On 05/11/14 09:39, Rob Vesse wrote:
I think this is expected and by-design behaviour
A Dataset does not carry any prefixes directly rather prefixes are only
carried on the individual Models that may be placed inside the Dataset.
AFAICT when writing datasets out ARQ only considers prefixes from the
default graph by default so if you don't have any data in your default
graph you would likely have no prefixes.
Part of the problem is that ARQ doesn't know what the underlying storage
of a Dataset instance is and so gathers the prefixes prior to carrying out
the actual writing. The actual writing is then typically done purely in
terms of quads rather than graphs/models so the writers doesn't have
access to the prefixes of specific models.
Changing the existing behaviour to try and gather the prefixes from all
graphs could be done but may produce a significant performance hit in some
cases.
Rob
Yes - in the case of the general "you can put any model in it", the
prefixes exist with the model.
If you parse into the dataset, the prefixes do end up in the dataset's
choice (default graph).
And of course there is the situation where one prefix is used twice for
different things. It's not the external vocabularies so much as the
local data with ":".
In all syntaxes that have prefixes, you can change prefix declaration
mid way through but that in itself can be unexpected.
| expectations | > | design choices |
Andy
On 04/11/2014 10:32, "Erich Bremer" <[email protected]> wrote:
Hi,
I am writing RDF datasets out from Jena to disk and I noticed that the
prefix definitions are missing, meaning, all predicates are written out
in
the long URI form rather than something compact like "foaf:knows" with
the
foaf prefix defined up at the top. The name spaces were defined the a
Jena
Model and then that model was added as a named graph to a empty dataset
then
written out to disk. Is this expected behavior from Jena or a bug or am
I
missing something here? - Erich
