[TINKERPOP-1905] TinkerGraphComputer doesn't handle grouping side-effect steps properly

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 3.3.1
• Component/s: process
• Assignee: Unassigned
• Reporter: Daniel Kuppitz
• Created: 27/Feb/18 16:46
• Updated: 27/Feb/18 16:46

Description

As noted in https://github.com/apache/tinkerpop/pull/803, there seems to be an issue in how TinkerGraphComputer processes grouping side-effect steps. Both grouping steps require an extra barrier() step to function properly; other side-effect steps (e.g. aggregate() show the correct / expected behavior.

gremlin> g = TinkerFactory.createModern().traversal().withComputer()
==>graphtraversalsource[tinkergraph[vertices:6 edges:6], graphcomputer]
gremlin> g.V().group("a").by("name").select("a")
==>[]
==>[]
==>[]
==>[]
==>[]
==>[]
gremlin> g.V().groupCount("a").by("name").select("a")
==>[]
==>[]
==>[]
==>[]
==>[]
==>[]
gremlin> g.V().aggregate("a").by("name").select("a")
==>[lop,josh,peter,vadas,ripple,marko]
==>[lop,josh,peter,vadas,ripple,marko]
==>[lop,josh,peter,vadas,ripple,marko]
==>[lop,josh,peter,vadas,ripple,marko]
==>[lop,josh,peter,vadas,ripple,marko]
==>[lop,josh,peter,vadas,ripple,marko]

Including an extra barrier() which, in my opinion, really shouldn't be necessary in an OLAP job:

gremlin> g.V().group("a").by("name").barrier().select("a")
==>[josh:[v[4]],ripple:[v[5]],lop:[v[3]],peter:[v[6]],vadas:[v[2]],marko:[v[1]]]
==>[josh:[v[4]],ripple:[v[5]],lop:[v[3]],peter:[v[6]],vadas:[v[2]],marko:[v[1]]]
==>[josh:[v[4]],ripple:[v[5]],lop:[v[3]],peter:[v[6]],vadas:[v[2]],marko:[v[1]]]
==>[josh:[v[4]],ripple:[v[5]],lop:[v[3]],peter:[v[6]],vadas:[v[2]],marko:[v[1]]]
==>[josh:[v[4]],ripple:[v[5]],lop:[v[3]],peter:[v[6]],vadas:[v[2]],marko:[v[1]]]
==>[josh:[v[4]],ripple:[v[5]],lop:[v[3]],peter:[v[6]],vadas:[v[2]],marko:[v[1]]]
gremlin> g.V().groupCount("a").by("name").barrier().select("a")
==>[josh:1,ripple:1,lop:1,peter:1,vadas:1,marko:1]
==>[josh:1,ripple:1,lop:1,peter:1,vadas:1,marko:1]
==>[josh:1,ripple:1,lop:1,peter:1,vadas:1,marko:1]
==>[josh:1,ripple:1,lop:1,peter:1,vadas:1,marko:1]
==>[josh:1,ripple:1,lop:1,peter:1,vadas:1,marko:1]
==>[josh:1,ripple:1,lop:1,peter:1,vadas:1,marko:1]

A pull request for this ticket should also remove the extra barrier() step from the test case g_VX1X_groupXaX_byXconstantXaXX_byXnameX_selectXaX_selectXaX.

I haven't checked 3.2.x yet, but this line of code most likely shows the same issue.