W3C home > Mailing lists > Public > public-gld-comments@w3.org > July 2013

Re: RDF Data Cube validator. Problem with one integrity constraint: IC-12

From: Jose Emilio Labra Gayo <jelabra@gmail.com>
Date: Tue, 16 Jul 2013 23:23:25 +0200
Message-ID: <CAJadXXJ7UoD4Rcw4BMqGAEwRj7BSxeKd8B8G7V6rxd8FdM71Dw@mail.gmail.com>
To: Dave Reynolds <dave.e.reynolds@gmail.com>
Cc: public-gld-comments@w3.org
Thanks for looking at it. I have now been able to validate the data
and submitted the validation report.

It was interesting to confirm that the validator was using a slightly
different algorithm than simply running the specification queries.

Best regards, Jose Labra

On Tue, Jul 16, 2013 at 5:46 PM, Dave Reynolds
<dave.e.reynolds@gmail.com> wrote:
> Hi Jose,
>
> Finally found time to look at this, sorry for the delay.
>
> Short answer:
>
> The data is incorrect (or at least incomplete) but in ways that the official
> integrity checks don't spot. The on line tool does use a different
> implementation of IC-12 (to avoid the scaling problem with the sparql query)
> that accidentally picks up the problem.
>
> Details:
>
> You have a number of qb:Observations which say they are in the data set
> dataset:Computation.
>
> However dataset:Computation is not defined anywhere and in particular does
> not have an associated qb:structure.
>
> The rule IC-12 checks for duplicate observations for which the value of
> every declared dimension is the same. In the case where no dimensions are
> apparently declared the official IC-12 SPARQL query will pass by default
> because it never finds any duplicate observation values. The hand coded more
> efficient version in the validator sees multiple observations in the same
> data set which don't differ by any dimensions. Essentially a difference how
> the null case is treated.
>
> This does reveal a limitation of the integrity checks.
>
> The rules check that every qb:Observation has a qb:dataSet value and that
> every declared qb:DataSet has a qb:structure but doesn't catch the case
> where there a qb:DataSet is implicitly used but not declared. That is a
> limitation of the simplified closure algorithm.
>
> The WG will need to consider what to do about that.
>
> Thanks for the report.
>
> Dave
>
>
> On 08/07/13 13:34, Jose Emilio Labra Gayo wrote:
>>
>> I have a running example that I am trying to validate with the RDF
>> Data Cube validator.
>>
>> Using my local copy of the queries it passes all the tests, however,
>> when I try to use the RDF Data Cube Validator, it fails for integrity
>> constraint 12.
>>
>> I attach the file and my local copy of the query, it is the query from
>> the specification plus the prefixes and it returns NO (=success).
>>
>> Is it possible that the validator is using a different version of the
>> query which makes it to fail?
>>
>> --
>> Best regards, Labra
>>
>



-- 
Saludos, Labra
Received on Tuesday, 16 July 2013 21:24:12 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 16 July 2013 21:24:12 UTC