Gremlin 查询中的条件执行

Question

Is there a way to conditionally execute part of a gremlin query, based on data external to the query, not based on the graph.

I'm using Go, the Gremlingo driver and AWS Neptune. I've been able to do this from my code, but it is not that pretty. Here's how I did it.

func (n NeptuneGremlinGraph) Put(assetID string, version string, records []les.DeltaEditRecord) error {
    g := gremlin.Traversal_().WithRemote(n.connection)
    anonT := gremlin.T__
    for _, r := range records {
        promise := n.addParent(g.V().HasLabel("Entity").
            Has("asset_id", assetID).
            Has("version", version).
            Has("entity_id", r.EntityID).
            Fold().
            Coalesce(anonT.Unfold(),
                anonT.AddV("Entity").
                    Property("asset_id", assetID).
                    Property("version", version).
                    Property("entity_id", r.EntityID)).
            Store("e").
            V().HasLabel("Component").
            Has("asset_id", assetID).
            Has("version", version).
            Has("entity_id", r.EntityID).
            Has("component_id", r.ComponentID).
            Fold().
            Coalesce(anonT.Unfold().
                Property("value", r.Value),
                anonT.AddV("Component").
                    Property("asset_id", assetID).
                    Property("version", version).
                    Property("entity_id", r.EntityID).
                    Property("component_id", r.ComponentID).
                    Property("value", r.Value)).
            AddE("ATTACHED_TO").To(anonT.Cap("e").Unfold()), anonT, "e", assetID, version, r).Iterate()
        err := <-promise
        if err != nil {
            return err
        }
    }

    return nil
}

func (n NeptuneGremlinGraph) addParent(graphTraversal *gremlin.GraphTraversal, anonT gremlin.AnonymousTraversal, e, assetID, version string, record les.DeltaEditRecord) *gremlin.GraphTraversal {
    if !n.hasParent(record) {
        return graphTraversal
    }

    parent := n.getParent(record)
    return graphTraversal.V().
        HasLabel("Entity").
        Has("asset_id", assetID).
        Has("version", version).
        Has("entity_id", parent).
        Fold().
        Coalesce(anonT.Unfold(),
            anonT.AddV("Entity").
                Property("asset_id", assetID).
                Property("version", version).
                Property("entity_id", parent)).
        AddE("CHILD_OF").From(anonT.Cap(e).Unfold())
}

It would be really slick if I could add a new method to gremlin.GraphTraversal, but I don't think you can add methods to a struct in another package. I suppose that I could also do it with two completely independent queries, in a transaction.

Just wondering if there is something I missed and that there is a way to do this within the query.

Answer 1

There are a couple of different approaches here. First, you can build up a query dynamically in code. As in your example above, if your second method accepted a graph traversal object as a parameter, that could be appended with additional steps before then being returned as a graph traversal object. Simple pseudocode here:

func AddFilter(query,parameter) {
    return query.has('key',parameter);
}

GraphTraversal new = g.V().has('somekey','somevalue');

queryResult = AddFilter(new,'value').next();

Graph traversals are not sent to the server until you append a Terminal Step at the very end.

A more involved way of doing this would be to create your own DSL in a way to "extend" the query language to simplify reused query components. The docs cover this in detail here: https://tinkerpop.apache.org/docs/current/reference/#gremlin-go-dsl