Graham Wallis created ATLAS-2197:
------------------------------------
Summary: Entities updated by changes in relationships depends on
tinkerpop version
Key: ATLAS-2197
URL: https://issues.apache.org/jira/browse/ATLAS-2197
Project: Atlas
Issue Type: Bug
Environment: Behavior detected during testing on Windows but probably
affects all platforms
Reporter: Graham Wallis
Priority: Minor
When running test:
mvn surefire:test -Dtest=AtlasRelationshipStoreSoftDeleteV1Test
-DfailIfNoTests=false -Ptitan1
there is a difference in the number of entities that are updated, compared to
the same test run with titan0.
The specific part of the test that is affected is the part that alters the
'friends' relationships between the person entity 'Max' and other person
entities 'John', 'Mike' and 'Julius'. Prior to the update, Max is friends with
Mike and John (only), Following the update, Max is friends with John and Julius
(only). It is obvious that Max is updated, as are Mike and Julius. With titan0
(and hence tinkerpop2) these 3 entities are updated, which makes sense
intuitively. However, when tested with titan1 (and hence tinkerpop3), the
number of updated entities is 4, because John also appears to be modified.
It is easy to change the testcase to tolerate either a result of 3 (for titan0)
or 4 (for titan1), but I think it would be worth investigating more precisely
*why* the difference occurs. It is most likely caused by the way we construct
the gremlin for TP3 compared to the earlier TP2 gremlin, but this has not been
investigated and there may be a related problem in the relationship code that
causes the additional and unexpected modification. It seems desirable that we
do not modify more entities than are strictly necessary.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
