fusepoolp3 / p3-proxy Goto Github PK

If I post to the transforming proxy:

curl -i -H "Content-Type: text/csv" -H "Accept: application/rdf+xml" -H "Slug: acco_150120_6.csv" -X POST --data-binary @accommodations_short_150120_1.csv -X POST http://sandbox.fusepool.info:8181/ldp/virtuoso-csv-transformer

the trace from the fusepool.openlinksw.com console shows
DB.DBA.ldgeneration: received format param: text/turtle
i.e. the originally supplied Accept header isn't received.

Test should be added that test the basic proxing functionality and allow to add tests for additional features.

The charset param confuses the proxy so that no triples are added to the RDF as its not parsed by the proxy.

I created a new T-LDPC using the dashboard on playground.fusepool.info. I still have an issue with the URI created from the pipeline UI (fusepoolP3/p3-pipeline-gui#8) so I add the correct URI manually which leads to this TLDP-C:

http://playground.fusepool.info:8181/ldp/dcr/default-configuration/ALOD-BAR-Neu

If I now post a file there via Dashboard UI or via curl I see the CSV-File ending up there. But it never seems to be transformed.

Note that I am using the sandbox UI to make sure I'm on the latest UI code (http://sandbox.fusepool.info:8200/index.html?platformURI=http://playground.fusepool.info/)

Literals with language are serialized wrongly when sent to the backend.

which is achieved by using the transfomer library

curl -i -H "Content-Type: text/csv" -H 'Slug:pubs141222_no_trailing_sep' -X POST --data-binary @pubs.csv http://sandbox.fusepool.info:8181/ldp/crawled-1
results in the supplied content being transformed.

curl -i -H "Content-Type: text/csv" -H 'Slug:pubs141222_trailing_sep' -X POST --data-binary @pubs.csv http://sandbox.fusepool.info:8181/ldp/crawled-1/
creates a resource in the LDPC, but the content isn't transformed.

Both these variants return the same description:
curl http://sandbox.fusepool.info:8181/ldp/crawled-1/
curl http://sandbox.fusepool.info:8181/ldp/crawled-1

Could the proxy be made more resilient so that it accepts POSTs to an LPDC with or without the trailing '/' ?

For transforming containers the transformer should be invoked (if the media-type matches) and the results posted to the same container.

While debugging fusepoolP3/p3-dashboard#10 we noticed that the request to BatchRefine seems to be wrong. A correct request should look similar to this:

GET /?refinejson=http%3A%2F%2Fsandbox.fusepool.info%3A8181%2Fldp%2Fbla-json.json HTTP/1.1
Accept: text/turtle
Host: sandbox.fusepool.info:8310

while we see it executing:

GET ?refinejson=http%3A%2F%2Fsandbox.fusepool.info%3A8181%2Fldp%2Fbla-json.json HTTP/1.1
Accept: text/turtle
Host: sandbox.fusepool.info:8310

So there is a /missing before the query parameters.

To support 201 responses

As an optional behavior the proxy should put a file structure to the backend, replacing a placeholder with the name of the host used to access the proxy on the first request.

Rationale: this allows to do an initial configuration when multiple transformer factories are shipped as vm or docker alongside the proxy.

Some problems have been reborted on different beaviour using batchrefine as trasformer.