Dydra.com SPARQL tests
These tests define the Dydra.com SPARQL dialect. They are based heavily on the W3C spec as well as custom tests. They are the best way to determine when and why we deviate from the SPARQL standard.
For every known deviation from a W3C test, the W3C test remains intact alongside, and is tagged with an rspec tag. Thus, not every test in the repository is expected to pass. The failing tests are there for comparison purposes.
To run only the tests that we would expect to pass:
$ bundle install $ bundle exec rspec -cfn \ --tag '~reduced:all' \ --tag '~arithmetic:boxed' \ --tag '~blank_nodes:unique' \ --tag '~status:bug' \ --tag '~tz:zoned' \ spec/w3c/data-r2/
Known buggy tests are skipped with
--tag '~status:bug', to enable them,
remove the tag.
The tags above represent our known quirks. W3C tests remain alongside these tests, tagged differently, with the original W3C expectations instead of our adjusted ones.
This tag marks tests which fail because of normalization of numeric types.
Dydra.com treats the literal "01", and many other variables of untyped
literals, as a number. Some W3C tests require the store to return the original
value, in some cases after doing arithmetic comparisons on it. A SPARQL
arithmetic:boxed tests would return the non-normalized
This tag marks tests which have been modified (not necessarily failing) due to
Dydra.com's treatment of blank node identifiers as canonical. Tests tagged with
blank_nodes:unique would be passed by a SPARQL endpoint that generates new
blank node identifiers for each serialization/query/etc.
The REDUCED modifier may or may not reduce all results. The expectations for
reduced:all are not reduced. Dydra.com passes tests which remove
Dydra.com treats some input as the lexical representations of their real values. In particular, dates, times, and datetimes are always returned in Zulu time, regardless of their input format. Time zone information is not saved. In the same way, string literal language tags are normalized to a lowercase form.