W3C home > Mailing lists > Public > public-rdb2rdf-comments@w3.org > September 2011

R2RML: comments on spec and test cases

From: Toby Inkster <tai@g5n.co.uk>
Date: Mon, 26 Sep 2011 11:19:30 +0100
To: public-rdb2rdf-comments@w3.org
Message-ID: <20110926111930.22b45f2a@miranda.g5n.co.uk>
SPEC
====

Section 8 of the spec includes an example R2RML mapping which uses the
property rr:refObjectMap - however, this property isn't mentioned
anywhere else. I think rr:objectMap is intended.

The first example R2RML mapping in section 9 uses rr:graph where I
believe rr:constant is intended.


TESTS
=====

This test case file is invalid Turtle:

	D011-M2MRelations/r2rmlb.ttl

(Missing two semicolons.)

Many of the tests seem to test against the March 2011 draft of R2RML,
and should be updated. Although my implementation now follows the Sept
2011 draft, it still supports the March 2011 draft except where they
conflict. The major place they conflict is the default term type for
object maps (literal in March, IRI in September). Test cases really
need updating.

It would be useful if the tests didn't use schema-qualified table
names (e.g. "SCOTT.DEPT"). Although my implementation does support
schema-qualified table names, I've been running the test cases on
SQLite, which does not support schemas (at least, not in a normal way).
Given that there is no longer a rr:tableOwner predicate, support for
schemas doesn't seem to be an "feature" that an R2RML test suite
should be testing for - i.e. because it's a feature of the underlying
database, not a feature of R2RML.

-- 
Toby A Inkster
<mailto:mail@tobyinkster.co.uk>
<http://tobyinkster.co.uk>
Received on Monday, 26 September 2011 10:18:49 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 26 September 2011 10:18:49 GMT