JarCastingThe Object Graph Mapping Library for Kotlin and Gremlin
Gremlin is the graph traversal language for the Apache TinkerPop graph framework and is supported by most graph database implementations.
Basic Usage:
Define a Vertex
@Element("Person")
data class Person(
@ID
val id: Long? = null,
@Property("name")
val name: String)
Define a Relationship
val friends = ManyToManySymmetricEdgeSpec<Person>("friends")
val subordinates = SingleToManyEdgeSpec<Person, Person>("boss_to_subordinates")
val boss = subordinates.inverse
Save a Vertex
val mighael = graphMapper.saveV(Person(name = "Michael Scott"))
val dwight = graphMapper.saveV(Person(name = "Dwight Schrute"))
Save an Edge
graphMapper.saveE(friends from michael to dwight)
graphMapper.saveE(boss from michael to dwight)
Traverse an Edge
graphMapper.traverse(friends from michael) // returns List<Person> [ dwight ]
graphMapper.traverse(friends from dwight) // returns List<Person> [ michael ]
graphMapper.traverse(subordinates from michael) // returns List<Person> [ dwight ]
graphMapper.traverse(boss from dwight) // returns non-optional Person micheal
The graphMapper is an implementation of the GraphMapper interface which requires two properties:
- A
GraphTraversalSourcespawned from a tinkerpop graph - A
GraphDescriptionwhich is easily instantiated usingCachedGraphDescription(vertices = setOf(Person::class))
Sample Starwars App
An interactive example can be run using the starwars example project. From the repo root directory, run:
./gradlew run
Then load GraphiQL at http://localhost:5000/graphiql.html to explore the data mapped with this library.
A small typescript + apollo + react web client that uses this starwars API can also be sampled:
cd example/frontend && yarn start
Then load http://localhost:3000 (server must also be running)
Installation:
-
Gradle
compile 'com.github.pm-dev:kotlin-gremlin-ogm:0.21.0' -
Maven
<dependency> <groupId>com.github.pm-dev</groupId> <artifactId>kotlin-gremlin-ogm</artifactId> <version>0.21.0</version> </dependency>
Library Extensions
Mix and match the following extensions to this library
- JanusGraph: kotlin-janusgraph-ogm
- GraphQL: kotlin-gremlin-graphql
- ReactiveX: kotlin-rx-ogm
Features:
- Take full advantage of Kotlin's type-safety. Traversals return either a list, non-optional, or optional based on whether a relationship is defined as to-many, to-single, or to-optional, respectively.
- Map and filter traversals using your Kotlin objects.
- No runtime code generation. Other ogms use third-party libraries that generate new classes at runtime.
- External dependencies are limited to: Kotlin's standard library, the gremlin-driver, and slf4j.
- Annotation-based so you can bring your current POJO domain objects.
- Kotlin compiler plugins 'all-open' and 'no-arg' are not required.
Why use a graph database and ogm?
- Graph databases are powerful for modeling data that is highly connected.
- This OGM enables for strong typing of domain objects in the application layer while removing the need for a schema enforced by the db.
- This allows for data to be backed by a NoSQL datastore. NoSQL datastores are horizontally scalable and can be partition tolerant.
- This makes migrations much easier.
Limitations:
- This library will not work if you're trying to connect to a Gremlin Server remotely. This library creates traversals that call back into the library, thus, your graph implementation must be running within the same JVM.
- For this reason, connecting to Amazon Neptune is not currently supported, as Amazon Neptune does not support calling arbitrary java from within a traversal.
Design Principles:
- Common use-cases should be easy. Uncommon use-cases should be possible.
- Performance is important.
- Fail fast with helpful exceptions.
Native property types are stored directly in the graph as property values:
BooleanByteDoubleFloatIntLongString
If your Gremlin implementation does not support one of these native types, make sure to register a property mapper for they type with your GraphDescription or declare a @Mapper for that property.
Built-in property mappers:
To use other property types, register your custom property mapper with GraphDescription by returning a PropertyMapper from the getScalarPropertyMapper function a @Mapper for that property.
How the mapping works:
- A description of your graph, based annotations, is processed and cached when your
GraphMapperis instantiated. - Using this description of the graph, we can create 'vertex mappers' that know how to serialize/deserialize objects marked with
@Elementto/from the graph. - Nested objects are not natively supported by Gremlin. When mapping a nested object to properties of a vertex, the property key uses periods ('.') to denote a nested object. For example:
Given:
class Name(val first: String, val last: String)
class Person(val name: Name)
...is serialized in the graph using vertex properties:
"name.first" -> "Lionel"
"name.last" -> "Messi"
ListandSettypes are supported. These collections are stored as follows:
Given:
class Name(val first: String, val last: String)
class Person(val names: Set<Name>)
...is serialized in the graph using vertex properties:
"names.0.first" -> "Cassius"
"names.0.last" -> "Clay"
"names.1.first" -> "Muhammad"
"names.1.last" -> "Ali"
To preserve the difference between a null and empty collection or map, we use a special UUID token. For example if the names Set was empty:
"names" -> "474A56F1-6309-41B5-A632-AD53F57DBDAE"
...or in the case of an empty map, the special UUID is: 9B94DCB9-D405-47C1-B56D-72F83C4E81D3.
Legal:
Licensed under the Apache Software License 2.0. This code is in no way affiliated with, authorized, maintained, sponsored or endorsed by the Apache Software Foundation.