Referring to this previously asked question: Default value when property is missing with project() step in Gremlin? Is there a more efficient way to do the same thing or preferably returning "null" in the latest gremlin version as each coalesce step takes 2 additional traversals?
Gremlin will not have a notion of a null traverser until 3.5.0 is released (there is no schedule for release at this time). Here are some of examples of what to expect from 3.5.0:
gremlin> g = TinkerGraph.open().traversal()
==>graphtraversalsource[tinkergraph[vertices:0 edges:0], standard]
gremlin> g.addV('person').property('name','allen')
==>v[0]
gremlin> g.V().has('person','name','allen').project('name','age').by('name').by('age')
==>[name:allen,age:null]
gremlin> g.V().has('person','name','allen').elementMap()
==>[id:0,label:person,name:allen]
gremlin> g.V().has('person','name','allen').elementMap('name','age')
==>[id:0,label:person,name:allen]
gremlin> g.V().has('person','name','allen').values('name','age')
==>allen
gremlin> g.V().has('person','name','allen').union(values('name'),values('age'))
==>allen
gremlin> g.V().has('person','name','allen').union(values('name'),coalesce(values('age'),constant(null)))
==>allen
==>null
There shouldn't be much loss in efficiency for using coalesce() even if you must use many of them. The cost of constant() is not an issue and values(String) assuming that's all you're doing will compile to a fast TokenTraversal. I suppose the big loss is having to type a lot. I'm not aware of any graph databases (maybe sqlg since it's based on relational backends?) that support default values as most don't have notion of a schema.