different behavior between 3.2.0-incubating and 3.2.3

[Damien Seguy]

Hi all,

The following query has different results 3.2.0-incubating and 3.2.3 : 

g.V().where( __.in('A').in('B').count().is(eq(0)) )

On 3.2.3, I always get 0 as results. Actually, I also get 0 when I use neq instead of ne

g.V().where( __.in('A').in('B').count().is(neq(0)) )

When I use only one in() inside the where(), I get consistent results, on both 3.2.3 and 3.2.0-incubating

g.V().where( __.in('A').count().is(eq(0)) ) // works as expected

g.V().where( __.in('B').count().is(eq(0)) ) // works as expected

g.V().where( __.in('B').count().is(neq(0)) ) // works as expected

I checked on 3.2.3, 3.2.2, 3.2.1 (couldn't compile on 3.2.0, only on 3.2.0-incubating) : they all have the same behavior.

I'm using the neo4j-gremlin plug-in from https://github.com/thinkaurelius/neo4j-gremlin-plugin.

I tested on Neo4j 2.3.9 and 2.3.7. I'm running java 1.8.0_121), on OSX.

I check neo4j/data/log/console.log, no error are mentioned there.

The data is the same. 

Is there anything I'm missing ? Is it forbidden to use twice in() in a where? 

May be, There is a better way to check that a node has no 'grand-parent'. 

Thanks for your attention.

Damien. 

[Damien Seguy]

Also checked on gremlinbin.com : 

g.V().where( __.in('created').in('knows').count().is(eq(0)) )

works as expected, though I don't know which version of Gremlin is on the site.

Also, I checked on Debian, and the behavior is the same as on OSX. 

I don't know where to go from here. 

Damien. 

[Robert Dale]

Looks related to this issue https://issues.apache.org/jira/browse/TINKERPOP-1391

Robert Dale

--
You received this message because you are subscribed to the Google Groups "Gremlin-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-users+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/e772d120-fcfe-4b9f-90cd-be388ca99fe7%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[Damien Seguy]

Hi Robert,

Good catch, yes, this looks like related : usage of where(), workaround by moving hasLabel() outside, 

and 3.2.0-incubating being immune. 

Although, this is marked as closed and resolved since september. It is also marked as fixed in 3.2.3,

but I am experiencing this with 3.2.3. 

My code only uses default behavior : no strategy is mentioned anywhere. 

May be that will cast a new light on the problem.

I'm reverting to 3.2.0-incubating for the moment. I'll try with a 3.2.4 when it is released.

Damien

Robert Dale

To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-user...@googlegroups.com.

[Damien Seguy]

Just a quick note : I tested gremlin 3.2.4 and I also get the same error. 

[Robert Dale]

I created an issue for this https://issues.apache.org/jira/browse/TINKERPOP-1638

Robert Dale

To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-users+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/ac779b45-d2ac-40d6-b9eb-f896305c0826%40googlegroups.com.

[Damien Seguy]

Hi. 

Just a follow up to say that Gremlin 3.2.5 doesn't display this problem anymore here. 

Robert Dale