Arthur Keen created JENA-814:
--------------------------------
Summary: Duration comparison returning incorrect results 28D = 1M
and 365D = 1Y
Key: JENA-814
URL: https://issues.apache.org/jira/browse/JENA-814
Project: Apache Jena
Issue Type: Bug
Components: ARQ
Affects Versions: Fuseki 1.1.1, Fuseki 1.0.2, Fuseki 1.0.1
Environment: OS X 10.10.1
Reporter: Arthur Keen
When comparing equality of 28 days to 1 month and 365 days to 1 year in Fuseki
1.1.1 and 1.0.1, Fuseki responds that they are the same, whereas the answer is
indeterminate.
How to reproduce:
SELECT ("P28D"^^<http://www.w3.org/2001/XMLSchema#duration>
="P1M"^^<http://www.w3.org/2001/XMLSchema#duration> as ?same) {}
--------
| same |
========
| true |
--------
SELECT ("P365D"^^<http://www.w3.org/2001/XMLSchema#duration>
="P1Y"^^<http://www.w3.org/2001/XMLSchema#duration> as ?same) {}
--------
| same |
========
| true |
--------
Reference:
The following table shows the strongest relationship that can be determined
between example durations. The symbol <> means that the order relation is
indeterminate. Note that because of leap-seconds, a seconds field can vary from
59 to 60. However, because of the way that addition is defined in Adding
durations to dateTimes, they are still totally ordered.
Relation
P1Y > P364D <> P365D <> P366D < P367D
P1M > P27D <> P28D <> P29D <> P30D <> P31D < P32D
P5M > P149D <> P150D <> P151D <> P152D <> P153D < P154D
Implementations are free to optimize the computation of the ordering
relationship. For example, the following table can be used to compare durations
of a small number of months against days.
However, Fuseki seem to have implemented this ordering differently.
According to Fuseki,
P1M == P28D => true - not OK(should be indeterminate according to
specification)
P1M <> P29D - OK
P1M <> P30D - OK
P1M <> P31D - OK
P1M < P32D - OK
P1Y == P365D -> true - not OK (should be indeterminate according to
specification)
P1Y <> P366D - OK
P1Y < 367D OK
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
