[
https://issues.apache.org/jira/browse/TINKERPOP-2895?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17701219#comment-17701219
]
Stephen Mallette commented on TINKERPOP-2895:
---------------------------------------------
> So do you mean I need to do it like:
> g.V().match(_.as('a').hasNot('vp2'),_.as('a').filter(both('e1').has('vp4'))).select('a')?
i'm not saying that you need to change the query. i was just trying to explain
how {{{match()}} works. I think the real explanation is my text representation
of the query: "find me all vertices that do not have a property of 'vp2' and
that have at least one edge to a vertex with a property of 'vp4'".
You now have a new example, but the explanation is the same:
1. `g.V().in('el2','el1').both('el1')` returns `[0, 11, 19, 4, 6]`.
2. `g.V().both('el2','el1')` returns `[0, 11, 13, 19, 4, 6]`.
3.
`g.V().match(_.as('a').in('el2','el1').both('el1'),_.as('a').filter(both('el2','el1'))).select('a')`
returns `[0, 11, 13, 19, 4]`.
The english representation of the new examples are:
1. For each vertex traverse "el2"and "el1" edges and then traverse to "el1"
edges and return those vertices
2. For each vertex traverse "el2" and "el1" edges and then return those vertices
3. Find me vertices where you can traverse "el2" or"el1" edges to then traverse
to other "el1" edges and that also can traverse one "el2" or "el1" edge.
The results of 1 and 2 can't be used to determine the results of 3. I'm not
sure how else to explain it i'm afraid. I think you're misunderstanding
something fundamental to how {{match()}} works.
Here's a simpler example and I'd encourage you to spend more time looking for
far more simple ways to represent these problems because i think it will help
lead to understanding:
{code}
gremlin> g = TinkerFactory.createModern().traversal()
==>graphtraversalsource[tinkergraph[vertices:6 edges:6], standard]
gremlin> g.V().out('created')
==>v[3]
==>v[5]
==>v[3]
==>v[3]
gremlin> g.V().out('knows')
==>v[2]
==>v[4]
{code}
Given how you've been figuring your expected results this should return no
results for {{match()}} the way you've been doing it, right?
{code}
gremlin> g.V().match(__.as('a').out('created'),
__.as('a').out('knows')).select('a')
==>v[1]
==>v[1]
{code}
Why does it return v[1]? Because the traversal is not saying "give me an
intersection of the results of the two sub-queries". It's saying, "Find me the
vertices that have at least one outgoing 'created' edge and that have at least
one outgoing 'knows' edge". v[1] is the only vertex in the "modern" graph that
satisfies those conditions.
> Merged query using logical operator AND returns false results
> -------------------------------------------------------------
>
> Key: TINKERPOP-2895
> URL: https://issues.apache.org/jira/browse/TINKERPOP-2895
> Project: TinkerPop
> Issue Type: Bug
> Components: driver, server
> Affects Versions: 3.6.2
> Environment: - TinkerGraph Version: 3.6.2
> - Operating system: macOS 13.2.1
> - API/Driver: Java
> Reporter: Zeyang Zhuang
> Priority: Major
>
> We discovered a bug that Merged query using logical operator AND returns
> false results.
> - TinkerGraph Version: 3.6.2
> - Operating system: macOS 13.2.1
> - API/Driver: Java
> *Expected behavior:*
> We construct the following scenario: we randomly generate two queries Q1, Q2,
> and merge these two queries using AND logical operator into a new query Q3.
> Based on the AND calculation. The Q3 query result set should be the
> intersection of result sets from Q1 and Q2.
> We generate graph schema and data based on random strings and values. Here is
> one of our examples that triggered the bug.
> 1) `g.V().outE('el0','el2').bothV()` returns `[0, 0, 11, 11, 13, 13, 13, 13,
> 18, 18, 18, 18, 18, 2, 2, 2, 2, 2, 2, 2, 6, 6, 6, 6, 6, 8, 8, 8]`
> 2) `g.V().hasLabel('vl1','vl0')` returns `[16, 18, 2, 21, 4, 6, 8]`
> 3) `g.V().and(outE('el0','el2').bothV(),hasLabel('vl1','vl0'))` returns `[18,
> 2, 6]`.
> We calculate the intersection result set of Q1 and Q2, which is `[18, 2, 6,
> 8]`.
> The intersection result set doesn't equal to Q3 result set.
> *Actual behavior:*
> The intersection result set should equal to Q3 result set. We did trigger
> some cases conform to this requirement, but still there're some cases that
> violate this constraint.
> *Steps to reproduce:*
> We create a graph with 10 nodes and 20 edges. We try to make it clear to
> reproduce the bugs, {*}{{*}}hope{{*}}{*} to not cause much inconvenience to
> your reviewing, but we believe the problem does exist.
> Following the following graph data generation query, we can reproduce the
> bugs:
> Create data
> ```
> Vertex:
> g.addV('vl2').property('vp2','0.11229777').property(T.id,0)
> g.addV('vl1').property('vp1','0.7532909965461835').property(T.id,2)
> g.addV('vl0').property('vp6','-882633195435277600').property(T.id,4)
> g.addV('vl1').property('vp4','7665355125606600882').property(T.id,6)
> g.addV('vl1').property('vp1','0.7532909965461835').property('vp4','3838824259062394782').property(T.id,8)
> g.addV('vl2').property('vp4','6730135576265295973').property(T.id,11)
> g.addV('vl2').property('vp3','-6896539503167143038').property('vp4','-2940639995931981142').property(T.id,13)
> g.addV('vl0').property('vp6','-2552831883676257311').property(T.id,16)
> g.addV('vl1').property('vp1','0.4100010612879974').property('vp4','-2768293334048120500').property(T.id,18)
> g.addV('vl0').property('vp6','8251747935808021903').property(T.id,21)
> Edge:
> g.V(0).as('0').V(2).as('2').addE('el1').from('0').to('2')
> g.V(8).as('8').V(6).as('6').addE('el0').from('8').to('6')
> g.V(0).as('0').V(18).as('18').addE('el1').from('0').to('18')
> g.V(6).as('6').V(13).as('13').addE('el2').from('6').to('13')
> g.V(18).as('18').V(2).as('2').addE('el0').from('18').to('2')
> g.V(18).as('18').V(0).as('0').addE('el2').from('18').to('0')
> g.V(6).as('6').V(18).as('18').addE('el0').from('6').to('18')
> g.V(2).as('2').V(11).as('11').addE('el2').from('2').to('11')
> g.V(8).as('8').V(13).as('13').addE('el2').from('8').to('13')
> g.V(11).as('11').V(8).as('8').addE('el1').from('11').to('8')
> g.V(18).as('18').V(13).as('13').addE('el2').from('18').to('13')
> g.V(8).as('8').V(0).as('0').addE('el2').from('8').to('0')
> g.V(6).as('6').V(11).as('11').addE('el2').from('6').to('11')
> g.V(11).as('11').V(18).as('18').addE('el1').from('11').to('18')
> g.V(2).as('2').V(6).as('6').addE('el0').from('2').to('6')
> g.V(13).as('13').V(6).as('6').addE('el1').from('13').to('6')
> g.V(0).as('0').V(6).as('6').addE('el1').from('0').to('6')
> g.V(2).as('2').V(13).as('13').addE('el2').from('2').to('13')
> g.V(2).as('2').V(18).as('18').addE('el0').from('2').to('18')
> g.V(2).as('2').V(2).as('2').addE('el0').from('2').to('2')
> ```
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
