Addressing the no-PK table issue (was: Re: Please review R2RML changes - diff available)

Ivan,

On 2 May 2012, at 08:53, Ivan Herman wrote:
> The cardinality issue is on the borderline, all the other issues are not substantive (although lead to a new LC because implementations may have to change, eg, for the URI). Even the cardinality issue seems to address a corner case if my understanding is correct.

Well, it depends on how we resolve it.

In discussion after the call yesterday, Eric proposed to add something like a “cardinality-preserving blank node constructor function” to R2RML. This would essentially be a term map that produces one blank node per row of the underlying logical table, even if the table contains duplicates. This would actually be a nice addition to the language, and would be handy in situations other than just the non-PK-table situation (e.g., when mapping some kinds of denormalized tables).

So, this addition would solve the DM/R2RML incompatibility problem, *if* we can figure out a way how it can actually be implemented without materialization and with an amount of additional effort that is justified for such a corner case. Eric seems confident that it can be done; Juan and I don't see yet how.

I'm not averse to exploring this, as it would make R2RML better, but it would require additional research and experimentation, so I don't see it happening in just a few weeks. And it may or may not yield a workable solution in the end, AFAICT.

As I see it, our options are:

1) Ship a DM that cannot be implemented in R2RML (Richard -1)
2) Choose a compromise like the one proposed by Juan (Eric -1)
3) Explore the option above (delay, and may not work)

:-/

>> Please send me a pointer to the appropriate SotD boilerplate whenever you're ready.
> 
> I guess you are talking about one subsection of http://www.w3.org/TR/r2rml/ only.
> 
> How does that sound? 
> 
> [[[
> This document is a 2nd Last Call Working Draft of the "R2RML: RDB to RDF Mapping Language". This version reflects changes made as a result of comments received during a Call for Implementations (Candidate Recommendation) period. The Working Group believes that all comments received have been addressed.
> 
> The Working Group welcomes reports of implementations, sent to the comments address. If the Group gathers sufficient evidence of interoperable implementations, based on the results of the Candidate Recommendation phase but also reflecting the changes on the document as an answer to the comments, the group may request to skip 2nd Candidate Recommendation drafts and have the next round of publications be Proposed Recommendations.
> ]]]

Assuming this is the way we go, this sounds good to me.

Just to be sure: Formally, this document would be a Working Draft?

Should this contain a date somewhere, to indicate the end of the 2nd LC phase (with a placeholder date for now)?

Best,
Richard

Received on Wednesday, 2 May 2012 10:05:26 UTC