When working on a difference algorithm's implementation, its generally
more important to care about how it behaves on real-world inputs than
it does on fake inputs created for unit test cases. Run each
implementation against a number of real-world repositories, looking at
changes between files in each commit. This gives a better picture of
how a particular algorithm performs.
This test suite run against JGit and linux-2.6 with the current
available algorithms shows HistogramDiff always out-performs
MyersDiff, and by a wide margin on the linux-2.6 sources. As
HistogramDiff has similar output properties as PatienceDiff, the
resulting edits are probably also more human-readable. These test
results show that HistogramDiff is a good choice for the default
implementation, and also show that PatienceDiff isn't worth keeping.
jgit: start at baa83ae
2686 files, 760 commits
N= 3 min lines, 3016 max lines
Algorithm Time(ns) ( Time(ns) on Time(ns) on )
( N=3 N=3016 )
---------------------------------------------------------------------
histogram_myers 314652100 ( 3900 298100 )
histogram 315973000 ( 3800 302100 )
patience 774724900 ( 4500 347900 )
patience_histogram_myers 786332800 ( 3700 351200 )
myers 819359300 ( 4100 379100 )
patience_myers 843416700 ( 3800 348000 )
linux-2.6.git: start at 85a3318
4001 files, 2680 commits
N= 2 min lines, 39098 max lines
Algorithm Time(ns) ( Time(ns) on Time(ns) on )
( N=2 N=39098 )
---------------------------------------------------------------------
histogram_myers 1229870000 ( 5900 2642700 )
histogram 1235654100 ( 6000 2695400 )
patience 3856546000 ( 5900 2627700 )
patience_histogram_myers 3866728100 ( 7000 2624000 )
patience_myers 4004875300 ( 8000 2651700 )
myers 9794679000 ( 7200 2716200 )
Change-Id: I2502684d31f7851e720356820d04d8cf767f7229
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Large objects stored as deltas get unpacked by JGit into a loose
object, so they are cheaper to access later on. This unpacking was
broken because TeeInputStream copied the wrong length into the loose
object, sometimes copying too many bytes into the result. This
created a loose object that did not have the correct content, and
whose length did not match the length denoted in the object header.
Change-Id: I3ce1fd9f3dc5bd195249c7872b3bec49570424a2
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
When passing to a fallback algorithm, we can avoid creating a new copy
of the hash codes for each sequence by passing in the hashed sequences
directly. This makes it cheaper to switch from HistogramDiff down to
MyersDiff in a single pass.
Change-Id: Ibf2e81be57c083862eeb134279aed676653bf9b5
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
There is a corner case where we get an EMPTY region during recursion,
but we didn't expect to receive that. Its harmless to ignore the
region since the region is empty and has no content, so do so rather
than throwing an exception
Change-Id: I50dcec81ecba763072bb739adfab5879fb48b23a
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Certain inputs caused an infinite loop because the prior match data
couldn't be used as expected. Rather than incrementing the match
pointer before looking at an element, do it after, so the loop breaks
when we wrap around to the starting point.
Change-Id: Ieab28bb3485a914eeddc68aa38c256f255dd778c
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
The need for branching becomes more pressing with pull
support: we need to make sure the upstream configuration entries
are written correctly when creating and renaming branches
(and of course are cleaned up when deleting them).
This adds support for listing, adding, deleting and renaming
branches including the more common options.
Bug: 326938
Change-Id: I00bcc19476e835d6fd78fd188acde64946c1505c
Signed-off-by: Mathias Kinzler <mathias.kinzler@sap.com>
Signed-off-by: Chris Aniszczyk <caniszczyk@gmail.com>
Application code, including unit tests for storage implementations,
should not extend RevCommit outside of the scope of using it for a
RevWalk. Its a lot of overhead and unlikely to work long-term.
Instead for the one test that matters, use a custom subclass of the
ObjectId type. This lets us measure exactly what we are looking for,
which is that the subclass isn't retained.
A lot of other tests were unnecessarily wrapping an object with a
RevCommit and storing that back into the RefUpdate. This is just
retesting what the earlier no-cache test was doing, and complicated
the test considerably. Drop that code and just rely on the value that
was configured by the helper method.
Change-Id: I5b31813484eaa306e9bc4de9622dd5bd4846b16d
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
In bug 323571 it is mentioned that if you call
'toURI().toURL().toString()' on a java.io.File you cannot pass
that string to jgit as an URIish. Problem is that the passed
URI looks like 'file:/C:/a/b.txt' and that we where expecting
double slashes after scheme':'. This fix adds support for this
single-slash file URLs.
Bug: 323571
Change-Id: I866a76a4fcd0c3b58e0d26a104fc4564e7ba5999
Signed-off-by: Christian Halstrick <christian.halstrick@sap.com>
This is the minimal implementation of a "Pull" command. It does not
have any parameters besides the generic progress monitor and timeout.
It works on the currently checked-out branch and assumes that the
configuration contains the keys "branch.<branch name>.remote" and
"branch.<branch name>.merge" to determine the remote configuration
for the fetch and the remote branch name for the merge.
Bug: 303404
Change-Id: I7fe09029996d0cfc09a7d8f097b5d6af1488fa93
Signed-off-by: Mathias Kinzler <mathias.kinzler@sap.com>
Signed-off-by: Chris Aniszczyk <caniszczyk@gmail.com>
There where quite some bugs regarding wrong URI parsing. In order
to solve them the parsing has to be refactored. We now have
specialized regexps for 'scheme://host/...', scp URIs and local
file names. Now we can detect problems while parsing 'git://host:/abc' which
was previously not possible.
Bug: 315571
Bug: 292897
Bug: 307017
Bug: 323571
Bug: 317388
Change-Id: If72576576ebb6b9d9dc8b7e51ddd87c9909e8b62
Signed-off-by: Christian Halstrick <christian.halstrick@sap.com>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
The regular expression which should handle the
user/password part in an URI was potentially
processing too many chars. This led to problems
when user/pwd and port was specified
Change-Id: I87db02494c4b367283e1d00437b1c06d2c8fdd28
Signed-off-by: Christian Halstrick <christian.halstrick@sap.com>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
URIs for the git protocol have to have a hostname.
(see http://www.kernel.org/pub/software/scm/git/docs
/git-clone.html#_git_urls_a_id_urls_a) Some tests tested
URIs like git:/abc.git which is not allowed. Fixed this.
Change-Id: Ia3b8b681ad6592f03b090a874a6e91068a8301fe
Signed-off-by: Christian Halstrick <christian.halstrick@sap.com>
The regular expressions used to parse URI's are constructed by
concatenating different segments to a big String. Introduce
String constants for these segements and document them.
Change-Id: If8b9dbaaf57ca333ac0b6c9610c3d3a515c540f9
Signed-off-by: Christian Halstrick <christian.halstrick@sap.com>
Some strings were not externalized. Also use them in HTTP tests to
ensure that they will also succeed when message bundles are
translated.
Change-Id: Id02717176557e7d57e676e1339cd89f2be88d330
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
Since 858b2c92 we have a HTTP authentication implementation hence
we now get different exception messages when required authentication
headers are not available. This broke the HTTP tests.
Change-Id: Ie08c1ec37e497c2a6f70a75f7c59f0805812a5cc
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
The strings used to construct the regex to parse
URIs are split differently. This makes it easier
to introduce meaningful String constants later on.
Change-Id: I9355fd42e57e0983204465c5d6fe5b6b93655074
Signed-off-by: Christian Halstrick <christian.halstrick@sap.com>
Projects like org.eclipse.mdt contain large XML files about 6 MiB
in size. So does the Android project platform/frameworks/base.
Doing a clone of either project with JGit takes forever to checkout
the files into the working directory, because delta decompression
tends to be very expensive as we need to constantly reposition the
base stream for each copy instruction. This can be made worse by
a very bad ordering of offsets, possibly due to an XML editor that
doesn't preserve the order of elements in the file very well.
Increasing the threshold to the same limit PackWriter uses when
doing delta compression (50 MiB) permits a default configured
JGit to decompress these XML file objects using the faster
random-access arrays, rather than re-seeking through an inflate
stream, significantly reducing checkout time after a clone.
Since this new limit may be dangerously close to the JVM maximum
heap size, every allocation attempt is now wrapped in a try/catch
so that JGit can degrade by switching to the large object stream
mode when the allocation is refused. It will run slower, but the
operation will still complete.
The large stream mode will run very well for big objects that aren't
delta compressed, and is acceptable for delta compressed objects that
are using only forward referencing copy instructions. Copies using
prior offsets are still going to be horrible, and there is nothing
we can do about it except increase core.streamFileThreshold.
We might in the future want to consider changing the way the delta
generators work in JGit and native C Git to avoid prior offsets once
an object reaches a certain size, even if that causes the delta
instruction stream to be slightly larger. Unfortunately native
C Git won't want to do that until its also able to stream objects
rather than malloc them as contiguous blocks.
Change-Id: Ief7a3896afce15073e80d3691bed90c6a3897307
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Chris Aniszczyk <caniszczyk@gmail.com>
Singleton references should be protected from multiple threads. As far as we
know this cannot happen as JUnit is used today since we currently don't run
tests in parallel, but now this code will not prevent anyone.
Change-Id: I29109344d2e8025fa2a3ccaf7c2c16469544ce05
Signed-off-by: Robin Rosenberg <robin.rosenberg@dewire.com>
Natively support the HTTP basic and digest authentication methods
by setting the Authorization header without going through the JREs
java.net.Authenticator API. The Authenticator API is difficult to
work with in a multi-threaded server environment, where its using
a singleton for the entire JVM. Instead compute the Authorization
header from the URIish user and pass, if available.
Change-Id: Ibf83fea57cfb17964020d6aeb3363982be944f87
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
As findbugs pointed out, there was a small risk for creating multiple instances of
translation bundles. If that happens, drop the second instance.
Change-Id: I3aacda86251d511f6bbc2ed7481d561449ce3b6c
Signed-off-by: Robin Rosenberg <robin.rosenberg@dewire.com>
HistogramDiff is an alternative implementation of patience diff,
performing a search over all matching locations and picking the
longest common subsequence that has the lowest occurrence count.
If there are unique common elements, its behavior is identical to
that of patience diff.
Actual performance on real-world source files usually beats
MyersDiff, sometimes by a factor of 3, especially for complex
comparators that ignore whitespace.
Change-Id: I1806cd708087e36d144fb824a0e5ab7cdd579d73
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Each algorithm should produce a particular number of results
given one of the standard inputs used during the performance
tests. To help ensure those tests are accurate, assert that
the edit list length is correct.
Change-Id: I292f8fde0cec6a60a75ce09e70814a00ca47cb99
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Pass through the addAll request to our underlying ArrayList.
This way the underlying ArrayList grows no more than once during the
call, which may be important if the list was originally allocated
at the default size of 16, but 64 Edits are being added.
Change-Id: I31c3261e895766f82c3c832b251a09f6e37e8860
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
FetchCommand was missing the ability to set dry run and thin
preferences on the transport operation.
Change-Id: I0bef388a9b8f2e3a01ecc9e7782aaed7f9ac82ce
Signed-off-by: Chris Aniszczyk <caniszczyk@gmail.com>
This is the test suite I was using to help understand why we had
such a high collision rate with RawTextComparator, and to select
a replacement function.
Since its not something we will run very often, lets make it a
program in the debug package rather than a JUnit test. This way
we can run it on demand against any corpus of files we choose,
but we aren't bottlenecking our daily builds running tests with
no assertions.
Adding a new hash function to this suite is simple, just define
a new instance member of type "Hash" with the logic applied to
the region passed in.
Change-Id: Iec0b176adb464cf95b06cda157932b79c0b59886
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Optional inCore parameter to Resolver/Strategy will
instruct it to perform all the operations in memory
and avoid modifying working folder even if there is one.
Change-Id: I5b873dead3682f79110f58d7806e43f50bcc5045
The hash code returned by RawTextComparator (or that is used
by the SimilarityIndex) play an important role in the speed of
any algorithm that is based upon them. The lower the number of
collisions produced by the hash function, the shorter the hash
chains within hash tables will be, and the less likely we are to
fall into O(N^2) runtime behaviors for algorithms like PatienceDiff.
Our prior hash function was absolutely horrid, so replace it with
the proper definition of the DJB hash that was originally published
by Professor Daniel J. Bernstein.
To support this assertion, below is a table listing the maximum
number of collisions that result when hashing the unique lines in
each source code file of 3 randomly chosen projects:
test_jgit: 931 files; 122 avg. unique lines/file
Algorithm | Collisions
-------------+-----------
prior_hash 418
djb 5
sha1 6
string_hash31 11
test_linux26: 30198 files; 258 avg. unique lines/file
Algorithm | Collisions
-------------+-----------
prior_hash 8675
djb 32
sha1 8
string_hash31 32
test_frameworks_base: 8381 files; 184 avg. unique lines/file
Algorithm | Collisions
-------------+-----------
prior_hash 4615
djb 10
sha1 6
string_hash31 13
We can clearly see that prior_hash performed very poorly, resulting
in 8,675 collisions (elements in the same hash bucket) for at least
one file in the Linux kernel repository. This leads to some very
bad O(N) style insertion and lookup performance, even though the
hash table was sized to be the next power-of-2 larger than the
total number of unique lines in the file.
The djb hash we are replacing prior_hash with performs closer to
SHA-1 in terms of having very few collisions. This indicates it
provides a reasonably distributed output for this type of input,
despite being a much simpler algorithm (and therefore will be much
faster to execute).
The string_hash31 function is provided just to compare results with,
it is the algorithm commonly used by java.lang.String hashCode().
However, life isn't quite this simple.
djb produces a 32 bit hash code, but our hash tables are always
smaller than 2^32 buckets. Mashing the 32 bit code into an array
index used to be done by simply taking the lower bits of the hash
code by a bitwise and operator. This unfortuntely still produces
many collisions, e.g. 32 on the linux-2.6 repository files.
From [1] we can apply a final "cleanup" step to the hash code to
mix the bits together a little better, and give priority to the
higher order bits as they include data from more bytes of input:
test_jgit: 931 files; 122 avg. unique lines/file
Algorithm | Collisions
-------------+-----------
prior_hash 418
djb 5
djb + cleanup 6
test_linux26: 30198 files; 258 avg. unique lines/file
Algorithm | Collisions
-------------+-----------
prior_hash 8675
djb 32
djb + cleanup 7
test_frameworks_base: 8381 files; 184 avg. unique lines/file
Algorithm | Collisions
-------------+-----------
prior_hash 4615
djb 10
djb + cleanup 7
This is a massive improvement, as the number of collisions for
common inputs drops to acceptable levels, and we haven't really
made the hash functions any more complex than they were before.
[1] http://lkml.org/lkml/2009/10/27/404
Change-Id: Ia753b695de9526a157ddba265824240bd05dead1
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Because PatienceDiff works by looking for common unique lines within
the region, the DiffTestDataGenerator needs to be modified to produce
a unique character for each region. If we don't give PatienceDiff
a few unique points, it will just offer back a single REPLACE edit
that covers the entire files, and this doesn't tell us very much.
Change-Id: I5129faea1e763c74739118ca20d86bd62e0deaef
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
For certain tiny input sequences, every DiffAlgorithm should produce
exactly the same results, as there should be no ambiguity. Package
these up in an abstract TestCase that algorithms can extend from in
order to perform basic validation of their implementation.
Since these tests are more complete than what we used to have for
the MyersDiff algorithm, throw away Johannes' tests and only use
this new package.
Change-Id: I9a044330887c849ad4c78aa5c7aa04c783c10252
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
As it turns out, every single diff algorithm we might try to
implement can benfit from using the SequenceComparator's native
concept of the simple reduceCommonStartEnd() step. For most inputs,
there can be a significant number of elements that can be removed
from the space the DiffAlgorithm needs to consider, which will
reduce the overall running time for the final solution.
Pool this logic inside of DiffAlgorithm itself as a default, but
permit a specific algorithm to override it when necessary.
Convert MyersDiff to use this reduction to reduce the space it
needs to search, making it perform slightly better on common inputs.
Change-Id: I14004d771117e4a4ab2a02cace8deaeda9814bc1
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
PatienceDiff always uses a HashedSequence, which promises to provide
constant time access for hash codes during the equals method and
aborts fast if the hash codes don't match. Therefore we don't need
to cache the hash codes inside of the index, saving us memory.
Change-Id: I80bf1e95094b7670e6c0acc26546364a1012d60e
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Shawn O. Pearce
Chris Aniszczyk
Mathias Kinzler
Christian Halstrick
Matthias Sohn
Robin Rosenberg
Christian Halstrick
Dmitry Fink