If you use an API that caches the HTTP connection, HTTP is quite cheap.
HTTP overhead is primarily connection setup/teardown.
After that binary would help.
Content type "application/sparql-results+thrift" for Fuseki.
QueryExecutionFactory.createServiceRequest
QueryEngineHTTP.setSelectContentType
Andy
On 29/04/17 09:38, Laura Morales wrote:
I've got code which allows binary connection via Thrift to a DatasetGraph but
it requires the server side to be bootstrapped. It's something under
consideration by Jean I believe. We use it to connect via Python.
Got any link?
What http overhead are you concerned with?
I'm just trying to make a program that uses Jena/Fuseki as a backend database
server, and the way I access the database right now is via the Fuseki HTTP
endpoint. So I was wondering, if instead of sending HTTP requests for each
operation I could use some library or binding (which should be more efficient,
since there isn't the overhead that comes with HTTP). What I'm talking to would
be the equivalent of using PostgreSQL through PHP PDO.
Hope it makes sense.
