Alex,
> Will this method work with relational data that I have setup an RDF view
for?
The idea will work, thw procedure will not. It access directly to RDF_QUAD
ignoring any quad maps ('RDF views').
> When I invoke this method with my graph, I get an empty file, which I
assume is because I have not figured out how to speicfy my quad storage...?
Yes,
To get data from your storage, the loop
for (select S as subj, P as pred, O as obj from RDF_QUAD where G = iri_to_id
(graph_iri)) do
should be replaced with something like
for (select subq."s" as subj, subq."p" as pred, subq."o" as obj from (
sparql
define output:valmode "LONG"
define input:storage <http://terapath.net/adb/quad_storage/default>
select ?s ?p ?o from <graph-in-question>
) subq) do
Best Regards,
IvAn.
- Alex
> -----Original Message-----
> From: imikhailov [mailto:imikhai...@openlinksw.com]
> Sent: Monday, April 30, 2007 11:36 AM
> To: Alex Black; virtuoso-users@lists.sourceforge.net
> Subject: RE: [Virtuoso-users] Export RDF to a file
>
> Alex,
>
> CONSTRUCT is definitely not the best way, because it creates in-memory
> dictionary of created triples, 1.8.Gb dictionary of this sort may
> abort client connection by 'memory low' alert.
>
> Latest commercial Virtuoso build contains function
>
> DB.DBA.RDF_GRAPH_TO_TTL (in graph_iri varchar, inout ses any)
>
> that gets a graph IRI and writes the content of the graph into string
> output object that is passes as second argument (if second argument is
> NULL then it will output to the current HTTP client connection, if the
> procedure is called not from an HTTP request then NULL argument will
> result in error).
>
> While the patch that contains this function is on its way to public
> Virtuoso Open Source repository, you may temporarily clone it.
>
> create procedure DB.DBA.RDF_GRAPH_TO_TTL (in graph_iri varchar, inout
> ses
> any)
> {
> declare tcount integer;
> declare res varchar;
> declare prev_s, prev_p IRI_ID;
> tcount := 0;
> prev_s := null;
> prev_p := null;
> for (select S as subj, P as pred, O as obj from RDF_QUAD where G =
> iri_to_id (graph_iri)) do
> {
> if (prev_s = subj)
> {
> if (prev_p = pred)
> {
> http (',\n\t\t', ses);
> goto print_o;
> }
> http (';\n\t', ses);
> goto print_p;
> }
> if (prev_s is not null)
> http ('.\n', ses);
> if (subj >= #i1000000000)
> http (sprintf ('_:b%d ', iri_id_num (subj)), ses);
> else
> {
> res := id_to_iri (subj);
> http ('<', ses);
> http_escape (res, 12, ses, 1, 1);
> http ('> ', ses);
> }
> prev_s := subj;
> prev_p := null;
> print_p:
> if (pred >= #i1000000000)
> signal ('RDFXX', 'DB.DBA.RDF_GRAPH_TO_TTL(): blank node as
> predicate');
> else
> {
> res := id_to_iri (pred);
> http ('<', ses);
> http_escape (res, 12, ses, 1, 1);
> http ('> ', ses);
> }
> prev_p := pred;
> print_o:
> DB.DBA.RDF_LONG_TO_TTL (obj, ses);
> tcount := tcount + 1;
> }
> if (0 = tcount)
> http ('# Empty TURTLE', ses);
> else
> http ('.\n', ses);
> }
> ;
>
> Please rename it to something else to avoid possible conflict when the
> function with this name will appear in the VOS release; you may also
> create it without renaming but drop it before the upgrade.
> You may also replace the condition 'where G = iri_to_id (graph_iri)'
> with something else. but in that case you may wish to add 'order by
> 1,2,3' or the result may become not so pretty-printed.
>
> Before next major release, Virtuoso will optimize some CONSTRUCT
> queries so they will avoid in-memory dictionary in some cases,
> including 'dump graph'
> case; this will eliminate the complication.
>
> Best Regards,
> IvAn Mikhailov.
>
>
> -----Original Message-----
> From: Alex Black [mailto:alex.bl...@terapath.net]
> Sent: Monday, April 30, 2007 10:11 PM
> To: imikhailov; virtuoso-users@lists.sourceforge.net
> Subject: RE: [Virtuoso-users] Export RDF to a file
>
> Ok, thanks Ivan.
>
> Related to this, are there any other options for exporting triples to
> a file?
>
> If I want to export my 1.8gb of data to RDF TTL, is this the most
> performant method in Virtuoso?
>
> - Alex
>
>
