SDB load prefixes

[Petr Křemen <pe...@gmail.com>]

Hi to all,

I have a single store and two named models opened on top of it, say
model1 and model2. When reading data using

model1.read(...,BASE1)
model2.read(...,BASE2)

the second command fails with

ERROR: duplicate key value violates unique constraint "prefixes_pk"
INSERT INTO Prefixes
   VALUES (':', NSOFTHESECOND)

I understand, that a single prefix should not be set to different
mapping within a single SDB store.

My question is - is there a possibility to throw away all prefix
information from the RDF/XMLs and to load the documents with no (or
artifically generated) prefixes, such that the above error disappears
?

Thanks,
Petr

SDB load prefixes

[Petr Křemen <pe...@gmail.com>]

Hi to all,

I have a single store and two named models opened on top of it, say
model1 and model2. When reading data using

model1.read(...,BASE1)
model2.read(...,BASE2)

the second command fails with

ERROR: duplicate key value violates unique constraint "prefixes_pk"
INSERT INTO Prefixes
  VALUES (':', NSOFTHESECOND)

I understand, that a single prefix should not be set to different
mapping within a single SDB store.

My question is - is there a possibility to throw away all prefix
information from the RDF/XMLs and to load the documents with no (or
artifically generated) prefixes, such that the above error disappears
?

Thanks,
Petr


-- 
Petr Křemen
Petr.Kremen@gmail.com

Re: SDB load prefixes

[Petr Křemen <pe...@gmail.com>]

Hi Andy,

thanks for your answer.

> Which version of SDB and which database are you using? (It looks like
> posgresql)  What would be helpful is a complete, minimal example, especially
> if it works on one of the embedded DBs like H2 or Derby.
SDB maven artifact ver.1.3.4, PostgreSQL DB.

Regarding per-graph (TDB) and overall (SDB) prefixes storing, it is
indeed a bit strange - although TDB and SDB are "just 2 different
backends", the semantics of model.read are so different in each case.
In my view, per-graph prefix storing for SDB would definitely be less
confusing.

> The prefixes are being set because you are parsing directly into database
> backed models.  If you read into a memory model then add that , the prefixes
> aren't copied over.
>
>   Model m = FileManager.get().read(..)
>   model1.add(m) ;
>   m = FileManager.get().read(..)
>   model2.add(m) ;
>
This works like a charm for me. Thanks for your help

Petr

Re: SDB load prefixes

[Andy Seaborne <an...@apache.org>]

On 11/03/12 09:05, Petr Křemen wrote:
> Hi to all,

Hi Petr,

Which version of SDB and which database are you using? (It looks like 
posgresql)  What would be helpful is a complete, minimal example, 
especially if it works on one of the embedded DBs like H2 or Derby.

Unlike TDB, SDB stores prefixes on a dataset basis, not a per-graph 
basis.  Changing this would need a change to the on-disk formats so 
while it is desirable to do, it's also not insignificant.

> I have a single store and two named models opened on top of it, say
> model1 and model2. When reading data using
>
> model1.read(...,BASE1)
> model2.read(...,BASE2)
>
> the second command fails with
>
> ERROR: duplicate key value violates unique constraint "prefixes_pk"
> INSERT INTO Prefixes
>     VALUES (':', NSOFTHESECOND)
>
> I understand, that a single prefix should not be set to different
> mapping within a single SDB store.
>
> My question is - is there a possibility to throw away all prefix
> information from the RDF/XMLs and to load the documents with no (or
> artifically generated) prefixes, such that the above error disappears

The prefixes are being set because you are parsing directly into 
database backed models.  If you read into a memory model then add that , 
the prefixes aren't copied over.

    Model m = FileManager.get().read(..)
    model1.add(m) ;
    m = FileManager.get().read(..)
    model2.add(m) ;

As a bonus, if there is a problem due to some data error, you only add 
data to the database if it was completely readable.

	Andy

> ?
>
> Thanks,
> Petr