Gremlin API: how to traverse vertices, collect properties conditionally, and stop when collected properties reach a certain limit

Question

I'm new to graphs and Gremlin API although I find it promising for my purpose so I am investigating and despite looking at documentation and several tutorials, I can't seem to find how to do this.

A portion of my graph would have vertices representing "documents" with "id" (1, 2, 3, ...) and "author" (a: Luc, Kim, ...) properties. They are linked with simple "next" edges. Like this:

V1{a:'Luc'} -> V2{a:'Kim'} -> V3{a:'Marc'} -> V4{a:'Kim'} -> V5{a:'Luc'} -> V6{a:'Luc'}

What I am trying to do:

• start at any known location from the vertice id (eg.: V(2))
• go either "up" with in() or "down" with out() (eg.: out())
• somehow "collect" the ids of vertices that have their author part of an array (eg.: ['Luc', 'Kim'])
• keep going until I have collected at most a certain number of ids (which will, in most cases be smaller than the number of vertices I will traverse to collect them), or until I reach the end (eg.: maximum 3)

So for this example, I would expect the result to be: [2, 4, 5].

From what I found so far, I would have something like:

g.V(2).repeat(somethingThatKeepsTheIdIfAuthorIsInArray(['Luc', 'Kim']).out()).until("I have found 3 ids")

Or if I word it differently, I want to "skip" vertex 3 because its author is not in the list, but keep going until I find at most 3 vertices that match my condition or I reach then end.

Any idea of what step(s) I should be looking at to accomplish this ?

Answer 1

Using this graph

g.addV('Document').property('author','Luc').property(id,'D1').as('d1').
  addV('Document').property('author','Kim').property(id,'D2').as('d2').
  addV('Document').property('author','Marc').property(id,'D3').as('d3').
  addV('Document').property('author','Kim').property(id,'D4').as('d4').
  addV('Document').property('author','Luc').property(id,'D5').as('d5').
  addV('Document').property('author','Luc').property(id,'D6').as('d6').
  addE('NEXT').from('d1').to('d2').
  addE('NEXT').from('d2').to('d3').
  addE('NEXT').from('d3').to('d4').
  addE('NEXT').from('d4').to('d5').
  addE('NEXT').from('d5').to('d6') 

We can write a Gremlin query that tests for certain author's names and uses a sideEffect to store the ones that match.

g.V('D2').
  repeat(out('NEXT').sideEffect(has('author',within('Luc','Kim')).id().store('documents'))).
  until(not(out('NEXT'))).
  select('documents')

Which will return

['D4', 'D5', 'D6']

With a small change to the query, we can also test for a maximum number of documents.

g.V('D2').
  repeat(out('NEXT').
         sideEffect(has('author',within('Luc','Kim')).id().store('documents'))).
  until(not(out('NEXT')).or().select('documents').count(local).is(2)).
  select('documents')

Which yields

['D4', 'D5']

To include the starting vertex in the results, we can just put the sideEffect first in the repeat step.

g.V('D2').
  repeat(out('NEXT').
         sideEffect(has('author',within('Luc','Kim')).id().store('documents'))).
  until(not(out('NEXT')).or().select('documents').count(local).is(2)).
  select('documents')

Which yields

['D2', 'D4', 'D5']

UPDATED 2021-11-17

In response to the comments/discussion below, here are two more examples of the query in use/ The fist starts at D2 with an is(3) and the second starts at D4. These were run using the Gremlin Console and TinkerGraph. While my query works, now that I fully understand your requirements, I think it is a bit over-engineered for your case. The answer you have since posted I think is fine for this case. The one caveat being, as we discussed in comments, that your query as written will fail on most TinkerPop enabled stores but seems to be working the way you need on CosmosDB. I up-voted your answer :-)

gremlin> g.V('D2').
......1>   repeat(out('NEXT').
......2>          sideEffect(has('author',within('Luc','Kim')).id().store('documents'))).
......3>   until(__.not(out('NEXT')).or().select('documents').count(local).is(3)).
......4>   select('documents')   

==>[D4,D5,D6]   



gremlin> g.V('D4').
......1>   repeat(out('NEXT').
......2>          sideEffect(has('author',within('Luc','Kim')).id().store('documents'))).
......3>   until(__.not(out('NEXT')).or().select('documents').count(local).is(3)).
......4>   select('documents')     

==>[D5,D6]  

Answer 2

I kept digging and found this to be a possible solution :

g.v('D2').emit(has('author', within('Luc', 'Kim'))).repeat(out()).limit(3).values('id').fold()

Which yields:

['D2', 'D4', 'D5']

I also confirmed that it works correctly no matter from which vertex we start. For example starting a D5 yields:

['D5', 'D6']

The only concern (question) I have about this is how the limit(3) is applied ? Does it actually break the repeat loop when 3 vertices are emitted or will the repeat loop traverse all vertices until there are no more "next" edge, THEN truncate the result to only return 3 ?