Of all the changes here, the most compelling one is the change made to `GraphSONTypeIdResolver` (the rest occur in methods that I don't think get called frequently enough to really make a noticeable impact in performance - just guessing from a cursory glance at this PR). You provided a small general benchmark which is nice, but could you please specifically do a benchmark over the `GraphSONTypeIdResolver` to see what kind of performance improvement we get there as a result of this?
You can try to benchmark the it directly but you could probably also just instantiate a `ObjectMapper` for GraphSON 3.0 and fire a bunch of different GraphSON types at it in succession to do the benchmark. Just an example with a `Vertex` - the full list of GraphSON types is [here](http://tinkerpop.apache.org/docs/3.3.4/dev/io/) ```text gremlin> mapper = GraphSONMapper.build().version(GraphSONVersion.V3_0).create().createMapper() ==>org.apache.tinkerpop.shaded.jackson.databind.ObjectMapper@352e612e gremlin> x = mapper.writeValueAsString(g.V(1).next()) ==>{"@type":"g:Vertex","@value":{"id":{"@type":"g:Int32","@value":1},"label":"person","properties":{"name":[{"@type":"g:VertexProperty","@value":{"id":{"@type":"g:Int64","@value":0},"value":"marko","label":"name"}}],"age":[{"@type":"g:VertexProperty","@value":{"id":{"@type":"g:Int64","@value":1},"value":{"@type":"g:Int32","@value":29},"label":"age"}}]}}} gremlin> v = mapper.readValue(x, Vertex.class) ==>v[1] ``` if you demonstrate a speed improvement there that would be huge :smile: [ Full content available at: https://github.com/apache/tinkerpop/pull/1001 ] This message was relayed via gitbox.apache.org for dev@tinkerpop.apache.org