Repository: tinkerpop
Updated Branches:
refs/heads/TINKERPOP-1564 8b2dcb053 -> 9349612e8
Have SparkGraphComputer integration test over TinkerGraph with HashPartitioner.
All tests pass. BAM. Tomorrow I'm going to create
TinkerGraphPartitionerProvider in tinkergraph-gremlin/ so it can be
minor nothing tweak.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/bb97c942
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/bb97c942
Diff: http://git-wip-us.apache.org/repos/asf/tinkerpop/diff/bb97c942
TINKERPOP-1565 Enabled more GraphSON IO tests
It was now possible to enable more tests as certain changes that were made to
master on the GraphSON 2.0 format could be reverted and then migrated to
GraphSON 3.0.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit:
Removed a PathProcessor.ID constraint from ComputerVerificationStrategy.
Moreover, sampling and ordering is more efficient as the projected data is
co-located with the traverser in the new ProjectedTraverser wrapper. Going to
leave it at this for tp32/... Moving forward, we can make it so we
TINKERPOP-1565 Reverted GraphSON 3.0 changes
These changes were part of the original work on this branch to alter the
structure of GraphSON 2.0. I'd migrated those changes to 3.0 to avoid breaking
change. Since that time there has been a lot of work on and off this branch
related to this work
updated CHANGELOG and tweaked ProjectTraverser variable naming.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/91e1f50c
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/91e1f50c
Diff:
TINKERPOP-1565 Added tests for GraphSON 3.0
And fixed some broken tests preventing a proper build.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/729af57f
Tree:
TINKERPOP-1565 Updated the IO dev docs with GraphSON 3.0.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/2dc9b51d
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/2dc9b51d
Diff:
first draft of new updates to GraphSON 2.0. The object models are concise, all
the elements/properties are attachable. Really wasn't that much of a headache
to change -- a few test cases in Gremlin-Python needed updating and in
GraphSONMessageSerializerV2d0Test. The corresponding ticket in JIRA
moved all the GroupStep work against tp32/
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/3496402a
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/3496402a
Diff:
Merge branch 'TINKERPOP-1248' into tp32
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/379a6e5e
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/379a6e5e
Diff:
Merge branch 'tp32'
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/66f1b235
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/66f1b235
Diff: http://git-wip-us.apache.org/repos/asf/tinkerpop/diff/66f1b235
introduced a minor bug into CollectingBarrierStep that would only be noticed by
asynchrnous traversal execution engines. I noticed it in the GraphActors
branch. Also, added a toString() to ProjectedTraverser. CTR.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit:
TINKERPOP-1565 Stub out GraphSON 3.0
Set up the new testing model and generated test files. Reverted changes on
property serialization that went in as breaking changes on GraphSON 2.0 and
moved them forward to GraphSON 3.0.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
added a private no-arg constructor to MultiComparator so it can be serialized.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/ee1ab08b
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/ee1ab08b
Diff:
conflicts in GryoVersion -- repeated values. fixed.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/1bc66ab0
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/1bc66ab0
Diff:
@dkuppitz found an optimization trick for MultiComparator. The startIndex for
comparing should start after the last Order.shuffle.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/341ebf98
Tree:
http://git-wip-us.apache.org/repos/asf/tinkerpop/blob/7db10c40/gremlin-tools/gremlin-io-test/src/test/resources/org/apache/tinkerpop/gremlin/structure/io/graphson/_3_3_0/sessionlessevalaliased-v2d0-partial.json
--
diff --git
Merge branch 'tp32'
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/bfe432eb
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/bfe432eb
Diff: http://git-wip-us.apache.org/repos/asf/tinkerpop/diff/bfe432eb
We now have ProjectedTraveser which is a Traverser with List of
projections. OrderGlobalStep uses this and thus, can order for everything
within the local star graph. Added MultiComparator which is like
ChainedComparator but doesn't contain traversal projections -- just comparators.
Project:
A proof that TINKERPOP-1261 is fixed given the refactor of GroupBiOperator.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/d54b490a
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/d54b490a
Diff:
http://git-wip-us.apache.org/repos/asf/tinkerpop/blob/7db10c40/gremlin-driver/src/test/java/org/apache/tinkerpop/gremlin/driver/ser/GraphSONMessageSerializerV3d0Test.java
--
diff --git
TINKERPOP-1565 Updated changelog
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/8ffd0b8f
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/8ffd0b8f
Diff:
updated CHANGELOG.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/60022999
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/60022999
Diff: http://git-wip-us.apache.org/repos/asf/tinkerpop/diff/60022999
Not going to touch GroupStep in this ticket. Too complicated ... will do for
the next release. Minor tweaks and cleanups.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/973484d1
Tree:
Repository: tinkerpop
Updated Branches:
refs/heads/TINKERPOP-1565 8b9ac6d6f -> 8ffd0b8fb (forced update)
Created OrdertedTraverser which allows us to move beyond the star graph for
OrderGlobalStep by()-projections. Moreover, it reduces the complexity of
ordering as the objects of comparison
TINKERPOP-1414 Made GraphSON 3.0 the default serialization format.
This was previously GraphSON 2.0 but recent consensus has this now at 3.0 since
2.0 will not be altered to support breaking change.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit:
Repository: tinkerpop
Updated Branches:
refs/heads/TINKERPOP-1564 6b6df35d0 -> 8b2dcb053
okay, more tests exposed. The final issue is nested
group().groupCount().select(keys) style traversals. I believe I have attachment
covered in this situation, but obviously not. I'm blasted done for
Repository: tinkerpop
Updated Branches:
refs/heads/TINKERPOP-1564 a2dbe7b0b -> 6b6df35d0
got LocalBarriers with SideEffects working in GraphActors. There are only 3
tests that fail -- having to do with nested group().groupCount()-style things.
Almost done.
Project:
refactored GraphActors packaging -- its not actor/, but actors/. JavaDoc and
various cleanups. Also, about to NOT serialize a traversal but instead use
Bytecode. Next push will do this with TraveraslVertexProgram.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit:
revert to master/ representation of GroupStep. TINKERPOP-1590 has the clone
safe model.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/16282914
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/16282914
okay, so I now have the GryoMapper classes programmically loaded in the
ActorSystem serialiation-bindings. This is slick. We are almost there. Done for
the night. That was a 13 hour day.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit:
added a GraphComputerTest that ensures that the GraphComputer configuration
evolves as expected and that it does not alter the Graph configuration.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/55919886
Tree:
okay. so I created ProcessActorsSuite and our first actors specific test --
GraphActorsTest (currently just makes sure toString() is valid :). Really
organizing things around Configuration... Just like Graphs, Computers and
Actors are defined by Configurations.
Project:
ahhh.. a really solid simplification using the Terminate token as the barrier
synchronizer. Lots of other little random clean me ups. Things are looking
brighter at close of business.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit:
GroupStep work... Moving to another branch so committing what I had left after
from last week.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/af5a6f6e
Tree:
GryoVersion.3_0 ... epic. @spmallette making my life easy. OrderedTraverser
registered and private no-arg constructor added. More
AkkaActorsProcessActorsTest cases passing. The only ones that are failing have
to do with GroupStep.GroupBiOperator. So lost as to why its a problem. :|
Project:
I have Master/Worker serialization working fully now. Prior to this moment, the
master was sent to an Akka server and then it executed its workers locally. Now
the workers are distributed to where their physical Partition is. Furthermore,
ActorProgams are fully serializable via
more alignment between GraphActors and GraphComputer.
GraphActors.partitioner().program().submit(). Unlike GraphComputer, there is no
Graph.actors(). HOWEVER -- since GraphActors is Partition-centric, perhaps we
do Graph.partitioner().actors(). eek. hehe..
Project:
every ActorSystem needs a unique system name. Thus, every GraphActors submit()
needs it. Added UUID creation.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/46a6a3b8
Tree:
small nothing.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/ad0f094f
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/ad0f094f
Diff: http://git-wip-us.apache.org/repos/asf/tinkerpop/diff/ad0f094f
OMG. That was a 3+ hour bug. The bug was not in Actors, but in GroupStep
cloning So painful. Renamed Actors to GraphActors so its like
GraphComputer. Renamed ActorStep to TraversalActorProgramStep so its like
TraversalVertexProgramStep. Next up, get rid of GraphComputing and add
PushBased
added OrderedTraverser to Gryo registration listing Crap, prolly means I
need to do it for GraphSON too...
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/2aa1dc3b
Tree:
can't really get my head around the best way to do partition attachment
keep playing with different models... each have their problems. Minor commit
with Partitioner.getPartition(Element) to Partition.find(Element).
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit:
commited to g.withProcessor(Processor.Description). Actors and Computer
implement Processor.Description. Following the VertexProgramStrategy model,
this makes it easy for language variant providers to support any arbitrary
Processor down the line.
Project:
taking a lunch break --- added more JavaDoc and started honing in on how
withProcessor() will generalize... sorta tricky.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/3bf924b4
Tree:
through hell and back to get IoRegistries loaded into Akka GryoSerializer.
Phew. Learned a bunch about Akka Config.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/4463d865
Tree:
We now have PartitionInputFormat which allows an Hadoop-based GraphComputer to
pull data from any Graph implementation. PartitionInputSplit is basically a
wrapper around Partition so that data access is data-local and thus, for
distributed Graph databases, it will simply Partition.vertices()
smarter distribution of initial GraphStep traversers.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/bd9d0477
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/bd9d0477
Diff:
HashPartitioner requires a positive hashCode() value. Thus, wrapped calls in
Math.abs().
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/ea577f20
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/ea577f20
We now have AkkaGraphActors configured fully via our Configuration model --
graph.configuration(), actorProgram.configuration(), etc... Generates
appropriate Config for Akka and we are off to the races. More work on Host.
I have some good ideas here... didn't have time to get to it today.
OrderedTraverser needs a equals() and hashCode().
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/61f73b2e
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/61f73b2e
Diff:
Repository: tinkerpop
Updated Branches:
refs/heads/TINKERPOP-1564 bc61e6a72 -> a2dbe7b0b (forced update)
more organization and cleaning and learning next up Host and Proxy.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit:
starting to fill out GraphActorsTest (like GraphComputerTest). Tweaks to
AkkaConfigFactory for default 'local' settings.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/48d860f3
Tree:
http://git-wip-us.apache.org/repos/asf/tinkerpop/blob/243ab6a7/gremlin-core/src/main/java/org/apache/tinkerpop/gremlin/process/actors/GraphActors.java
--
diff --git
Created AkkaGremlinPlugin so you can import akka-gremlin into GremlinConsole
(and GremlinServer--not sure). The concept of an Actor Address now has to
identifies -- the physical location of the actor and the unique (provider
specific) guid of the actor. By knowing the physical location and
got new GraphComputer.configuration() model working for SparkGraphComputer and
GiraphGraphComputer.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/eaf95a5e
Tree:
what a day. So I have Akka remoting working (not fully). We now have
akka.io.GryoSerializer :). And TraversalActorProgram is smart about detach()
and attach(). Learned a bunch about how to do Partitioner/Partitions in
TinkerPop stuff is coming along nicely. Need a break though...been coding
http://git-wip-us.apache.org/repos/asf/tinkerpop/blob/243ab6a7/gremlin-core/src/main/java/org/apache/tinkerpop/gremlin/process/actor/ActorProgram.java
--
diff --git
added Pushing and Distributing which are currently the two constructs of
GraphComputing. GraphComputing is now deprecated with its corresponding methods
calling the respective Pushing/Distributing methods. Pushing means the step is
using push-based semantics. Distributing means that the step
So much. withProcessor(Processor). No more Compute. Process.submit(Graph) as we
are now staging it so that every Processor (GraphComputer/GraphAgents) can work
over any Graph. This will all be via Partitioner. withComputer() deprecated.
Lots of cool stuff with Process strategies --
finally figured out what is wrong with GroupStep. Just sending over the Barrier
object -- no need to send over the whole traversal during serialization. Phew.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/6c212c31
came up with a clever idea for maintaining order in a distributed traversal.
OrdereredTraverser is a wrapper around a Traverser where all methods are
delegated to the internal traverser. However, OrderedTraverser has an order()
method which returns an int. This allows an ordered stream of
trivial tweaks. going to take a break.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/50367dd2
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/50367dd2
Diff:
opened up more test cases to AkkaGraphActors. A few that fail don't make sense
too me...but very close to the entire test suite passing.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/2cd936a3
Tree:
Added AkkaConfigFactory which deals with creating an Akka Config. Identical in
nature to SparkConf and Hadoop's Configuration. Akka GryoSerializer pimp'd up a
bit. I had a default pool size of 256. Test cases were slow. Changed to 1 --
BOOMIN fast. Wonder what the nature of serializer reuse is
Was able to open up a bunch more test cases to AkkaGraphActors. There are a few
that are still busted -- and I believe these are related to nested collections
and complexities with attachment of elements in the collections.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit:
Tweaks to SparkGraphComputer configuration settings. More GraphComputerTest
things. Before the break, I want to have Configuration all stubbed out for
GraphComputer and GraphActors and then this branch is simply a function of
testing GraphActors at scale when we return.
Project:
trying to surpress the logging during testing, but I'm clueles.. Need
@spmallette help.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/49a028e2
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/49a028e2
having to redesign our concept of Attachable.attach() and detach().
@spmallette's Host interface now has a default method of Host.attach(t).
Moreover, Traverser now has Traverser.attach(Host). Vertex extends Host. Graph
extends Host... and now Partition extends Host. Instead of leaving the
lots of cleanup and organization of the interfaces. trying to line up the
GraphActors interfaces as much as possible with GraphComputer interfaces.
JavaDoc'in.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit:
added the @dkuppitz example to the AkkaGraphActors test/ package for playing
and benchmarking.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/a65b0bdc
Tree:
added StringFactory rules for GraphActors, Partitioner, and Partition. Realized
someting profoundly sucky about DedupGlobalStep. Exposed more test cases to
AkkaGraphActors.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit:
http://git-wip-us.apache.org/repos/asf/tinkerpop/blob/67849635/gremlin-test/src/main/java/org/apache/tinkerpop/gremlin/process/computer/GraphComputerTest.java
--
diff --git
learned about preStart() and postStop() methods in Akka Actors. Lined those up
with actor.terminate() and actor.setup(). JavaDoc here and there nada much.
done for the night.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit:
split up start and terminate messages.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/aa0ae03a
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/aa0ae03a
Diff:
added graphNotSupported coverage to GraphComputerTest. Added JavaDocs here and
there while I was waiting for the test suite to complete.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/9852a60e
Tree:
Repository: tinkerpop
Updated Branches:
refs/heads/TINKERPOP-1565 7f23d31f4 -> 8b9ac6d6f
TINKERPOP-1565 Updated changelog
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/8b9ac6d6
Tree:
Repository: tinkerpop
Updated Branches:
refs/heads/TINKERPOP-1565 c952b741f -> 7f23d31f4
TINKERPOP-1565 Reverted GraphSON 3.0 changes
These changes were part of the original work on this branch to alter the
structure of GraphSON 2.0. I'd migrated those changes to 3.0 to avoid breaking
Repository: tinkerpop
Updated Branches:
refs/heads/master 1bc66ab06 -> bfe432eb8
introduced a minor bug into CollectingBarrierStep that would only be noticed by
asynchrnous traversal execution engines. I noticed it in the GraphActors
branch. Also, added a toString() to ProjectedTraverser.
Merge branch 'tp32'
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/bfe432eb
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/bfe432eb
Diff: http://git-wip-us.apache.org/repos/asf/tinkerpop/diff/bfe432eb
Repository: tinkerpop
Updated Branches:
refs/heads/tp32 600229992 -> 8ad291134
introduced a minor bug into CollectingBarrierStep that would only be noticed by
asynchrnous traversal execution engines. I noticed it in the GraphActors
branch. Also, added a toString() to ProjectedTraverser. CTR.
Repository: tinkerpop
Updated Branches:
refs/heads/TINKERPOP-1564 80adc4273 -> bc61e6a72
wow. found a bug in my OrderGlobalStep work in the lastest PR that only shows
up in GraphActors.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit:
what a day. So I have Akka remoting working (not fully). We now have
akka.io.GryoSerializer :). And TraversalActorProgram is smart about detach()
and attach(). Learned a bunch about how to do Partitioner/Partitions in
TinkerPop stuff is coming along nicely. Need a break though...been coding
Added AkkaConfigFactory which deals with creating an Akka Config. Identical in
nature to SparkConf and Hadoop's Configuration. Akka GryoSerializer pimp'd up a
bit. I had a default pool size of 256. Test cases were slow. Changed to 1 --
BOOMIN fast. Wonder what the nature of serializer reuse is
http://git-wip-us.apache.org/repos/asf/tinkerpop/blob/825f5156/gremlin-test/src/main/java/org/apache/tinkerpop/gremlin/process/computer/GraphComputerTest.java
--
diff --git
refactored GraphActors packaging -- its not actor/, but actors/. JavaDoc and
various cleanups. Also, about to NOT serialize a traversal but instead use
Bytecode. Next push will do this with TraveraslVertexProgram.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit:
Tweaks to SparkGraphComputer configuration settings. More GraphComputerTest
things. Before the break, I want to have Configuration all stubbed out for
GraphComputer and GraphActors and then this branch is simply a function of
testing GraphActors at scale when we return.
Project:
HashPartitioner requires a positive hashCode() value. Thus, wrapped calls in
Math.abs().
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/63f5c0f2
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/63f5c0f2
We now have AkkaGraphActors configured fully via our Configuration model --
graph.configuration(), actorProgram.configuration(), etc... Generates
appropriate Config for Akka and we are off to the races. More work on Host.
I have some good ideas here... didn't have time to get to it today.
ahhh.. a really solid simplification using the Terminate token as the barrier
synchronizer. Lots of other little random clean me ups. Things are looking
brighter at close of business.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit:
added a GraphComputerTest that ensures that the GraphComputer configuration
evolves as expected and that it does not alter the Graph configuration.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/097f5409
Tree:
trivial tweaks. going to take a break.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/0029b6fc
Tree: http://git-wip-us.apache.org/repos/asf/tinkerpop/tree/0029b6fc
Diff:
starting to fill out GraphActorsTest (like GraphComputerTest). Tweaks to
AkkaConfigFactory for default 'local' settings.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit: http://git-wip-us.apache.org/repos/asf/tinkerpop/commit/95d08570
Tree:
okay. so I created ProcessActorsSuite and our first actors specific test --
GraphActorsTest (currently just makes sure toString() is valid :). Really
organizing things around Configuration... Just like Graphs, Computers and
Actors are defined by Configurations.
Project:
We now have PartitionInputFormat which allows an Hadoop-based GraphComputer to
pull data from any Graph implementation. PartitionInputSplit is basically a
wrapper around Partition so that data access is data-local and thus, for
distributed Graph databases, it will simply Partition.vertices()
added Pushing and Distributing which are currently the two constructs of
GraphComputing. GraphComputing is now deprecated with its corresponding methods
calling the respective Pushing/Distributing methods. Pushing means the step is
using push-based semantics. Distributing means that the step
added StringFactory rules for GraphActors, Partitioner, and Partition. Realized
someting profoundly sucky about DedupGlobalStep. Exposed more test cases to
AkkaGraphActors.
Project: http://git-wip-us.apache.org/repos/asf/tinkerpop/repo
Commit:
Created AkkaGremlinPlugin so you can import akka-gremlin into GremlinConsole
(and GremlinServer--not sure). The concept of an Actor Address now has to
identifies -- the physical location of the actor and the unique (provider
specific) guid of the actor. By knowing the physical location and
more alignment between GraphActors and GraphComputer.
GraphActors.partitioner().program().submit(). Unlike GraphComputer, there is no
Graph.actors(). HOWEVER -- since GraphActors is Partition-centric, perhaps we
do Graph.partitioner().actors(). eek. hehe..
Project:
1 - 100 of 155 matches
Mail list logo