On 21/04/14 21:57, Andy Seaborne wrote:
On 21/04/14 21:21, Tim Harsch wrote:
Hmmm... it seems then that this test poses some challenge for
implementors
because how can you go both choose to return anything, and pass the test?
EricP did a comprehensive job of annotating the tests with the assumptions.
This goes back to SPARQL 1.0.
ARQ runs SPARQL tests (1.1 and 1.0) on every build. It does change
strict/non-strict on different parts of the test suite. That's the
point of extension points!
PS it is strictly impossible to pass the whole SPARQL 1.0 test suite!
There is a pair of tests that captures an ambiguity with FILTERS and
OPTIONAL with {{}} - but they aren't approved tests.
The interpretation of dateTime comparison for timezone/non-timezone
changes as well (as Holger and other found out).
F&O defines one thing (indeterminate if the difference is +/- 14 hours)
and XQuery another (insert a default timezone for some "default").
The trouble with default timezone is which one? The publisher of the
data? The SPARQL server? The client asking? On the web, they can all
be different. XQuery assumes they are the same.
Andy
