• Bumped to Jackson Databind 2.9.8

Improvements

• Removed conflicting non-indy groovy core dependency

• Bumped jython-standalone 2.7.1

• Added a delegate to the Gremlin.Net driver that can be used to configure the WebSocket connection.

• SSL security enhancements

• Added Gremlin version to Gremlin Server startup logging output.

• Fixed problem with Gremlin Server sometimes returning an additional message after a failure.

• Allowed spaces in classpath for gremlin-server.bat.

• Fixed bug in traversals that used Python lambdas with strategies in gremlin-python.

• Modified Maven archetype for Gremlin Server to use remote traversals rather than scripts.

• Added an system error code for failed plugin installs for Gremlin Server -i option.

• Fixed bug in keep-alive requests from over-queuing cancelled jobs.

• Match numbers in choose() options using NumberHelper (match values, ignore data type).

• Added support for GraphSON serialization of Date in Javascript.

• Added synchronized Map to Gryo 1.0 registrations.

• Added Triple to Gryo 1.0 registrations.

• Added support for Double.NaN, Double.POSITIVE_INFINITY and Double.NEGATIVE_INFINITY.

• Improved escaping of special characters in strings passed to the GroovyTranslator.

• Added Cluster configuration option to set a custom validation script to use to test server connectivity in the Java driver.

• Improved ability of GroovyTranslator to handle more types supported by GraphSON.

• Improved ability of GroovyTranslator to handle custom types.

• Added better internal processing of Column in by(Function).

• Added hasNext() support on Traversal for gremlin-python.

• Added support for additional extended types in Gremlin.Net with decimal, TimeSpan, BigInteger, byte, byte[], char and short.

• Fixed bug in Java driver where an disorderly shutdown of the server would cause the client to hang.

• Added a dotnet template project that should make it easier to get started with Gremlin.Net.

• Removed ThreadInterruptCustomizerProvider from documentation as a way to timeout.

• Changed behavior of withRemote() if called multiple times so as to simply throw an exception and not perform the side-effect of auto-closing.

• Added Docker images for Gremlin Console and Gremlin Server.

• Fixed bug in branch() where reducing steps as options would produce incorrect results.

• Removed recursive handling of streaming results from Gremlin-Python driver to avoid max recursion depth errors.

• Improved performance of TraversalVertexProgram and related infrastructure.

• Checked web socket state before closing connection in the .NET driver.

• Deprecated BulkLoaderVertexProgram and related infrastructure.

• Deprecated BulkDumperVertexProgram with the more aptly named CloneVertexProgram.

• Added createGratefulDead() to TinkerFactory to help make it easier to try to instantiate that toy graph.

• Added identifiers to edges in the Kitchen Sink toy graph.

• Ordered the loading of plugins in the Gremlin Console by their position in the configuration file.

• Refactored the Gremlin Server integration testing framework and streamlined that infrastructure.

• Logged the seed used in initializing Random for tests.

• Fixed bug in GroovyTranslator that didn’t properly handle empty Map objects.

• Added concrete configuration methods to SparkGraphComputer to make a more clear API for configuring it.

• Fixed a bug in TinkerGraphCountStrategy, which didn’t consider that certain map steps may not emit an element.

• Fixed a bug in JavaScript GLV where DriverRemoteConnection close() method didn’t returned a Promise instance.

• Bumped to Jackson 2.9.6.

• Sasl Plain Text Authentication added to Gremlin Javascript.

• Ability to send scripts to server added to Gremlin Javascript.

• Translator class added to Gremlin Javascript to translate bytecode to script clientside.

Bugs

Improvements

• Fixed bug where path history was not being preserved for keys in mutations.

• Bumped to httpclient 4.5.5.

• Bumped to Groovy 2.4.15 - fixes bug with Lambda construction.

• Improved performance of GraphSON deserialization of Bytecode.

• Improved performance of traversal construction.

Bugs

Improvements

• Added a Lambda class to Gremlin.Net that makes it possible to use Groovy and Python lambdas with Gremlin.Net.

• Enums are now represented as classes in Gremlin.Net which allows to use them as arguments in more steps.

• Bumped to Groovy 2.4.14.

• Added checkAdjacentVertices option to SubgraphStrategy.

• Modified GremlinDslProcessor so that it generated the getAnonymousTraversalClass() method to return the DSL version of __.

• Added the "Kitchen Sink" test data set.

• Fixed deserialization of P.not() for GraphSON.

• Bumped to Jackson 2.9.4.

• Improved performance of JavaTranslator by caching reflected methods required for traversal construction.

• Ensure that RemoteStrategy is applied before all other DecorationStrategy instances.

• Added idleConnectionTimeout and keepAliveInterval to Gremlin Server that enables a "ping" and auto-close for seemingly dead clients.

• Fixed a bug where lambdas in gremlin-python would trigger a failure if steps using python-only symbols were present (such as as_()).

• Fixed a bug in NumberHelper that led to wrong min/max results if numbers exceeded the Integer limits.

• Delayed setting of the request identifier until RequestMessage construction by the builder.

• ReferenceElement avoids UnsupportedOperationException handling in construction thus improving performance.

• Improved error messaging for failed serialization and deserialization of request/response messages.

• Fixed handling of Direction.BOTH in Messenger implementations to pass the message to the opposite side of the StarGraph.

• Removed hardcoded expectation in metrics serialization test suite as different providers may have different outputs.

• Added IndexedTraverserSet which indexes on the value of a Traverser thus improving performance when used.

• Utilized IndexedTraverserSet in TraversalVertexProgram to avoid extra iteration when doing Vertex lookups.

• Bumped to Netty 4.0.56.Final.

• Fixed .NET GraphSON serialization of P.Within() and P.without() when passing a Collection as an argument.

• Fixed a bug in Gremlin Console which prevented handling of gremlin.sh flags that had an "=" between the flag and its arguments.

• Fixed bug where SparkMessenger was not applying the edgeFunction from MessageScope.

• Fixed a bug in ComputerAwareStep that didn’t handle reset() properly and thus occasionally produced some extra traversers.

• Removed TraversalPredicate class in Gremlin.Net. It is now included in the P class instead.

Bugs

Improvements

• Added core GraphSON classes for Gremlin-Python: UUID, Date, and Timestamp.

• Documented the recommended method for constructing DSLs with Gremlin.Net.

• Provided a method to configure detachment options with EventStrategy.

• Fixed a race condition in TinkerIndex.

• Fixed bug in handling of the long forms of -e and -i (--execute and --interactive respectively) for Gremlin Console.

• Fixed bug in LambdaRestrictionStrategy where traversals using Lambda scripts weren’t causing the strategy to trigger.

• Improved error messaging for bytecode deserialization errors in Gremlin Server.

• Fixed an ArrayOutOfBoundsException in hasId() for the rare situation when the provided collection is empty.

• Bumped to Netty 4.0.53

• TraversalVertexProgram profile() now accounts for worker iteration in GraphComputer OLAP.

• Returned the Builder instance from the DetachedEdge.Builder methods of setOutE and setOutV.

• Added test framework for GLVs.

• Fixed bug in TraversalHelper.replaceStep() where the step being replaced needed to be removed prior to the new one being added.

• Added alias support in the .NET DriverRemoteConnection.

• Added a test for self-edges and fixed Neo4jVertex to provided repeated self-edges on BOTH.

• Better respected permissions on the plugins.txt file and prevented writing if marked as read-only.

• Added getters for the lambdas held by LambdaCollectingBarrierStep, LambdaFlatMapStep and LambdaSideEffectStep.

• Fixed an old hack in GroovyTranslator and PythonTranslator where Elements were being mapped to their id only.

• Fixed an "attachement"-bug in InjectStep with a solution generalized to StartStep.

• Truncate the script in error logs and error return messages for "Method code too large" errors in Gremlin Server.

• Fixed a bug in LambdaRestrictionStrategy where it was too eager to consider a step as being a lambda step.

• ReferenceVertex was missing its label() string. ReferenceElement now supports all label handling.

• Fixed a bug where bytecode containing lambdas would randomly select a traversal source from bindings.

• Deprecated GremlinScriptEngine.eval() methods and replaced them with new overloads that include the specific TraversalSource to bind to.

• Added GraphHelper.cloneElements(Graph original, Graph clone) to the gremlin-test module to quickly clone a graph.

• Added GremlinDsl.AnonymousMethod annotation to help provide explicit types for anonymous methods when the types are not easily inferred.

• Bumped to GMavenPlus 1.6.

• Added better error message for illegal use of repeat()-step.

• Fixed a bug in RangeByIsCountStrategy that led to unexpected behaviors when predicates were used with floating point numbers.

• Bumped to Jackson 2.8.10.

• Deprecated MutationListener.vertexPropertyChanged() method that did not use VertexProperty and added a new method that does.

• Added an EmbeddedRemoteConnection so that it’s possible to mimic a remote connection within the same JVM.

• Supported interruption for remote traversals.

• Allow the :remote command to accept a Cluster object defined in the console itself.

• The Console’s plugin.txt file is only updated if there were manually uninstalled plugins.

• Fixed a bug in MatchStep where mid-traversal where() variables were not being considered in start-scope.

• Generalized MatchStep to locally compute all clauses with barriers (not just reducing barriers).

• Ensured that plugins were applied in the order they were configured.

• Fixed a bug in Neo4jGremlinPlugin that prevented it from loading properly in the GremlinPythonScriptEngine.

• Fixed a bug in ComputerVerificationStrategy where child traversals were being analyzed prior to compilation.

• Fixed a bug that prevented Gremlin from ordering lists and streams made of mixed number types.

• Fixed a bug where keepLabels were being corrupted because a defensive copy was not being made when they were being set by PathRetractionStrategy.

• Cancel script evaluation timeout in GremlinExecutor when script evaluation finished.

• Added a recipe for OLAP traversals with Spark on YARN.

• Added spark-yarn dependencies to the manifest of spark-gremlin.

Bugs

Improvements

This release also includes changes from 3.1.8.

• Bumped to Netty 4.0.50

• Registered HashMap$TreeNode to Gryo.

• Fixed a lambda-leak in SackValueStep where BiFunction must be tested for true lambda status.

• Fixed a bug in RangeByIsCountStrategy that broke any ConnectiveStep that included a child traversal with an optimizable pattern.

• Allowed access to InjectStep.injections for TraversalStrategy analysis.

• Exceptions that occur during result iteration in Gremlin Server will now return SCRIPT_EVALUATION_EXCEPTION rather than SERVER_ERROR.

• AddEdgeStep attaches detached vertices prior to edge creation.

• Added graph element GraphSON serializers in Gremlin-Python.

• Initialization scripts for Gremlin Server will not timeout.

• Added Gremlin.Net.

• ProfileTest is now less stringent about assertions which will reduce burdens on providers.

• GremlinExecutor begins timeout of script evaluation at the time the script was submitted and not from the time it began evaluation.

• Added Gremlin.Net.

• ReferenceFactory and DetachedFactory now detach elements in collections accordingly.

• Deprecated GryoLiteMessageSerializerV1d0 in favor of HaltedTraverserStrategy.

• Deprecated the useMapperFromGraph configuration option for Gremlin Server serializers.

• JavaTranslator is now smart about handling BulkSet and Tree.

• Added annotations to the traversal metrics pretty print.

• EdgeOtherVertexStep is no longer final and can be extended by providers.

• EdgeVertexStep is no longer final and can be extended by providers.

• Deprecated Transaction.submit(Function).

• Fixed HADOOP_GREMLIN_LIBS parsing for Windows.

• Improved GraphSON serialization performance around VertexProperty.

• Changed some tests in EventStrategyProcessTest which were enforcing some unintended semantics around transaction state.

• Added WsAndHttpChannelizer and SaslAndHttpBasicAuthenticationHandler to be allow for servicing Http and Websocket requests to the same server

• Added deep copy of Bytecode to DefaultTraversal.clone().

Bugs

Improvements

This release also includes changes from 3.1.7.

• Fixed folding of multiple hasId() steps into GraphStep.

• Added string performance options to StarGraph.

• Fixed a bug in until(predicate) where it was actually calling emit(predicate).

• Fixed inconsistency in GraphSON serialization of Path where properties of graph elements were being included when serialized.

• Improved performance and memory usage of GraphSON when serializing TinkerGraph and graph elements.

• Removed use of stream() in DetachedEdge and DetachedVertex.

• Deprecated a constructor in DetachedEdge that made use of Pair in favor of a new one that just uses the objects that were in the Pair.

• Improved error messaging on the g.addV(Object…​) when passing an invalid arguments.

• Reduced memory usage for TinkerGraph deserialization in GraphSON by streaming vertices and edges.

• Added the gremlin-archetype-dsl to demonstrate how to structure a Maven project for a DSL.

• Developed and documented patterns for Domain Specific Language implementations.

• Removed the Groovy dependency from gremlin-python and used Groovy Templates and the gmavenplus-plugin to generate the python GLV classes.

• Now using Groovy […​] map notation in GroovyTranslator instead of new LinkedHashMap(){{ }}.

• Maintained type information on Traversal.promise().

• Propagated exception to Future instead of calling thread in RemoteConnection.

• Fixed a bug in RepeatUnrollStrategy where LoopsStep and LambdaHolder should invalidate the strategy’s application.

• Deprecated authentication.className setting in favor of using authentication.authenticator.

• Added authentication.authenticationHandler setting.

• Added abstraction to authentication to allow users to plug in their own AbstractAuthenticationHandler implementations.

• Fixed a NullPointerException bug in B_LP_O_S_SE_SL_Traverser.

• PathRetractionStrategy now uses the marker-model to reduce recursive lookups of invalidating steps.

• ProfileStrategy now uses the marker-model to reduce recursive lookups of ProfileSideEffectStep.

• Mutating steps now implement Scoping interface.

• Fixed a step id compilation bug in AddVertexStartStep, AddVertexStep, AddEdgeStep, and AddPropertyStep.

• Added more details to Gremlin Server client side messages - exception hierarchy and stack trace.

• Deprecated "Exception-Class" in the Gremlin Server HTTP protocol in favor of the new "exceptions" field.

• De-registered metrics on Gremlin Server shutdown.

• Added "help" command option on :remote config for plugins that support that feature in the Gremlin Console.

• Allowed for multiple scripts and related arguments to be passed to gremlin.sh via -i and -e.

• LABELED_PATH requirement is now set if any step in the traversal is labeled.

• Updated PathRetractionStrategy to not run if the provided traversal contains a VertexProgramStep that has a LABELED_PATH requirement.

• Added various metrics to the GremlinGroovyScriptEngine around script compilation and exposed them in Gremlin Server.

• Moved the caffeine dependency down to gremlin-groovy and out of gremlin-server.

• Improved script compilation in GremlinGroovyScriptEngine to use better caching, log long compile times and prevent failed compilations from recompiling on future requests.

• Synchronized script compilation.

• Logged Script compilation times.

• Prevented failed scripts from recompiling.

• Logged warnings for scripts that take "too long" to compile.

• Improved memory usage of the GremlinGroovyScriptEngine.

• Added cyclicPath().from().to().by() support to GraphTraversal.

• Added simplePath().from().to().by() support to GraphTraversal.

• Added path().from().to() support to GraphTraversal so sub-paths can be isolated from the current path.

• Added FromToModulating interface for use with to()- and from()-based step modulators.

• Added Path.subPath() which supports isolating a sub-path from Path via to/from-labels.

• Fixed NullPointerException in GraphMLReader that occurred when an <edge> didn’t have an ID field and the base graph supported ID assignment.

• Added ScopingStrategy which will computer and provide all Scoping steps with the path labels of the global Traversal.

• Split ComputerVerificationStrategy into two strategies: ComputerVerificationStrategy and ComputerFinalizationStrategy.

• Removed HasTest.g_V_hasId_compilationEquality from process test suite as it makes too many assumptions about provider compilation.

• Deprecated CustomizerProvider infrastructure.

• Deprecated PluginAcceptor infrastructure.

• Improved consistency of the application of bindings to GremlinScriptEngine implementations in the BindingsGremlinPlugin.

• Fixed a bug in OLAP ComputerAwareStep where end-step labels were not being appended to the traverser correctly.

• Refactor SparkContext handler to support external kill and stop operations.

• Fixed an optimization bug in LazyBarrierStrategy around appending barriers to the end of a Traversal.

• Fixed an optimization bug in PathRetractionStrategy around appending barriers to the end of a Traversal.

• TraverserIterator in GremlinServer is smart to try and bulk traversers prior to network I/O.

• Improved error handling of compilation failures for very large or highly parameterized script sent to Gremlin Server.

• Fixed a bug in RangeByIsCountStrategy that changed the meaning of inner traversals.

• Improved Gremlin-Python Driver implementation by adding a threaded client with basic connection pooling and support for pluggable websocket clients.

• Changed GraphManager from a final class implementation to an interface.

• Updated GraphManager interface to include methods for opening/instantiating a graph and closing a graph.

• Implemented DefaultGraphManager to include previous GraphManager functionality and adhere to updated interface.

• Deprecated GraphManager.getGraphs() and added GraphManager.getGraphNames().

• Deprecated GraphManager.getTraversalSources() and added GraphManager.getTraversalSourceNames().

• Fixed a bug so now users can supply a YAML with an empty staticVariableTypes to be used by the FileSandboxExtension

Bugs

Improvements

This release also includes changes from 3.1.6.

• Fixed a bug where PathProcessor.keepLabels were not being pushed down into child traversals by PathRetractionStrategy.

• Added default MessagePassingReductionStrategy for GraphComputer that can reduce the number of message passing iterations.

• Fixed a bug associated with user-provided maps and GroupSideEffectStep.

• GroupBiOperator no longer maintains a detached traversal and thus, no more side-effect related OLAP inconsistencies.

• Added ProjectedTraverser which wraps a traverser with a List<Object> of projected data.

• Fixed an optimization bug in CollectingBarrierSteps where the barrier was being consumed on each addBarrier().

• OrderGlobalStep and SampleGlobalStep use ProjectedTraverser and now can work up to the local star graph in OLAP.

• SASL negotiation supports both a byte array and Base64 encoded bytes as a string for authentication to Gremlin Server.

• Deprecated all test suites in gremlin-groovy-test - Graph Providers no longer need to implement these.

• Deprecated TinkerIoRegistry replacing it with the more consistently named TinkerIoRegistryV1d0.

• Made error messaging more consistent during result iteration timeouts in Gremlin Server.

• Fixed a memory leak in the classloader for the GremlinGroovyScriptEngine where classes in the loader were not releasing from memory as a strong reference was always maintained.

• PathRetractionStrategy does not add a NoOpBarrierStep to the end of local children as its wasted computation in 99% of traversals.

• Fixed a bug in AddVertexStartStep where if a side-effect was being used in the parametrization, an NPE occurred.

• Fixed a bug in LazyBarrierStrategy where profile() was deactivating it accidentally.

• Fixed a bug in RepeatUnrollStrategy where stateful DedupGlobalStep was cloned and thus, maintained two deduplication sets.

• Added documentation around "terminal steps" in Gremlin: hasNext(), next(), toList(), etc.

• Added specific GraphSON serializers for RequestMessage and ResponseMessage in GraphSON 2.0.

• Added CloseableIterator to allow Graph providers who open expensive resources a way to let users release them.

• Fixed minor bug in gremlin-driver where closing a session-based Client without initializing it could generate an error.

• Relieved synchronization pressure in various areas of TinkerGraphComputer.

• Fixed an optimization bug in OLAP-based DedupGlobalStep where deduping occurred twice.

• MemoryComputeKey now implements Cloneable which is useful for BiOperator reducers that maintain thread-unsafe state.

• TinkerGraphComputer now supports distributed Memory with lock-free partition aggregation.

• TinkerGraph Gryo and GraphSON deserialization is now configured to use multi-properties.

• Changed behavior of ElementHelper.areEqual(Property, Property) to not throw exceptions with null arguments.

• Added GryoVersion for future flexibility when introducing a new verison of Gryo and moved serializer registrations to it.

• Fixed Gryo serialization of ConnectiveP instances.

• Lessened the severity of Gremlin Server logging when it encounters two or more serializers addressing the same mime type.

• Bumped to Netty 4.0.42.final.

• Added ByteBuffer, InetAddress, Timestamp to the list of Gryo supported classes.

• Fixed Gryo serialization of Class.

• Fixed GraphSON serialization of enums like T, P, etc. where values were overriding each other in the GraphSON type registry.

• Fixed a bug in Gremlin-Python around .() and __.start().

• Fixed a bug around long serialization in Gremlin-Python when using Python3.

• Deprecated TraversalSource.withBindings() as it is no longer needed in Gremlin-Java and never was needed for other variants.

• Fixed a bug in Gremlin-Java Bytecode where anonymous traversals were not aware of parent bindings.

• Fixed a bug in Gremlin-Java GraphSON deserialization around P.within() and P.without().

• Converted Spark process suite tests to "integration" tests.

• Fixed a bug in InlineFilterStrategy having to do with folding HasContainers into VertexStep.

• Deprecated HasContainer.makeHasContainers() which was used to dissect AndP and shouldn’t be used at the TinkerPop-level.

• GraphTraversal.has() now will try and fold-left HasContainer if end step is a HasContainerHolder.

• Created explicit P-predicate methods for GraphTraversal.hasXXX().

• Fixed a bug in FilterRankStrategy around where().by() ordering.

• Added another optimization in RangeByIsCountStrategy, that removes count().is() altogether if it’s not needed.

• Fixed a OLAP MatchStep.clone()-bug that occurs when the match() is in a local child.

• Added another optimization in RangeByIsCountStrategy, that removes count().is() altogether if it’s not needed.

• Fixed a bug in RangeByIsCountStrategy where labeled parents shouldn’t have the strategy applied to their children.

• Fixed a bug in PathRetractionStrategy where MatchEndStep labels were being dropped when they shouldn’t be.

• Added TinkerGraphCountStrategy which translates g.V().map*.count() patterns into direct Map.size() calls in TinkerGraph.

• Added Path.head() and Path.isEmpty() with default method implementations.

• Fixed a NoSuchElementException bug with GroupXXXStep where if the reduced TraverserSet is empty, don’t add the key/value.

• Fixed a NullPointerException bug with profiling GroupSideEffectStep in OLTP.

• Improved ability to release resources in GraphProvider instances in the test suite.

• Factored GremlinPlugin functionality out of gremlin-groovy and into gremlin-core - related classes were deprecated.

• Added a force option for killing sessions without waiting for transaction close or timeout of a currently running job or multiple jobs.

• Deprecated Session.kill() and Session.manualKill().

• Added Traversal.promise() method to allow for asynchronous traversal processing on "remote" traversals.

• Deprecated RemoteConnection.submit(Bytecode) in favor of submitAsync(Bytecode).

• Added choose(predicate,traversal) and choose(traversal,traversal) to effect if/then-semantics (no else). Equivalent to choose(x,y,identity()).

• Removed ImmutablePath.TailPath as it is no longer required with new recursion model.

• Removed call stack recursion in ImmutablePath.

• Gremlin-Python serializes Bytecode as an object (instead of a JSON string) when submit over the RemoteConnection.

• Fixed the handling of the DriverRemoteConnection pass-through configurations to the driver.

• IncidentToAdjacentStrategy now uses a hidden label marker model to avoid repeated recursion for invalidating steps.

• PathProcessorStrategy can inline certain where(traversal)-steps in order to increase the likelihood of star-local children.

• SparkGraphComputer no longer starts a worker iteration if the worker’s partition is empty.

• Added ProjectStep.getProjectKeys() for strategies that rely on such information.

• Added VertexFeatures.supportsDuplicateMultiProperties() for graphs that only support unique values in multi-properties.

• Deprecated the "performance" tests in OptIn.

• Deprecated getInstance() methods in favor of instance() for better consistency with the rest of the API.

• Block calls to "remote" traversal side-effects until the traversal read is complete which signifies an end to iteration.

• Added Pick.none and Pick.any to the serializers and importers.

• Added a class loader to TraversalStrategies.GlobalCache which guarantees strategies are registered prior to GlobalCache.getStrategies().

• Fixed a severe bug where GraphComputer strategies are not being loaded until the second use of the traversal source.

• The root traversal now throws regular NoSuchElementException instead of FastNoSuchElementException. (breaking)

• Added a short sleep to prevent traversal from finishing before it can be interrupted during TraversalInterruptionComputerTest.

• Added support for SSL client authentication

Bugs

Improvements

This release also includes changes from 3.1.5.

• Restructured Gremlin-Python’s GraphSON I/O package to make it easier for users to register serializers/deserializers. (breaking)

• Fixed a bug with TraversalOpProcessor that was returning a final result prior to committing the transaction.

• Fixed a bug in ConnectiveStrategy where infix and/or was not correctly reasoning on choose() HasNextStep injections.

• Increased performance of CredentialGraph authentication.

• Removed Java 8 stream usage from TraversalHelper for performance reasons.

• Fixed a bug in RepeatStep where emit().as('x') wasn’t adding the step labels to the emit-traverser.

• Added GraphComputing.atMaster(boolean) to allow steps to know whether they are executing at master or distributed at workers.

• Fixed a bug in OLAP where DedupGlobalStep wasn’t de-duping local master traversers.

• Added HasContainerHolder.removeHasContainer()-method with default UnsupportedOperationException implementation.

• TraversalSource.withComputer() is simplified to add a VertexProgramStrategy. Easier for language variants.

• Fixed a Set, List, Map bug in the various Translators where such collections were not being internally translated.

• Fixed a Bytecode bug where nested structures (map, list, set) were not being analyzed for bindings and bytecode conversions.

• Fixed a String bug in GroovyTranslator and PythonTranslator where if the string has double-quotes it now uses """ """.

• Added a default TraversalStrategy.getConfiguration() which returns the configuration needed to construct the strategy.

• Computer instances can be created with Computer.create(Configuration) and accessed via Computer.getConf().

• Every TraversalStrategy can be created via a Configuration and a static MyStrategy.create(Configuration).

• Added language-agnostic TraversalStrategy support in Bytecode.

• Added PartitionStrategy.Builder.readPartitions() and deprecated PartitionStrategy.Builder.addPartition().

• A new version of LazyBarrierStrategy has been created and added to the default strategies.

• FilterRankStrategy now propagates labels "right" over non-Scoping filters.

• Fixed a bug in ConnectiveP where nested equivalent connectives should be inlined.

• Fixed a bug in IncidentToAdjacentStrategy where TreeStep traversals were allowed.

• Fixed a end-step label bug in MatchPredicateStrategy.

• Fixed a bug in MatchPredicateStrategy where inlined traversals did not have strategies applied to it.

• Fixed a bug in RepeatUnrollStrategy where inlined traversal did not have strategies applied to it.

• Fixed padding of prompt in Gremlin Console when the number of lines went beyond a single digit.

• Fixed GraphSON 2.0 namespace for TinkerGraph to be "tinker" instead of "gremlin".

• Dropped serialization support in GraphSON 2.0 for Calendar, TimeZone, and Timestamp.

• Added TraversalHelper.copyLabels() for copying (or moving) labels form one step to another.

• Added TraversalHelper.applySingleLevelStrategies() which will apply a subset of strategies but not walk the child tree.

• Added the concept that hidden labels using during traversal compilation are removed at the end during StandardVerificationStrategy. (breaking)

• Added InlineFilterStrategy which will determine if various TraversalParent children are filters and if so, inline them.

• Removed IdentityRemovalStrategy from the default listing as its not worth the clock cycles.

• Removed the "!" symbol in NotStep.toString() as it is confusing and the NotStep-name is sufficient.

• Fixed a bug in TraversalVertexProgram (OLAP) around ordering and connectives (i.e. and() and or()).

• Added AbstractGremlinProcessTest.checkOrderedResults() to make testing ordered results easier.

• AbstractLambdaTraversal now supports a bypassTraversal where it is possible for strategies to redefine such lambda traversals.

• Added an internal utility ClassFilterStep which determines if the traverser object’s class is an instance of the provided class.

• ConnectiveStep extends FilterStep and thus, is more appropriately categorized in the step hierarchy.

• PropertyMapStep supports a provided traversal for accessing the properties of the element. (breaking)

• SubgraphStrategy now supports vertex property filtering.

• Fixed a bug in Gremlin-Python P where predicates reversed the order of the predicates.

• Added tests to DedupTest for the dedup(Scope, String…​) overload.

• Added more detailed reference documentation for IO formats.

• Fixed a bug in serialization of Lambda instances in GraphSON, which prevented their use in remote traversals.

• Fixed a naming bug in Gremlin-Python where P.and and P._or should be P.and and P.or_. (breaking)

• where() predicate-based steps now support by()-modulation.

• Added Gryo serialization for Bytecode.

• Moved utility-based serializers to UtilSerializers for Gryo - these classes were private and hence this change is non-breaking.

• TraversalRing returns a null if it does not contain traversals (previously IdentityTraversal).

• Deprecated Graph.Exceptions.elementNotFoundException() as it was not used in the code base outside of the test suite.

• Fixed a JavaTranslator bug where Bytecode instructions were being mutated during translation.

• Added Path to Gremlin-Python with respective GraphSON 2.0 deserializer.

• Traversal and TraversalSource now implement AutoCloseable.

• Added "keep-alive" functionality to the Java driver, which will send a heartbeat to the server when normal request activity on a connection stops for a period of time.

• Renamed the empty.result.indicator preference to result.indicator.null in Gremlin Console

• If result.indicator.null is set to an empty string, then no "result line" is printed in Gremlin Console.

• Deprecated reconnectInitialDelay on the Java driver.

• Added some validations to Cluster instance building.

• Produced better errors in readGraph of GryoReader and GraphSONReader if a Vertex cannot be found in the cache on edge loading.

• VertexPrograms can now declare traverser requirements, e.g. to have access to the path when used with .program().

• New build options for gremlin-python where -DglvPython is no longer required.

• Added missing InetAddress to GraphSON extension module.

• Added new recipe for "Pagination".

• Added new recipe for "Recommendation".

• Added functionality to Gremlin-Server REST endpoint to forward Exception Messages and Class in HTTP Response

• Gremlin Server TraversalOpProcessor now returns confirmation upon Op close.

• Added close method Java driver and Python driver DriverRemoteTraversalSideEffects.

Bugs

Improvements

This release also includes changes from 3.1.4.

• Included GraphSON as a default serializer (in addition to Gryo, which was already present) in Gremlin Server if none are defined.

• Added gremlin-python package as a Gremlin language variant in Python.

• Added Bytecode which specifies the instructions and arguments used to construct a traversal.

• Created an experimental GraphSON representation of Bytecode that will be considered unstable until 3.3.0.

• Added Translator which allows from the translation of Bytecode into some other form (e.g. script, Traversal, etc.).

• Added JavaTranslator, GroovyTranslator, PythonTranslator, and JythonTranslator for translating Bytecode accordingly.

• Added TranslationStrategy to gremlin-test so translators can be tested against the process test suite.

• Added Traversal.Admin.nextTraverser() to get the next result in bulk-form (w/ default implementation).

• Added TraversalSource.getAnonymousTraversalClass() (w/ default implementation).

• Added GremlinScriptEngine interface which specifies a eval(Bytecode, Bindings) method.

• Deprecated RemoteGraph in favor of TraversalSource.withRemote() as it is more technically correct to tie a remote traversal to the TraversalSource than a Graph instance.

• GremlinGroovyScriptEngine implements GremlinScriptEngine.

• Added GremlinJythonScriptEngine which implements GremlinScriptEngine.

• Removed support for submitting a Java serialized Traversal to Gremlin Server.

• Removed a largely internal feature that supported automatic unrolling of traversers in the Gremlin Driver.

• Made it possible to directly initialize OpProcessor implementations with server Settings.

• Included GraphSON as a default serializer (in addition to Gryo, which was already present) in Gremlin Server if none are defined

• Introduced GraphSON 2.0.

• Deprecated embedTypes on the builder for GraphSONMapper.

• Bumped to Netty 4.0.40.final.

• Defaulted the gremlinPool setting in Gremlin Server to be zero, which will instructs it to use Runtime.availableProcessors() for that settings.

• Changed scope of log4j dependencies so that they would only be used in tests and the binary distributions of Gremlin Console and Server.

• Deprecated Io.Builder.registry() in favor of the newly introduced Io.Builder.onMapper().

• Added new recipe for "Traversal Induced Values".

• Fixed a potential leak of a ReferenceCounted resource in Gremlin Server.

• Added class registrations for Map.Entry implementations to GryoMapper.

• Added methods to retrieve Cluster settings in gremlin-driver.

• Fixed a severe bug in SubgraphStrategy.

• Deprecated SubgraphStrategy.Builder.vertexCriterion()/edgeCriterion() in favor of vertices()/edges().

• Fixed a small bug in StandardVerificationStrategy that caused verification to fail when withPath was used in conjunction with ProfileStep.

• Added color preferences

• Added input, result prompt preferences

• Added multi-line indicator in Gremlin Console

Bugs

Improvements

This release also includes changes from 3.1.3.

• PathProcessor steps now have the ability (if configured through a strategy) to drop Traverser path segments.

• MatchStep in OLTP has a lazy barrier to increase the probability of bulking.

• Added PathRetractionStrategy which will remove labeled path segments that will no longer be referenced.

• Added Path.retract() to support retracting paths based on labels.

• Optimized ImmutablePath and MutablePath equality code removing significant unnecessary object creation code.

• Bumped to Groovy 2.4.7.

• Added RepeatUnrollStrategy to linearize a repeat()-traversal if loop amount is known at compile time.

• Fixed a bug in BranchStep around child integration during clone().

• Fixed a bug in AbstractStep around label set cloning.

• Added TraversalStrategyPerformanceTest for verifying the performance gains of optimization-based traversal strategies.

• TraversalExplanation.prettyPrint() exists which provides word wrapping and GremlinConsole is smart to use console width to control toString().

• TraversalOpProcessor (RemoteConnection) uses HaltedTraverserStrategy metadata to determine detachment procedure prior to returning results.

• Allow DFS paths in HADOOP_GREMLIN_LIBS.

• Added a safer serializer infrastructure for use with SparkGraphComputer that uses KryoSerializer and the new GryoRegistrator.

• Added HaltedTraverserStrategy to allow users to get back different element detachments in OLAP.

• Fixed a NullPointerException bug around nested group()-steps in OLAP.

• Fixed a severe bug around halted traversers in a multi-job OLAP traversal chain.

• Ensure a separation of GraphComputer and VertexProgram configurations in SparkGraphComputer and GiraphGraphComputer.

• PeerPressureVertexProgram now supports dynamic initial vote strength calculations.

• Added EmptyMemory for ease of use when no memory exists.

• Updated VertexComputing.generateProgram() API to include Memory. (breaking)

• ImmutablePath.TailPath is now serializable like ImmutablePath.

• Added ConfigurationCompilerProvider which allows fine-grained control of some of the internal GremlinGroovyScriptEngine settings at the Groovy compilation level.

• Introduced the application/vnd.gremlin-v1.0+gryo-lite serialization type to Gremlin Server which users "reference" elements rather than "detached".

• GryoMapper allows overrides of existing serializers on calls to addCustom on the builder.

• Added a traversal style guide to the recipes cookbook.

• Fixed a bug in master-traversal traverser propagation.

• Added useful methods for custom VertexPrograms to be used with program()-step.

• Increased the test coverage around traverser propagation within a multi-job OLAP traversal.

• Added tests to validate the status of a transaction immediately following calls to close.

• Added tests to ensure that threaded transactions cannot be re-used.

• GraphFilter helper methods are now more intelligent when determining edge direction/label legality.

• Added GraphFilterStrategy to automatically construct GraphFilters via traversal introspection in OLAP.

• Updated the Gephi Plugin to support Gephi 0.9.x.

• Increased the testing and scope of TraversalHelper.isLocalStarGraph().

• Changed signature of get_g_VXlistXv1_v2_v3XX_name and get_g_VXlistX1_2_3XX_name of VertexTest to take arguments for the Traversal to be constructed by extending classes.

• Added VertexProgramInterceptor interface as a general pattern for GraphComputer providers to use for bypassing GraphComputer semantics where appropriate.

• Added SparkStarBarrierInterceptor that uses Spark DSL for local star graph traversals that end with a ReducingBarrierStep.

• Added SparkInterceptorStrategy which identifies which interceptor to use (if any) given the submitted VertexProgram.

• Added SparkSingleIterationStrategy that does not partition nor cache the graph RDD if the traversal does not message pass.

• Added more helper methods to TraversalHelper for handling scoped traversal children.

• Deprecated all "performance" tests based on "JUnit Benchmarks".

• SparkGraphComputer no longer shuffles empty views or empty outgoing messages in order to save time and space.

• TraversalVertexProgram no longer maintains empty halted traverser properties in order to save space.

• Added List<P<V>> constructors to ConnectiveP, AndP, and OrP for ease of use.

• Added support for interactive (-i) and execute (-e) modes for Gremlin Console.

• Displayed line numbers for script execution failures of -e and -i.

• Improved messaging around script execution errors in Gremlin Console.

• Added "help" support to Gremlin Console with the -h flag.

• Added options to better control verbosity of Gremlin Console output with -Q, -V and -D.

• Deprecated the ScriptExecutor - the -e option to gremlin.sh is now handled by Console.

• Traversal now allows cancellation with Thread.interrupt().

• Added a Gremlin language variant tutorial teaching people how to embed Gremlin in a host programming language.

Bugs

Improvements

This release also includes changes from 3.1.2-incubating.

• Bumped to Neo4j 2.3.3.

• Renamed variable local to fs in HadoopGremlinPlugin to avoid a naming conflict with Scope.local. (breaking)

• Added GraphTraversal.optional() which will use the inner traversal if it returns results, else it won’t.

• GroupStep and GroupSideEffectStep make use of mid-traversal reducers to limit memory consumption in OLAP.

• Added GraphTraversal.program(VertexProgram) to allow arbitrary user vertex programs in OLAP.

• Added GraphTraversal.project() for creating a Map<String,E> given the current traverser and an arbitrary number of by()-modulators.

• HADOOP_GREMLIN_LIBS can now reference a directory in HDFS and will be used if the directory does not exist locally.

• Added gremlin-benchmark module with JMH benchmarking base classes that can be used for further benchmark development.

• TraversalStrategies.GlobalCache supports both Graph and GraphComputer strategy registrations.

• select("a","b").by("name").by("age")-style traversals now work in OLAP with new PathProcessorStrategy.

• DedupGlobalStep can now handle star-bound by()-modulators and scoped keys on GraphComputer.

• Added Computer which is a builder for GraphComputers that is serializable.

• PersistedOutputRDD now implements PersistResultGraphAware and thus, no more unneeded warnings when using it.

• Renamed StandardTraversalMetrics to DefaultTraversalMetrics given the DefaultXXX-convention throughout. (breaking)

• Bumped to Apache Hadoop 2.7.2.

• Fixed a bug around profiling and nested traversals.

• Added gremlin.hadoop.defaultGraphComputer so users can use graph.compute() with HadoopGraph.

• Added gremlin.hadoop.graphReader and gremlin.hadoop.graphWriter which can handled XXXFormats and XXXRDDs.

• Deprecated gremlin.hadoop.graphInputFormat, gremlin.hadoop.graphOutputFormat, gremlin.spark.graphInputRDD, and gremlin.spark.graphOutputRDD.

• If no configuration is provided to HadoopPools it uses the default configuration to create a pool once and only once per JVM.

• Implemented RemoteGraph, RemoteConnection, and RemoteStrategy.

• Added validation to GryoMapper Kryo identifiers before construction to prevent accidental duplicates.

• Added GraphStep.addIds() which is useful for HasContainer "fold ins."

• Added a static GraphStep.processHashContainerIds() helper for handling id-based HasContainers.

• GraphStep implementations should have g.V().hasId(x) and g.V(x) compile equivalently. (breaking)

• Optimized ExpandableStepIterator with simpler logic and increased the likelihood of bulking.

• Optimized TraverserRequirement calculations.

• Step.addStart() and Step.addStarts() now take Traverser.Admin<S> and Traverser.Admin<S>, respectively. (breaking)

• Step.processNextStart() and Step.next() now return Traverser.Admin<E>. (breaking)

• Traversal.addTraverserRequirement() method removed. (breaking)

• Fixed a hashCode() bug in OrderGlobalStep and OrderLocalStep.

• Added OrderLimitStrategy which will ensure that partitions are limited before being merged in OLAP.

• ComparatorHolder now separates the traversal from the comparator. (breaking)

• Bumped to Apache Spark 1.6.1.

• If no Spark serializer is provided then GryoSerializer is the default, not JavaSerializer.

• Added Operator.sumLong as a optimized binary operator intended to be used by Memory reducers that know they are dealing with longs.

• Traversers from ComputerResultStep are no longer attached. Attaching is only used in TinkerPop’s test suite via System.getProperties().

• Fixed a hashCode()/equals() bug in MessageScope.

• Fixed a severe Traversal cloning issue that caused inconsistent TraversalSideEffects.

• TraversalSideEffects remain consistent and usable across multiple chained OLAP jobs.

• Added MemoryTraversalSideEffects which wraps Memory in a TraversalSideEffects for use in OLAP.

• TraversalSideEffects are now fully functional in OLAP save that an accurate global view is possible at the start of an iteration (not during).

• Updated the TraversalSideEffects API to support registered reducers and updated get()-semantics. (breaking)

• Split existing profile() into ProfileStep and ProfileSideEffectStep.

• The profile()-step acts like a reducing barrier and emits TraversalMetrics without the need for cap(). (breaking)

• Added LocalBarrier interface to allow traversers to remain distributed during an iteration so as to reduce cluster traffic.

• Added NoOpBarrierStep as a LocalBarrier implementation of LambdaCollectingBarrierStep(noOp).

• AggregateStep implements LocalBarrier and thus, doesn’t needlessly communicate its barrier traversers.

• Fixed an OLAP-based Barrier synchronization bug.

• Fixed a semantic bug in BranchStep (and inheriting steps) where barriers reacted locally. (breaking)

• Added MemoryComputeKey for specification of Memory keys in VertexProgram. (breaking)

• Added VertexComputeKey for specification of vertex compute properties in VertexProgram. (breaking)

• Added and, or, and addAll to Operator.

• Memory API changed to support setting and adding values for reduction. (breaking)

• Memory keys can be marked as broadcast and only those values are sent to workers on each iterator.

• Memory keys can be marked transient and thus deleted at the end of the OLAP job.

• Vertex compute keys can be marked transient and thus deleted at the end of the OLAP job.

• VertexProgram API changed to support MemoryComputeKey and VertexComputeKey. (breaking)

• TraversalVertexProgram able to execute OLAP and OLTP traversal sections dynamically within the same job.

• Removed FinalGet interface as all post processing of reductions should be handled by the reducing step explicitly. (breaking)

• Simplified all SupplyingBarrierStep implementations as they no longer require MapReduce in OLAP.

• Simplified all CollectingBarrierStep implementations as they no longer require MapReduce in OLAP.

• Simplified all ReducingBarrierStep implementations as they no longer require MapReduce in OLAP.

• All steps in OLAP that used MapReduce now use Memory to do their reductions which expands the list of legal traversals.

• GroupStep simplified with GroupHelper.GroupMap no longer being needed. Related to the removal of FinalGet.

• OLAP side-effects that are no longer generated by MapReduce are simply stored in ComputerResult.Memory w/ no disk persistence needed. (breaking)

• Added Generate step interface which states that there could be a final generating phase to a side-effect or reduction (e.g. GroupStep).

• Barrier step interface is now the means by which non-parallel steps communicate with their counterparts in OLAP.

• Added MemoryComputing step interface which states that the step uses MemoryComputeKeys for its computation in OLAP.

• Added PeerPressureVertexProgramStep and GraphTraversal.peerPressure().

• Added PureTraversal for handling pure and compiled versions of a Traversal. Useful in OLAP.

• Added ScriptTraversal which allows for delayed compilation of script-based Traversals.

• Simplified VertexProgram implementations with a PureTraversal-model and deprecated ConfigurationTraversal.

• Simplified script-based Traversals via ScriptTraversal and deprecated TraversalScriptFunction and TraversalScriptHelper.

• Added TimesModulating interface which allows the Step to decide how a times()-modulation should be handled.

• Added ByModulating interface which allows the Step to decide how a by()-modulation should be handled. (breaking)

• Simplified the by()-modulation patterns of OrderGlobalStep and OrderLocalStep.

• Added GraphComputerTest.shouldSupportPreExistingComputeKeys() to ensure existing compute keys are "revived." (breaking)

• Added GraphComputerTest.shouldSupportJobChaining() to ensure OLAP jobs can be linearly chained. (breaking)

• Fixed a bug in both SparkGraphComputer and GiraphGraphComputer regarding source data access in job chains.

• Expanded job chaining test coverage for GraphComputer providers.

• Added TraversalHelper.onGraphComputer(traversal).

• MapReduce.map() no longer has a default implementation. This method must be implemented. (breaking)

• TraversalVertexProgram can work without a GraphStep start.

• Added PageRankVertexProgramStep and GraphTraversal.pageRank().

• Added TraversalVertexProgramStep to support OLAP traversal job chaining.

• Added VertexProgramStrategy which compiles multiple OLAP jobs into a single traversal.

• Simplified the comparator model in OrderGlobalStep and OrderLocalStep.

• 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 are 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.

• Added GraphFilter to support filtering out vertices and edges that won’t be touched by an OLAP job.

• Added GraphComputer.vertices() and GraphComputer.edges() for GraphFilter construction. (breaking)

• SparkGraphComputer, GiraphGraphComputer, and TinkerGraphComputer all support GraphFilter.

• Added GraphComputerTest.shouldSupportGraphFilter() which verifies all filtered graphs have the same topology.

• Added GraphFilterAware interface to hadoop-gremlin/ which tells the OLAP engine that the InputFormat handles filtering.

• GryoInputFormat and ScriptInputFormat implement GraphFilterAware.

• Added GraphFilterInputFormat which handles graph filtering for InputFormats that are not GraphFilterAware.

• Fixed a bug in TraversalHelper.isLocalStarGraph() which allowed certain illegal traversals to pass.

• Added TraversalHelper.isLocalProperties() to verify that the traversal does not touch incident edges.

• GraphReader I/O interface now has Optional<Vertex> readGraph(InputStream, GraphFilter). Default UnsupportedOperationException.

• GryoReader does not materialize edges that will be filtered out and this greatly reduces GC and load times.

• Created custom Serializers for SparkGraphComputer message-passing classes which reduce graph sizes significantly.

Bugs

Improvements