Re: Bug in DM test D024?

Richard, Eric

TCs doc and repos are updated without the null-values related triples.

Boris



On Jun 7, 2012, at 7:08 PM, Eric Prud'hommeaux wrote:

> * Richard Cyganiak <richard@cyganiak.de> [2012-06-07 17:32+0200]
>> Hi Boris, Eric,
>> 
>> The DM spec says that the row graph contains:
>> 
>> [[
>> a reference triple for each <column name list> in a table's foreign keys where none of the column values is NULL.
>> ]]
> 
> Yeah, that could have been "has at least one non-NULL value", which is the semantics of MATCH TYPE FULL.
> Unique keys permit any unique arrangements of NULLs so it really is up to the foreign key match type semantics we want to support.
> That said, we haven't spec'd how to embed NULL values in IRIs.
> Most importantly, most joins are of the form (referrer.a1=referent.a1 … AND referrer.aⁿ=referent.aⁿ), which corresponds to MATCH TYPE FULL.
> 
> It will be a lot easier for future versions to extend the DM (to suport MATCH TYPE DEFAULT) than to reduce the DM so we have plenty of time to reconsider this later.
> 
> 
>> In create.sql for D024 we have:
>> 
>> [[
>> INSERT INTO "Source" ("ID", "attrA", "attrB")
>>             VALUES (1101, 'K2A22', NULL   );
>> ]]
>> 
>> And then we have a triple in directGraph.ttl:
>> 
>> [[
>> <Source/ID=1101> <Source#ref-attrA;attrB> <Target/PK=1011> .
>> ]]
>> 
>> I believe this triple shouldn't be there, because one of the involved column values is NULL.
>> 
>> Or am I misunderstanding something?
> 
> Nope, I totally agree.
> 
> also, D021:
> [[
> - <Source/ID-1101> <Source#ref-attrA.attrB> <Target/PK-1011> .
> - <Source/ID-1102> <Source#ref-attrA.attrB> <Target/PK-1012> .
> ]]
> 
> 
>> Best,
>> Richard
> 
> -- 
> -ericP
>

Received on Tuesday, 12 June 2012 15:03:28 UTC