Release Notes

Gremlin: A Graph Traversal Language
http://gremlin.tinkerpop.com

Gremlin 2.y.z

Version 2.3.0 (March 19, 2013)

<dependency>
  <groupId>com.tinkerpop.gremlin</groupId>
  <artifactId>gremlin-java</artifactId>
  <version>2.3.0</version>
</dependency>

<dependency>
  <groupId>com.tinkerpop.gremlin</groupId>
  <artifactId>gremlin-groovy</artifactId>
  <version>2.3.0</version>
</dependency>

• Added support for Vertex.addEdge() with Map<String,Object> properties for the edge
• A semantic change to store().cap() where StorePipe now implements LazySideEffectPipe
• Updated all TransformPipes to new typing interface rules in Pipes
• except()/retain() support String[] of named steps to reference as collection
• GroovyPipeFunction provides support for AsMap referencing and closure access to named-steps
• Bumped to support Groovy 1.8.9
• Added GremlinPipeline.remove() which removes all elements at the end of the pipeline (overrides Iterator.remove())
• Added LinkPipe and linkOut, linkIn, linkBoth steps to allow for the addition of linking current vertex with past vertex
• Added orderMap step which takes a Map as input, orders its entries, and then emits its keys
• Added Tokens.T.decr and Tokens.T.incr for default ordering in order and orderMap steps
• ScriptExecutor provides support for args[] as well as a1, a2, etc. type parameters
• Fixed a StackOverflowError in GremlinGroovyScriptEngine having to do with infinite looping on function calls
• GremlinGroovyScriptEngine now maintains an internal cache reset to ensure no OutOfMemoryException issues
• The map step now is map(String... keys) with user provided keys to get (id and label possible options)
• The console can be initialized with a script to pre-define steps and initialize other aspects of the environment
• Added shuffle step which gathers all objects into a list and randomizes the order of their output

Version 2.2.0 (December 19, 2012)

<dependency>
  <groupId>com.tinkerpop.gremlin</groupId>
  <artifactId>gremlin-java</artifactId>
  <version>2.2.0</version>
</dependency>

<dependency>
  <groupId>com.tinkerpop.gremlin</groupId>
  <artifactId>gremlin-groovy</artifactId>
  <version>2.2.0</version>
</dependency>

• Revised method signatures for loadGraphSON() and saveGraphSON() that now use the GraphSONMode enum
• Bumped to support Groovy 1.8.8
• Fixed a bug in QueryPipe limit, where limit was not decrementing via counting already emitted elements
• Provided a vertex-centric query optimization for out[0..1] type traversals
• Added support for non-zero low range limits in QueryPipe
• All transform steps implement new TransformPipe interface (marker interface)
• Various thread safety test cases added for GremlinGroovyScriptEngine
• Significant performance improvements in GremlinGroovyScriptEngine (mapped to GroovyScriptEngine)

---

Version 2.1.0 (August 4, 2012)

<dependency>
  <groupId>com.tinkerpop.gremlin</groupId>
  <artifactId>gremlin-java</artifactId>
  <version>2.1.0</version>
</dependency>

<dependency>
  <groupId>com.tinkerpop.gremlin</groupId>
  <artifactId>gremlin-groovy</artifactId>
  <version>2.1.0</version>
</dependency>

• Optimized the pom.xml for both the parent and child projects
• Added Travis continuous integration support
• Restructured the distribution.xml to provide a more “standard” looking directory structure
• Provide support for Query.limit() in query optimization
• Added import static support for Direction.*
• Made gremlin-groovy.sh/bat into simply gremlin.sh/bat

---

Version 2.0.0 (May 23, 2012)

<dependency>
  <groupId>com.tinkerpop.gremlin</groupId>
  <artifactId>gremlin-java</artifactId>
  <version>2.0.0</version>
</dependency>

<dependency>
  <groupId>com.tinkerpop.gremlin</groupId>
  <artifactId>gremlin-groovy</artifactId>
  <version>2.0.0</version>
</dependency>

• Removed Gremlin-Scala from the main Gremlin distribution
• Added tree-step based on the TreePipe to support representing the branch of a traversal
• Added order-step based on the OrderPipe to support in-stream ordering
• Removed VerticesPipe and EdgesPipes as should be called via getVertices() and getEdges() in GremlinPipeline constructor
• Added QueryPipe support automated optimizations around Vertex.query()
• Added GremlinPipeline.enablePath() to allow for non-automated path enabling
• Provided GremlinPipeline.optimize(boolean) to control whether the pipeline is query optimized
• The Imports class now provides static imports for Direction and Transaction enums
• Gremlin-Groovy
  • Renamed paths-step to simply path as now in Pipes the method is called Pipe.getCurrentPath()
  • Bumped to support GMaven 1.4

---

Gremlin 1.X

Version 1.5 (No More Mr. Nice Guy — February 28, 2012)

<dependency>
  <groupId>com.tinkerpop.gremlin</groupId>
  <artifactId>gremlin-java</artifactId>
  <version>1.5</version>
</dependency>

<dependency>
  <groupId>com.tinkerpop.gremlin</groupId>
  <artifactId>gremlin-groovy</artifactId>
  <version>1.5</version>
</dependency>

<dependency>
  <groupId>com.tinkerpop.gremlin</groupId>
  <artifactId>gremlin-scala</artifactId>
  <version>1.5</version>
</dependency>

• Added select step for the respective SelectPipe
• Added has and hasNot step for concise property/id/label filtering
• Removed idFilter step, where has and hasNot is the respective replacement
• Removed labelFilter step, where has and hasNot is the respective replacement
• Removed propertyFilter step, where has and hasNot is the respective replacement
• Removed index step as it was not used and there are other ways of getting that functionality
• Altered the semantics of groupCount value function such that it now takes a Pair (much more expressive)
• Fixed classpath issues in the .bat files of respective Gremlin implementations
• Lots of JavaDoc work to make the GremlinPipeline JavaDoc be a single point of reference
• Updated distribution.xml files to support data/ and LICENSE.txt inclusion in sub-modules
• Added groupBy step for the underlying GroupByPipe and GroupByReducePipe in Pipes (support for MapReduce)
• Gremlin-Groovy
  • Removed the JSONSlurper vertex and edge overloaded methods
  • Gremlin.load() now makes use of method reflection to build the language tokens
  • Increased the performance of Element.key by approximately a factor of 10

---

Version 1.4 (Ain’t No Thing But a Chicken Wing — December 7, 2011)

<dependency>
  <groupId>com.tinkerpop.gremlin</groupId>
  <artifactId>gremlin-java</artifactId>
  <version>1.4</version>
</dependency>

<dependency>
  <groupId>com.tinkerpop.gremlin</groupId>
  <artifactId>gremlin-groovy</artifactId>
  <version>1.4</version>
</dependency>

<dependency>
  <groupId>com.tinkerpop.gremlin</groupId>
  <artifactId>gremlin-scala</artifactId>
  <version>1.4</version>
</dependency>

• Generalized Gremlin to be a meta-language
  • Added gremlin-test as a universal test suite for all Gremlin implementations
  • Added gremlin-java as the Java representation of the Gremlin language
  • Added gremlin-groovy as the Groovy representation of the Gremlin language
  • Added gremlin-scala as the Scala representation of the Gremlin language
  • Added IndexElementsPipe as a start pipe that pulls elements from an index
  • Renamed uniqueObject step to dedup
  • Renamed orFilter and andFilter to or and and, respectively
  • Renamed uniquePath to simplePath
• Added store step to support lazily loading of objects into collection
• Added dedup{closure} to allow for uniqueness determination as a function
• gremlin-groovy updates
  • Renamed Graph.saveGraphJSON() to Graph.saveGrapSON() (likewise for load method)
  • No more _() required off of graph-based objects
  • No more [[ ]] filtering supported
  • No more support for >> syntax
    • >>1 is next()
    • >>2 is next(2)
    • >>-1 is iterate()
    • >>[] is toList()
    • >>collection is fill(collection)

---

Version 1.3 (On the Case — September 18, 2011)

<dependency>
  <groupId>com.tinkerpop</groupId>
  <artifactId>gremlin</artifactId>
  <version>1.3</version>
</dependency>

• Added SailGraph.loadRDF() meta method to make it easier to load RDF files
• Optimized >>-1 to use a try/catch model which is more efficient given the architecture of AbstractPipe
• Updated gremlin.sh to respect previous classpaths and to make sym-linking easier
• Fixed table width bug in Table
• Added Graph.loadGraphJSON() and Graph.saveGraphJSON() with respective data/graph-example-1.json
• Moved to FluentPipeline model provided by Pipes
• Filter functions must be made explicit now (e.g. out{it.id > 2} must now be out.filter{it.id > 2})
• Migrated TablePipe and Table to Pipes
• Renamed propFilter step to propertyFilter
• Added multi-vertex/edge lookup support via g.v(1,2,3) -> [v[1],v[2],v[3]] (and g.e() as well)
• Added memoize() step to provide the MemoizePipe functionality to Gremlin
• Added support for simpler PipeFunction model introduced by Pipes
• Bumped to Groovy 1.8.2

---

Version 1.2 (New Sheriff in Town — August 1, 2011)

<dependency>
  <groupId>com.tinkerpop</groupId>
  <artifactId>gremlin</artifactId>
  <version>1.2</version>
</dependency>

• Added respective pipe support for Vertex.getOutEdges(String...) and Vertex.getInEdges(String...)
• Added path functions to paths to allow for post-processing a returned path
• Added map entry functions to groupCount to allow for pre-processing an entry’s key/value pairs
• Added Gremlin.version() to get version information (returns GremlinTokens.VERSION)
• Added Gremlin.defineStep() to make it simple for developers to add new steps
• Added support for Graph.loadGraphML(URL)
• Renamed ifelse to ifThenElse and provided ‘scatter’ functionality if branch yields an iterator/iterable
• Implemented Pipe.reset() method for all Gremlin-specific pipes
• Migrated all Gremlin-specific pipes to Pipes
  • Added GroovyPipeClosure to map a Groovy Closure to a PipeClosure
• Update g.idx() to reflect null behavior of Blueprints index API update
• Added map step to emit an element’s property map
• Added support for aggregate step function to process object prior to insertion into the aggregate collection
• Bumped to Groovy 1.8.1

---

Version 1.1 (Clownin’ Around — June 15, 2011)

<dependency>
  <groupId>com.tinkerpop</groupId>
  <artifactId>gremlin</artifactId>
  <version>1.1</version>
</dependency>

• The .gremlin_history file is always saved to the users home directory
• Supports an inclusive max on RangeFilterPipe
• Supports integer range filtering for Map — e.g. m[0..10]
• Added meta methods loadGraphML and saveGraphML to Graph
• SailGraph is no longer a strict dependency (however, loads meta methods if in classpath)
• Added optional step for doing back, but without filtering
• Extended support for graph pattern matching
• Added AsPipe for naming steps.
  • Added table step with respective TablePipe and Table data structure
  • Added support for named steps in loop
  • Added support for named steps in back
  • Added support for named steps in optional

---

Version 1.0 (Gremlin — May 8, 2011)

<dependency>
  <groupId>com.tinkerpop</groupId>
  <artifactId>gremlin</artifactId>
  <version>1.0</version>
</dependency>

• Added out(label?), in(label?), and both(label?)
• Added copySplit, fairMerge, exhaustMerge
• SailLoader is loaded on startup (no longer required by developer)
• Faster Gremlin.compile() using GremlinScriptEngine
• Renamed andf, orf, and propf to andFilter, orFilter, and propFilter
• Moved to the more optimal vertex/edge incidence Pipes
• Renamed unique to uniqueObject
• Updated Gremlin-specific pipes to support reset() API

---

Gremlin 0.X

Version 0.9 (Gremlin the Grouch — April 4, 2011)

<dependency>
  <groupId>com.tinkerpop</groupId>
  <artifactId>gremlin</artifactId>
  <version>0.9</version>
</dependency>

• Iterable and Iterator meta methods are now Pipe exclusive
• Index results are auto-wrapped in an IdentityPipe to ensure consistent behavior across graphs
• Added DexGraph as imported graph database
• Object.propertyMissing is no longer defined to ensure Gremlin safety
• emit step renamed to transform
• foreach step renamed to sideeffect
• Removed futuref step as back supplies necessary computation
• The it of the gather step is now the gathered List

---

Version 0.8 (Grem Stefani — March 2, 2011)

<dependency>
  <groupId>com.tinkerpop</groupId>
  <artifactId>gremlin</artifactId>
  <version>0.8</version>
</dependency>

• Added new steps
  • emit to emit results of provided function
  • ifelse to perform a if/then/else branch
  • uniquePath to filter looping paths
• Added new methods
  • Object.mean() to calculate the average of delegate
  • Pipe.count() to calculate the number of objects in pipe
• Added support for stack trace printing during an error in the console
• loop step now requires “back integer”
• Added step caching model to increase speed of property notation
• Added outE(label), inE(label) and bothE(label) support
• Added support for dynamic counter adjustment in groupCount
• Bumped to using Groovy 1.7.8
• Created a type-cast friendly groupCount step

---

Version 0.7 (Gremopoly — January 22, 2011)

<dependency>
  <groupId>com.tinkerpop</groupId>
  <artifactId>gremlin</artifactId>
  <version>0.7</version>
</dependency>

• Added many new steps
  • loop step for in-line looping
  • foreach step for function based operations on flowing objects
  • back step for backtracking in pipeline (i.e. ../)
  • aggregate, except, and retain for except/retain pattern
  • groupCount for map aggregations (i.e. g:op-value())
  • step for anonymous/lambda steps
• Massive refactoring of documentation
• Changed to Groovy as the host language

---

Version 0.6 (Gremalicious — December 14, 2010)

<dependency>
  <groupId>com.tinkerpop</groupId>
  <artifactId>gremlin</artifactId>
  <version>0.6</version>
</dependency>

• Fixed foreach bug that occurred on single object looping
• Fixed if/else bug that occurred on function call
• null and empty iterator/iterable result now printed in console
• Empty path expression returns an empty iterable, not null
• Added g:copy-e() function that conveniently copies an element from one graph into another
• Fixed numerous bugs in how functions were computing atoms
• GPath optimization through auto-reset of Gremlin-specific range-based pipes
• Comparing numbers is now possible without casting
• Fixed a major bug in error message handling (many exceptions w/o message being provided in console)
• Improved Gremlin evaluator load time with static function and step loading
• Added support for Blueprints 0.3 multi-indexes
  • g:add-aidx-key(), g:add-idx(), g:drop-idx(), g:show-aidx-keys(), g:show-idx(), g:remove-aidx-keys()
• Added return ?object to return from a user defined function or step
• Added support for path generation through g:path() (main body of work in Pipes)

---

Version 0.5.5 (Gremlivich — October 28, 2010)

<dependency>
  <groupId>com.tinkerpop</groupId>
  <artifactId>gremlin</artifactId>
  <version>0.5.5</version>
</dependency>

• A rework of the user-defined step model
  • Renamed the concept of “path” to “step”
  • Users can now define computations in steps
• Added gather and scatter steps to allow for aggregation and unrolling of objects within a path
• Added g:append() function to concatenate objects to an iterable (e.g. list, set)

Version 0.5 (Gremlivich — September 22, 2010)

<dependency>
  <groupId>com.tinkerpop</groupId>
  <artifactId>gremlin</artifactId>
  <version>0.5</version>
</dependency>

• A complete redesign and implementation of the Gremlin compiler
  • Compiler written using ANTLR 3.2
  • Virtual machine written using Pipes 0.1 and JSR 223.
• Significant performance improvements
  • Order of magnitude increase in speed for numerous operations
  • Lazy evaluation for better memory usage
• Functions implement the ServiceLoader pattern for ease of incorporating 3rd-party extensions
• Numerous changes to the Gremlin function library
  • All core functions are now in the g namespace (e.g. g:count(), g:string())
  • g:id() changed to g:id-v() and g:id-e() for getting vertices and edges, respectively
  • g:key() changed to g:key-v()
  • Added g:version() function to retrieve version information
  • Added g:g() (group) to semi-replace () model in Gremlin 0.2.2
  • Added g:flatten() to flatten embedded iterables
  • Added g:includes() to determine if an object is in an iterable
  • Added g:power() to implement raising a number by a power
  • Added g:range() to allow for integer ranges to be created
  • Changed g:difference() to g:diff() (abbreviation policy enacted for function names)
  • Added g:set() to create a set (unordered, unique collection of objects)
  • Added g:integer(), g:long(), g:float(), and g:double() to support number casting
• Additions to the Gremlin type system
  • Supports iterable as a super type in order to have direct support for lazy evaluation
  • Supports set as an unordered, unique collection of objects
  • Supports integer (1), long (1l), float (1.0), double (1.0d) as different number types
• Print streams can be dynamically set in GremlinScriptContext
• true(), false(), and null() are now keywords—true, false, null
• String concatenation is supported with the + operation
• Range filter support through [1..10] type syntax
• Index of lists starts at 0 instead of 1 as in Gremlin 0.2.2
• Test cases now provide performance statistics
• Removed the ‘show all aspects’ * step (i.e. ./outE/inV/*)
• Allow for the use of assignment and other operations within a path (e.g. ./outE/inV[g:p($x := .)]
• Provided support for OrientDB graph database (main body of work in Blueprints)
• Update to dependency inclusions
  • Distribution includes all supported graph databases in Blueprints (gremlin-*-standalone.jar)
  • Dependency provide <scope>provided</scope>

---

Version 0.2.2 (Gremicide — March 7, 2010)

<dependency>
  <groupId>com.tinkerpop</groupId>
  <artifactId>gremlin</artifactId>
  <version>0.2.2</version>
</dependency>

• Fixed a bug introduced in 0.2.1 having to do with variables copied from a parent path context
• Added dependency on Neo4j RDF and SAIL SNAPSHOTS
• Added Neo4jSailGraph BETA to the core distribution

Version 0.2.1 (Gremicide — February 25, 2010)

<dependency>
  <groupId>com.tinkerpop</groupId>
  <artifactId>gremlin</artifactId>
  <version>0.2.1</version>
</dependency>

• Instance naming convention for Functions changed to FunctionLibrary
• list g:list(object+) is now list g:list(object*)
• Added support for Java 1.6+ JSR 223 ScriptEngine and ScriptEngineFactory
  • Renamed GremlinEngine to GremlinScriptEngine
  • Renamed GremlinEngineFactory to GremlinScriptEngineFactory

Version 0.2 (Gremicide — February 14, 2010)

<dependency>
  <groupId>com.tinkerpop</groupId>
  <artifactId>gremlin</artifactId>
  <version>0.2</version>
</dependency>

• Added user defined functions
  • Gremlin and Java defined functions supported
• Added user defined paths
  • Gremlin and Java defined functions supported
• Added .gremlinrc load script
• Added support for manual transaction handling of Neo4j graphs
  • boolean neo4j:start-tx(), boolean neo4j:stop-tx(boolean) added
• Added support for embedded maps and lists
• Added support for JSON encoding and decoding of Gremlin types
  • g:json(object) was removed
  • string g:to-json(object) and object g:from-json(string) added
• Added better support for vertex and edge property indexing
• Added working graph $_g variable and updated graph functions to support it
• Added support for SPARQL queries for SAIL-based graphs
  • list sail:sparql(graph?, string) was added
• Added support for Linked Data SAIL for real-time analysis of the Web of Data
• Added support for script loading
  • script gremlin-script.grm
• Added support for JUNG algorithms over Gremlin graphs
  • map jung:pagerank(graph?, map?)
  • list jung:dijkstra(graph?, vertex, vertex, map?)
• Renamed g:append(object, object+) to g:list(object+)
• Renamed the general graph model to property graph model
  • Moved the model and its implementations to Blueprints
  • Dependency is on Blueprints 0.1
• Various bug fixes and optimizations

---

Version 0.1 (XMas Alpha — December 25, 2009)

<dependency>
  <groupId>com.tinkerpop</groupId>
  <artifactId>gremlin</artifactId>
  <version>0.1</version>
</dependency>

• Initial release of the language and virtual machine