GitHub user okram opened a pull request:
https://github.com/apache/incubator-tinkerpop/pull/215
TINKERPOP-971: TraversalSource should be fluent like GraphComputer
https://issues.apache.org/jira/browse/TINKERPOP-971
The concept of a `TraversalSource` is now very explicit and used
extensively. Any traversal is composed of both a `TraversalSource` and a
`Traversal`. Prior to this PR, `TraversalSources` were manipulated via a very
awkward/error prone "builder model." Moreover, there were objects that didn't
need to exist -- e.g. `TraversalEngine`. With this PR, the API now looks like
this:
```
g =
graph.traversal().withCompute(SparkGraphComputer.class).withStrategies(ReadOnlyStrategy.instance())
g.V().count()
```
Note that `withComputer()` has lots of useful overloads.
```
g = graph.traversal().withComputer(h ->
h.compute(SparkGraphComputer.class).workers(10))
```
This change is "lightly breaking." For users, nothing changes. The previous
`Builder` model still exists in a `Deprecated` state. All graph system
providers have to do is:
* Implement `XXXGraphProvider.getGraphComputer()` (if and only if they
support `GraphComputer`).
* Change `traverser.getEngine().isGraphComputer()` to
`traversal.getStrategies().onGraphComputer()` if they have custom
`TraversalStrategy` implementations.
* Change `implements EngineDependent` to `implements GraphComputing` if
they have custom steps that implemented `EngineDependent`.
CHANGELOG:
```
* Refactored `TraversalSource` model to allow fluent-method construction of
`TraversalSources`.
* Deprecated the concept of a `TraversalSource.Builder`.
* Removed the concept of a `TraversalEngine`. All `Traversal` modulations
now mediated by `TraversalStrategies`. (*breaking*)
* Added `SideEffectStrategy` for registering sideEffects in a spawned
`Traversal`.
* Added `SackStrategy` for registering a sack for a spawned `Traversal`.
* Added `RequirementsStrategy` and `RequirementsStep` for adding dynamic
`TraverserRequirements` to a `Traversal`.
* Removed `EngineDependentStrategy`.
* Renamed step interface `EngineDependent` to `GraphComputing` with method
`onGraphComputer()`. (*breaking*)
* Cleaned up various `TraversalStrategy` tests now that `TraversalEngine`
no longer exists.
```
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/apache/incubator-tinkerpop TINKERPOP-971
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/incubator-tinkerpop/pull/215.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #215
----
commit 5f2cd6770d56b69c3b0eebf07152315237837f12
Author: Marko A. Rodriguez <[email protected]>
Date: 2016-02-09T00:35:46Z
The most brutal refactor. There is no such thing as a
TraversalSource.Builder anymore. All there are are TraversalSources. These are
simply a wrapper of a TraversalStrategy and a Graph. There is no such thing as
a TraversalEngine anymore. What was ComputerTraversalEngine is now simply a
strategy of TraversalVertexProgramStrategy. There is no such thing as
EngineDependentStrategy -- this is simply TraversalVertexProgramStrategy. In
short, all there are are Graphs and TraversalStrategies. So so so so so much
cleaner. This WILL NOT be backwards compatible for graph system providers. They
will need to update their GraphProvider implementations -- two methods changed.
Moreover, if they have an XXXStrategy that calls
traversal.getEngine().isComputer(), they will need to change that to
traversal.getStrategies().onGraphComputer(). Really simple to for them to fix.
It took me less a minute to fix up Spark, Giraph, and Neo4j. These changes WILL
be backwards compatible for users at the leve
l of graph.traversal(computer(SparkGraphComputer)) still working (though
Depcrecation introduced). The GraphTraversal.Builder will simply turn into a
bunch of TravesalSource calls. It is not used outside of this context. This
commit is NOT complete and will be completed on the next sprint. Moreover,
there are lots of cleanups and JavaDocing that I need to do -- I just don't
want to lose this work so pushing.
commit 908433faf631f6d585dd2ff662c430da030f2857
Author: Marko A. Rodriguez <[email protected]>
Date: 2016-02-09T12:36:13Z
added back in ComputerTraversalEngine, StandardTraversalEngine, and
TraversalSource.Builder. All are deprecated but are functional with respects to
Graph.traversal(TravesalSource.Builder). Both TinkerGraphProvider and
SparkGraphProvider have a 50-percent chance of either using the new traversal
source model or the old deprecated builder model. This way, we are certain
that, for users, this push is backwards compatible. Note that for graph system
providers, there is only one new method they need to implement in their
XXXGraphProvider test code -- getGraphComputer(). Thus, this ticket is
extremely gentle on both users and providers even though it is a massive
refactor.
commit 229474ab17a20de9be93325f8d170d915f60414a
Author: Marko A. Rodriguez <[email protected]>
Date: 2016-02-09T14:29:08Z
Added lots of JavaDoc. Lots of organization and clean up. Ready for PR.
commit 2db6faacff947a3ee456585714101d6ce0399f35
Author: Marko A. Rodriguez <[email protected]>
Date: 2016-02-09T14:42:14Z
renamed EngineDependent to GraphComputing as there is no longer the concept
of a TraversalEngine.
----
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
