we might store a loose
object whose SHA-1 doesn't match the expected SHA-1, such as if the
remote repository was corrupted. This is fairly harmless, as the
incorrectly named object will now be stored under its proper name,
and will eventually be garbage collected, as its not referenced by
the local repository.
We have to flush the inserter after the object is stored because
we aren't sure if we need to read the object later, or not.
Change-Id: Idb1e2b1af1433a23f8c3fd55aeb20575e6047ef0
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
The WindowCache is an implementation detail of PackFile and how its
used by ObjectDirectory. Lets start to hide it and replace the public
API with a more generic concept, ObjectReader.
Because PackedObjectLoader is also considered a private detail of
PackFile, we have to make PackWriter temporarily dependent upon the
WindowCursor and thus FileRepository and ObjectDirectory in order to
just start the refactoring. In later changes we will clean up the
APIs more, exposing sufficient support to PackWriter without needing
the file specific implementation details.
Change-Id: I676be12b57f3534f1285854ee5de1aa483895398
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Not every object storage system will have the concept of alternate
object databases to search, and even if they do, they may not have
the notion of fast-access / slow-access split like we do within
the ObjectDirectory code for pack files and loose objects.
Push all of that down below the generic API so that it is a hidden
detail of the ObjectDirectory and its related supporting classes.
Change-Id: I54bc1ca5ff2ac94dfffad1f9a9dad7af202b9523
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Some newer style APIs are updated to use the newer ObjectInserter
interface instead of the now deprecated ObjectWriter. In many of
the unit tests we don't bother to release the inserter, these are
typically using the file backend which doesn't need a release,
but in the future should use an in-memory HashMap based store,
which really wouldn't need it either.
Change-Id: I91a15e1dc42da68e6715397814e30fbd87fa2e73
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
The ObjectInserter API permits ObjectDatabase implementations to
control their own object insertion behavior, rather than forcing
it to always be a new loose file created in the local filesystem.
Inserted objects can also be queued and written asynchronously to
the main application, such as by appending into a pack file that
is later closed and added to the repository.
This change also starts to open the door to non-file based object
storage, such as an in-memory HashMap for unit testing, or a more
complex system built on top of a distributed hash table.
To help existing application code port to the newer interface we
are keeping ObjectWriter as a delegation wrapper to the new API.
Each ObjectWriter instances holds a reference to an ObjectInserter
for the Repository's top-level ObjectDatabase, and it flushes and
releases that instance on each object processed.
Change-Id: I413224fb95563e7330c82748deb0aada4e0d6ace
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
A repository implementation might support storing configurations
on a non-file storage system, so widen the return value to be any
type of configuration.
Change-Id: If9a0928f4b3ef29a24d270b0ce585a6e77f6fac6
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
When the surrounding code is already heavily based upon the
assumption that we have a FileRepository (e.g. because it
created that type of repository) keep the type around and
use it directly. This permits us to continue to do things
like save the configuration file.
Change-Id: Ib783f0f6a11acd6aa305c16d61ccc368b46beecc
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
To support other storage models other than just the local filesystem,
we split the Repository class into a nearly abstract interface and
then create a concrete subclass called FileRepository with the file
based IO implementation.
We are using an abstract class for Repository rather than the much
more generic interface, as implementers will want to inherit a large
array of utility functions, such as resolve(String). Having these in
a base class makes it easy to inherit them.
This isn't the final home for lib.FileRepository. Future changes
will rename it into storage.file.FileRepository, but to do that we
need to also move a number of other related class, which we aren't
quite ready to do.
Change-Id: I1bd54ea0500337799a8e792874c272eb14d555f7
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
If the working tree isn't available, it doesn't make any sense to
obtain the merge heads, or the buffered commit message. The
repository shouldn't have a partial merge state to read. Throw back
the same exception we do when invoking getWorkDir() on a bare
repository instance.
Change-Id: I762c55890b7fe272a183da583f910671d1cadf71
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
This permits us to leave the implementation of these methods here in
the Repository class, but later refactor how the directory is accessed
into a subclass.
Change-Id: I5785b2009c5b7cca0fb070a968e50814ce847076
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
A bare repository cannot be checked out, committed to, etc. as it
doesn't have a working directory. Define this as a state since the
state enumeration exists only to describe how a working directory
can be modified.
Change-Id: I0a299013c6e42fef6cae3f6a9446f8f6c8e0514a
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
This better matches with the other configuration variable,
'userConfig', and helps to make it clear what config object
we are dealing with.
Change-Id: I2c585649aecc805e8e66db2f094828cd2649e549
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Change the Repository API to use straight-up FileBasedConfig.
This lets us remove the subclass RepositoryConfig and stop having
a specialized configuration type for repository, letting us instead
focus the config type heirarchy on type-of-storage rather than use.
Change-Id: I7236800e8090624453a89cb0c7a9a632702691c6
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Instead of using the internal field directly to access references
or objects, use the getter method to obtain the proper type of
database, and follow down from there. This permits us to later
do a refactoring that makes those methods abstract and strips the
field out of the Repository class, moving it into a concrete base
class that is more storage implementation specific.
Change-Id: Ic21dd48800e68a04ce372965ad233485b2a84bef
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
This method doesn't need to be synchronized, as its only a proxy to
create(boolean), which is the real worker. While we are touching
it try to improve the Javadoc and whitespace nearby.
Change-Id: Ibdddec6e518ca6d7439cfad90fedfcdc2d6b7a2e
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
This permits the repository implementation to know what its
alternates concept means, and avoids needing to expose finer details
about the ObjectDatabase to network code like the RefAdvertiser.
Change-Id: Ic6d173f300cb72de34519c7607cf7b0ff3ea6882
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
In the future getDirectory() can return null. Avoid an NPE here by
refactoring the code to support conditionally skipping a check for
the properties file in the repository directory, falling to only
the user's ~/ file location.
Change-Id: I76f5503d4063fdd9d24b7c1b58e1b09ddf1a5670
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
If the destination repository doesn't use an ObjectDirectory to
store its objects, we can't download to the object directory.
Instead pull the pack-*.idx files down to temporary files in the
JVM's default temporary directory.
Change-Id: Ied16bc89be624d87110ba42ba52d698a6ea7d982
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
upload-pack and receive-pack take a git repository as an argument,
but its a lenient path format. Locate the repository and open it.
Change-Id: I4b377e57b28ba3b1717c13d9ab51a602de1ad257
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
When reading or locking the index of a repository, we need to use
the index file specified by the repository, to ensure we correctly
honor what the repository was configured with.
Change-Id: I5be366ce32d7923b888dc01d19335912b01b7c4c
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Its not strictly required that we sort topologically in order to
produce a valid pack file. This was just something that Linus
thought would be a good idea to do. In practice its not that
important for most repositories. Local file IO quickly falls
out of the pattern that topological sorting provides any sort
of benefit for, so expending extra resources to enforce it when
we make a pack isn't really worth it.
I'm removing this sort in the pipeline because later changes
would support really efficient COMMIT_TIME_DESC sorting on a
non-file storage system, but TOPO sorting would be a bit more
ugly to run, due to the in-memory delays it imposes.
Change-Id: I0121453461c2140c6917cb10c6df584eb47e5795
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
If UploadPack invokes flush() on the output stream we pass it, its
most likely the progress messages coming down the side band stream.
As pack generation can take a while, we want to push that down
at the client as early as we can, to keep the connection alive,
and to let the user know we are still working on their behalf.
Ensure we dump the temporary buffer whenever flush() is invoked,
otherwise the messages don't get sent in a timely fashion to the
user agent (in this case, git fetch).
We specifically don't implement flush() for ReceivePack right now,
as that protocol currently does not provide progress messages to
the user, but it does invoke flush several times, as the different
streams include '0000' type flush-pkts to denote various end points.
Change-Id: I797c90a2c562a416223dc0704785f61ac64e0220
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
We want to eventually get rid of the mapCommit, mapTree APIs on
Repository and force everyone into the faster parsers that exist
in RevWalk. Rewriting resolve in terms of the faster parsers is
a good first step.
It actually simplifies the code a bit, as we no longer need to
keep track of an ObjectId and an Object (the parsed form), since
all RevObjects implicitly have their ObjectId readily available.
Change-Id: I4d234630195616e2c263e7e70038b55a1be4e7a3
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Instead of peeling things by hand in application level code, defer
the peeling logic into RevWalk's new peel utility method.
Change-Id: Idabd10dc41502e782f6a2eeb56f09566b97775a8
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
We already have these objects parsed and cached in our object pool.
We shouldn't be looking them up via the legacy mapObject API, but
instead can use the pool and the faster parsing routines available
through the RevWalk that we extend.
While we are here fixing the code, lets also correct the tag date
sorting to accept tags that have no tagger identity, because they
were created before Git knew how to store that field.
Change-Id: Id49a11f6d9c050c82b876e5e11058840c894b2d7
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Rather than relying on the helpers in RepositoryConfig to get
these objects, obtain them directly through the Config API.
Its only slightly more verbose, but permits us to work with the
base Config class, which is more flexible than the highly file
specific RepositoryConfig.
This is what I really meant to do when I added the section parser
and caching support to Config, we just failed to finish updating
all of the call sites.
Change-Id: I481cb365aa00bfa8c21e5ad0cd367ddd9c6c0edd
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
We don't have to assume/depend on RepositoryConfig here, these
two tests can use higher level versions of the class and still
come up with the same test. That frees us up to do some changes
to the RepositoryConfig API.
Change-Id: Ia7b263c8c5efa3fae1054416d39c546867288132
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Long ago we stopped supporting the core.legacyHeaders variable,
as JGit (like C Git) stopped creating the new pack-style loose
objects, rendering this variable pointless. The test is still
valid, it proves we write the standard loose object format for
a commit, but the variable assignment has no impact on the test
so drop it from the code.
Change-Id: I051336ada23033c05e86bbff73ae5d78a37b1640
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Under smart HTTP the biDirectionalPipe flag is false, and we return
back immediately at this point in the negotiation process. There is
no need to flush the stream to the client, the request is over and
it will be automatically flushed out by the higher level servlet
that invoked us. Avoiding flush here allows us to only use flush
after a progress message is sent during pack generation.
Change-Id: Id0c8b7e95e3be6ca4c1b479e096bed6b0283b828
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
This simplifies the PackIndex code, which is trying to quickly copy
an existing ObjectId into a MutableObjectId. Rather than having
the PackIndex violate the ObjectId's internals, expose a copy from
function similar to the other ones for copying from raw byte arrays
or hex formatted strings.
Change-Id: I142635cbece54af2ab83c58477961ce925dc8255
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Storage systems can use these implementations to compare a passed
AnyObjectId with a stored representation of an ObjectId in the
canonical network byte order format. This can be useful to do a
binary search, or just linear scan, over an encoded storage file.
Change-Id: I8c72993c4f4c6e98d599ac2c9867453752f25fd2
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Shawn O. Pearce