Re: Restricting SPARQL update to a single named graph
That was a naive implementation... This one based on regex is somewhat
more robust:
https://github.com/AtomGraph/LinkedDataHub/blob/develop/src/main/java/com/atomgraph/linkeddatahub/server/model/impl/GraphStoreImpl.java#L322
If anyone has examples of how to do a similar thing without string
manipulation, that would be appreciated.
On Thu, Jun 9, 2022 at 12:21 PM Martynas Jusevičius
wrote:
>
> On Thu, Jun 9, 2022 at 11:03 AM Martynas Jusevičius
> wrote:
> >
> > On Wed, Jun 8, 2022 at 12:22 PM Andy Seaborne wrote:
> > >
> > >
> > >
> > > On 08/06/2022 09:22, Martynas Jusevičius wrote:
> > > > On Tue, Jun 7, 2022 at 9:15 PM Andy Seaborne wrote:
> > > >>
> > > >> On 07/06/2022 10:47, Martynas Jusevičius wrote:
> > > >>> Hi,
> > > >>>
> > > >>> I have implemented PATCH method for the Graph Store Protocol:
> > > >>> https://www.w3.org/TR/sparql11-http-rdf-update/#http-patch
> > > >>>
> > > >>> The PATCH is applied to a named graph. I am missing this bit however:
> > > >>> " If a SPARQL 1.1 Update request is used as the RDF payload for a
> > > >>> PATCH request that makes changes to more than one graph or the graph
> > > >>> it modifies is not the one indicated, it would be prudent for the
> > > >>> server to respond with a 422 Unprocessable Entity status."
> > > >>
> > > >> I read that in the context of GSP resource naming.
> > > >>
> > > >> ?graph=
> > > >>
> > > >> and so the update does not name a graph - it'll look like the default
> > > >> graph in the update.
> > > >>
> > > >> So look for GRAPH in the update.
> > > >
> > > > Thanks, that makes sense. GRAPH is also easy to check.
> > > >
> > > > But then I need to forward the update to a triplestore that does not
> > > > support PATCH.
> > >
> > > It's the GSP naming that matters.
> > >
> > > > Which means I would need to wrap INSERT/DELETE/WHERE templates into
> > > > GRAPH { }.
> > >
> > > WITH DELETE {} INSERT {} WHERE {}
> > >
> > > Also: USING. And protocol.
> >
> > Thanks, forgot about those. I could definitely use WITH.
> >
> > And re. SPARQL protocol, would ?using-named-graph-uri=uri have the
> > same effect as WITH in this case?
>
> I realized ?using-named-graph-uri=uri would restrict the dataset but
> still require GRAPH in the update string.
>
> WITH on the other hand does what I need. I add it by string
> manipulation -- the current code is ugly but seems to do the job:
>
> String updateString = updateRequest.toString();
> // append WITH before DELETE or INSERT
> if (updateString.toUpperCase().contains("DELETE"))
>updateString = updateString.replaceAll("(?i)" +
> Pattern.quote("DELETE"), "WITH <" + graphUri + ">\nDELETE");
> else
> {
> if (updateString.toUpperCase().contains("INSERT"))
> updateString = updateString.replaceAll("(?i)" +
> Pattern.quote("INSERT"), "WITH <" + graphUri + ">\nINSERT");
> else throw new BadRequestException("SPARQL update contains
> no DELETE or INSERT?"); // cannot happen
> }
> updateRequest = UpdateFactory.create(updateString);
>
> >
> > >
> > > > Is there some builder code that can help with that?
> > >
> > > Have you looked at UpdateBuilder?
> > >
> >
> > I looked at the Javadoc, it seems quite complicated. I'll see if I can
> > avoid modifying the update string.
> >
> >
> > > >
> > > >>
> > > >>> What would be the way to make sure that an update only affects a
> > > >>> single specific graph?
> > > >>
> > > >> A dataset of one graph and no others. c.f. DatasetGraphOne but for a
> > > >> single named graph and read-only dft graph.
> > > >>
> > > >> Or a dataset which yields read-only graphs except for the target graph.
> > > >>
> > > >> Or analyse the update - no GRAPH in templates if the target comes from
> > > >> the URL.
> > > >>
> > > >>>
> > > >>>
> > > >>> Martynas
> > > >>> atomgraph.com
Re: Restricting SPARQL update to a single named graph
On Thu, Jun 9, 2022 at 11:03 AM Martynas Jusevičius
wrote:
>
> On Wed, Jun 8, 2022 at 12:22 PM Andy Seaborne wrote:
> >
> >
> >
> > On 08/06/2022 09:22, Martynas Jusevičius wrote:
> > > On Tue, Jun 7, 2022 at 9:15 PM Andy Seaborne wrote:
> > >>
> > >> On 07/06/2022 10:47, Martynas Jusevičius wrote:
> > >>> Hi,
> > >>>
> > >>> I have implemented PATCH method for the Graph Store Protocol:
> > >>> https://www.w3.org/TR/sparql11-http-rdf-update/#http-patch
> > >>>
> > >>> The PATCH is applied to a named graph. I am missing this bit however:
> > >>> " If a SPARQL 1.1 Update request is used as the RDF payload for a
> > >>> PATCH request that makes changes to more than one graph or the graph
> > >>> it modifies is not the one indicated, it would be prudent for the
> > >>> server to respond with a 422 Unprocessable Entity status."
> > >>
> > >> I read that in the context of GSP resource naming.
> > >>
> > >> ?graph=
> > >>
> > >> and so the update does not name a graph - it'll look like the default
> > >> graph in the update.
> > >>
> > >> So look for GRAPH in the update.
> > >
> > > Thanks, that makes sense. GRAPH is also easy to check.
> > >
> > > But then I need to forward the update to a triplestore that does not
> > > support PATCH.
> >
> > It's the GSP naming that matters.
> >
> > > Which means I would need to wrap INSERT/DELETE/WHERE templates into
> > > GRAPH { }.
> >
> > WITH DELETE {} INSERT {} WHERE {}
> >
> > Also: USING. And protocol.
>
> Thanks, forgot about those. I could definitely use WITH.
>
> And re. SPARQL protocol, would ?using-named-graph-uri=uri have the
> same effect as WITH in this case?
I realized ?using-named-graph-uri=uri would restrict the dataset but
still require GRAPH in the update string.
WITH on the other hand does what I need. I add it by string
manipulation -- the current code is ugly but seems to do the job:
String updateString = updateRequest.toString();
// append WITH before DELETE or INSERT
if (updateString.toUpperCase().contains("DELETE"))
updateString = updateString.replaceAll("(?i)" +
Pattern.quote("DELETE"), "WITH <" + graphUri + ">\nDELETE");
else
{
if (updateString.toUpperCase().contains("INSERT"))
updateString = updateString.replaceAll("(?i)" +
Pattern.quote("INSERT"), "WITH <" + graphUri + ">\nINSERT");
else throw new BadRequestException("SPARQL update contains
no DELETE or INSERT?"); // cannot happen
}
updateRequest = UpdateFactory.create(updateString);
>
> >
> > > Is there some builder code that can help with that?
> >
> > Have you looked at UpdateBuilder?
> >
>
> I looked at the Javadoc, it seems quite complicated. I'll see if I can
> avoid modifying the update string.
>
>
> > >
> > >>
> > >>> What would be the way to make sure that an update only affects a
> > >>> single specific graph?
> > >>
> > >> A dataset of one graph and no others. c.f. DatasetGraphOne but for a
> > >> single named graph and read-only dft graph.
> > >>
> > >> Or a dataset which yields read-only graphs except for the target graph.
> > >>
> > >> Or analyse the update - no GRAPH in templates if the target comes from
> > >> the URL.
> > >>
> > >>>
> > >>>
> > >>> Martynas
> > >>> atomgraph.com
Re: Restricting SPARQL update to a single named graph
On Wed, Jun 8, 2022 at 12:22 PM Andy Seaborne wrote:
>
>
>
> On 08/06/2022 09:22, Martynas Jusevičius wrote:
> > On Tue, Jun 7, 2022 at 9:15 PM Andy Seaborne wrote:
> >>
> >> On 07/06/2022 10:47, Martynas Jusevičius wrote:
> >>> Hi,
> >>>
> >>> I have implemented PATCH method for the Graph Store Protocol:
> >>> https://www.w3.org/TR/sparql11-http-rdf-update/#http-patch
> >>>
> >>> The PATCH is applied to a named graph. I am missing this bit however:
> >>> " If a SPARQL 1.1 Update request is used as the RDF payload for a
> >>> PATCH request that makes changes to more than one graph or the graph
> >>> it modifies is not the one indicated, it would be prudent for the
> >>> server to respond with a 422 Unprocessable Entity status."
> >>
> >> I read that in the context of GSP resource naming.
> >>
> >> ?graph=
> >>
> >> and so the update does not name a graph - it'll look like the default
> >> graph in the update.
> >>
> >> So look for GRAPH in the update.
> >
> > Thanks, that makes sense. GRAPH is also easy to check.
> >
> > But then I need to forward the update to a triplestore that does not
> > support PATCH.
>
> It's the GSP naming that matters.
>
> > Which means I would need to wrap INSERT/DELETE/WHERE templates into
> > GRAPH { }.
>
> WITH DELETE {} INSERT {} WHERE {}
>
> Also: USING. And protocol.
Thanks, forgot about those. I could definitely use WITH.
And re. SPARQL protocol, would ?using-named-graph-uri=uri have the
same effect as WITH in this case?
>
> > Is there some builder code that can help with that?
>
> Have you looked at UpdateBuilder?
>
I looked at the Javadoc, it seems quite complicated. I'll see if I can
avoid modifying the update string.
> >
> >>
> >>> What would be the way to make sure that an update only affects a
> >>> single specific graph?
> >>
> >> A dataset of one graph and no others. c.f. DatasetGraphOne but for a
> >> single named graph and read-only dft graph.
> >>
> >> Or a dataset which yields read-only graphs except for the target graph.
> >>
> >> Or analyse the update - no GRAPH in templates if the target comes from
> >> the URL.
> >>
> >>>
> >>>
> >>> Martynas
> >>> atomgraph.com
