GraphSON Reader and Writer Library
<dependency>
<groupId>com.tinkerpop.blueprints</groupId>
<artifactId>blueprints-core</artifactId>
<version>??</version>
</dependency>The GraphSON reader and writer package allows an entire graph to be streamed to and from the standard JSON format which is utilized across the TinkerPop stack. The format comes in two flavors: one that uses explicit data typing within the JSON and one that does not. For most scenarios, standard JSON without data typing should generally be acceptable. Using the more verbose outputting of explicit data types only provides the added value of ensuring that numeric values are converted properly (ie. float versus double).
The following example shows the format without explicit data types:
{
"graph": {
"mode":"NORMAL",
"vertices": [
{
"name": "lop",
"lang": "java",
"_id": "3",
"_type": "vertex"
},
{
"name": "vadas",
"age": 27,
"_id": "2",
"_type": "vertex"
},
{
"name": "marko",
"age": 29,
"_id": "1",
"_type": "vertex"
},
{
"name": "peter",
"age": 35,
"_id": "6",
"_type": "vertex"
},
{
"name": "ripple",
"lang": "java",
"_id": "5",
"_type": "vertex"
},
{
"name": "josh",
"age": 32,
"_id": "4",
"_type": "vertex"
}
],
"edges": [
{
"weight": 1,
"_id": "10",
"_type": "edge",
"_outV": "4",
"_inV": "5",
"_label": "created"
},
{
"weight": 0.5,
"_id": "7",
"_type": "edge",
"_outV": "1",
"_inV": "2",
"_label": "knows"
},
{
"weight": 0.4000000059604645,
"_id": "9",
"_type": "edge",
"_outV": "1",
"_inV": "3",
"_label": "created"
},
{
"weight": 1,
"_id": "8",
"_type": "edge",
"_outV": "1",
"_inV": "4",
"_label": "knows"
},
{
"weight": 0.4000000059604645,
"_id": "11",
"_type": "edge",
"_outV": "4",
"_inV": "3",
"_label": "created"
},
{
"weight": 0.20000000298023224,
"_id": "12",
"_type": "edge",
"_outV": "6",
"_inV": "3",
"_label": "created"
}
]
}
}The following example shows the format with explicit data types:
{
"mode":"EXTENDED",
"vertices": [
{
"name": {
"type": "string",
"value": "lop"
},
"lang": {
"type": "string",
"value": "java"
},
"_id": "3",
"_type": "vertex"
},
{
"name": {
"type": "string",
"value": "vadas"
},
"age": {
"type": "integer",
"value": 27
},
"_id": "2",
"_type": "vertex"
},
{
"name": {
"type": "string",
"value": "marko"
},
"age": {
"type": "integer",
"value": 29
},
"_id": "1",
"_type": "vertex"
},
{
"name": {
"type": "string",
"value": "peter"
},
"age": {
"type": "integer",
"value": 35
},
"_id": "6",
"_type": "vertex"
},
{
"name": {
"type": "string",
"value": "ripple"
},
"lang": {
"type": "string",
"value": "java"
},
"_id": "5",
"_type": "vertex"
},
{
"name": {
"type": "string",
"value": "josh"
},
"age": {
"type": "integer",
"value": 32
},
"_id": "4",
"_type": "vertex"
}
],
"edges": [
{
"weight": {
"type": "float",
"value": 1
},
"_id": "10",
"_type": "edge",
"_outV": "4",
"_inV": "5",
"_label": "created"
},
{
"weight": {
"type": "float",
"value": 0.5
},
"_id": "7",
"_type": "edge",
"_outV": "1",
"_inV": "2",
"_label": "knows"
},
{
"weight": {
"type": "float",
"value": 0.4000000059604645
},
"_id": "9",
"_type": "edge",
"_outV": "1",
"_inV": "3",
"_label": "created"
},
{
"weight": {
"type": "float",
"value": 1
},
"_id": "8",
"_type": "edge",
"_outV": "1",
"_inV": "4",
"_label": "knows"
},
{
"weight": {
"type": "float",
"value": 0.4000000059604645
},
"_id": "11",
"_type": "edge",
"_outV": "4",
"_inV": "3",
"_label": "created"
},
{
"weight": {
"type": "float",
"value": 0.20000000298023224
},
"_id": "12",
"_type": "edge",
"_outV": "6",
"_inV": "3",
"_label": "created"
}
]
}There are a few differences between the formats. If types are embedded, the JSON must start with a mode key with a value of EMBEDDED. This key acts as a hint to the reader that it must extract property values in a different manner. If the key is omitted, that setting is assumed to be NORMAL.
There is a third option for the mode, which is COMPACT. This mode allows more complete control over exactly what element properties get serialized to the GraphSON output. As there is complete control, there is a possibility that the GraphSON will not be viable for reading (ie. if an important property is not serialized out, like the vertex _id).
All values of keys, short of the values for reserved keys that start with an underscore, must contain an object that has two keys: type and value. The type must be one of the following: boolean, string, integer, long, float, double, list, or map.
If the type is a map or a list, then each component object that make up that key must use that same format. For example:
"someMap": {
"name": {"type":"string", "value":"william"},
"age": {"type":"int", "value":76}
},
"someList" [{"type":"int", "value":1},{"type":"int", "value":2},{"type":"int", "value":3}]Please note that complex objects stored as properties will be converted to strings by way of the object’s toString method.
To output a graph in JSON format, pass the graph into the GraphSONWriter constructor, then call outputGraph:
Graph graph = ...
OutputStream out = ...
GraphSONWriter.outputGraph(graph, out);The GraphSONReader works in a similar format. Simply pass what would likely be an empty graph into the constructor, then call inputGraph:
Graph graph = ...
InputStream in = ...
GraphSONReader.inputGraph(graph, in);There are a number of static method overloads that offer more options and control.
The GraphSONUtility class is used by both GraphSONReader and GraphSONWriter to convert individual graph elements (vertices and edges) to and from the GraphSON format with conversion options to both a Jettison JSONObject and a Jackson ObjectNode. Usage is as follows:
Vertex v = graph.getVertex(1);
JSONObject json = GraphSONUtility.jsonFromElement(v);
System.out.println(json.toString())
Vertex convertedBack = GraphSONUtility.vertexFromJson(json,
new GraphElementFactory(graph), GraphSONMode.NORMAL, null);The GraphElementFactory is an implementation of the ElementFactory class, that utilizes a Graph instance to construct Vertex and Edge instances. In most cases, the GraphElementFactory is all that is needed to use the GraphSONUtility, though in cases where vertices or edges need to be constructed outside of the context of a Graph implementation, it might be necessary to implement a custom implementation.