[OTDev] Algorithm/Model/Task related API documentation
Martin Guetlein martin.guetlein at googlemail.comMon Nov 8 10:40:39 CET 2010
- Previous message: [OTDev] Algorithm/Model/Task related API documentation
- Next message: [OTDev] Algorithm/Model/Task related API documentation
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
On Mon, Nov 8, 2010 at 10:09 AM, Nina Jeliazkova <jeliazkova.nina at gmail.com> wrote: > Hi Martin, All, > > On 8 November 2010 10:55, Martin Guetlein <martin.guetlein at googlemail.com>wrote: > >> Hi All, >> >> I adjusted the Task API. The relevant section can be found at >> http://www.opentox.org/dev/apis/api-1.2/AsyncTask#creating-a-task-post >> and below. >> >> These points could be controversial: >> >> * I defined that a GET task/{id} with result header 'text/uri-list' >> returns the resultURI field of the task (That's the way we implemented >> it at ALU-FR and IST ;-)). >> I am well aware that this is a bit un-restfull, but I think it serves >> our purposes. >> > > Less consistent, but might be reasonable exception for text/uri-list. > > >> * A service may return a Task __or__ the result directly. Even the >> same service may do both, depending on the input. This can be >> distinguished via status code. >> > > Just to clarify, it should be result URI , not the result object > representation. > > >> >> * If a service does not return a task and Accept:application/rdf+xml >> is set, should the complete result object be returned? >> > > No, I strongly vote against returning object content, because it > deteriorates performance and is less consistent. > URI of the result object should be returned in some way, thus my preference > to return Task object always. text/uri-list might be an exception. > The client should go to the result resource and read the resource in most > appropriate way. (e.g. not a large dataset entirely,but paginated, as > decided by the client). On Mon, Nov 8, 2010 at 10:00 AM, Tobias Girschick <tobias.girschick at in.tum.de> wrote: > I would say yes, as this is what the client requests and expects. Are > there cases that could be problematic? I share Nina's opinion that we should not return the whole data due to performance issues. The question is, what to return if RDF/XML is requested by the client? IMHO, we have the following options: * Always return the task object (as Nina proposed). Drawback: inappropriate for services that are very fast, and do not need a task. * Return value for RDF/XML is undefined. Drawback: we should always support RDF/XML as this is the default format for OT web services. * Return the URI, wrapped in RDF/XML I would go for the last option. Best regards, Martin > > > Best regards, > Nina > > >> >> What do you think? >> >> >> Best regards, >> Martin >> >> >> On Tue, Nov 2, 2010 at 10:46 AM, Nina Jeliazkova >> <jeliazkova.nina at gmail.com> wrote: >> > On 2 November 2010 11:40, Martin Guetlein < >> martin.guetlein at googlemail.com>wrote: >> > >> >> On Tue, Nov 2, 2010 at 10:31 AM, Nina Jeliazkova >> >> <jeliazkova.nina at gmail.com> wrote: >> >> > On 2 November 2010 11:21, Martin Guetlein < >> >> martin.guetlein at googlemail.com>wrote: >> >> > >> >> >> On Tue, Nov 2, 2010 at 9:56 AM, Nina Jeliazkova >> >> >> <jeliazkova.nina at gmail.com> wrote: >> >> >> > Martin, >> >> >> > >> >> >> > On 2 November 2010 10:35, Martin Guetlein < >> >> >> martin.guetlein at googlemail.com>wrote: >> >> >> > >> >> >> >> Dear Nina, Christoph, >> >> >> >> >> >> >> >> What I am doing (implementation wise), is using the http code to >> >> >> >> identify whether the result of a POST call is a task (202) or the >> >> >> >> actual object (200). Doing that there is no unneeded additional >> HTTP >> >> >> >> call, and each service can return the task or the object itself, >> >> >> >> according to how long the process takes. >> >> >> >> >> >> >> > >> >> >> > Do you return object content, or object URI? >> >> >> > >> >> >> > Do you handle "Accept" header with different mime types the same >> way? >> >> >> >> >> >> If the result is computed as task, a POST always returns content-type >> >> >> "text/uri-list", content is the TASK uri, http-code is 202 (The task >> >> >> itself is available in different content-types). >> >> >> >> >> >> Returning the task if the POST accept header rdf, is still on todo >> >> >> list (nobody complained so far). >> >> >> >> >> > >> >> > OK. >> >> > >> >> > BTW, the http code could be 201 (as per wiki page) >> >> >> >> I think it was 202. 201 was for linking to another task see >> >> http://opentox.org/dev/apis/api-1.2/AsyncTask >> >> >> >> >> > Exactly, get URI of a task can return 201, if means it "redirects" the >> > processing to another task. It is a valid code to expect and has been >> > changed from 303. >> > >> > >> >> > >> >> > >> >> >> >> >> >> On Sat, Oct 30, 2010 at 9:53 AM, Nina Jeliazkova >> >> >> <jeliazkova.nina at gmail.com> wrote: >> >> >> > However, if the "Accept:text/uri-list" is requested instead of the >> >> >> default >> >> >> > RDF, the URL returned by completion will be the one of the result >> >> itself >> >> >> > (e.g. URL of a dataset with predicted results). >> >> >> >> >> >> Why don't you return the TASK uri in that case? IMHO the accept >> header >> >> >> should only affect the format of the object, not the returned object >> >> >> itself. >> >> >> >> >> > >> >> > I guess, because it is redundant - GET of an URI to return the same >> URI. >> >> > >> >> > Makes sense for consistency only. >> >> > >> >> > >> >> >> >> >> >> However, IMHO it is no problem that the web services return the >> object >> >> >> result or the task alternately, as long as it can be distinguished >> via >> >> >> http-code. >> >> >> >> >> >> >> >> > Fine as far as it is clearly specified. Can we document on the API >> page >> >> > what is the expected behaviour in all cases? >> >> >> >> I think the documentation at the Task API is quite substantial. >> >> What we could additionally point out more clearly is the __or__ case, >> >> that the service could return the object uri directly without task >> >> with http-code 200. >> >> We could add an internal link from the other API pages to this >> >> documentation. I could do that. >> >> >> >> Any objections? >> >> >> > >> > >> > No. In addition, the doc should also explicitly list the behaviour on >> > different Accept headers . >> > >> > Best regards, >> > Nina >> > >> > >> >> >> >> Best regards, >> >> Martin >> >> >> >> >> >> > >> >> > Best regards, >> >> > Nina >> >> > >> >> > >> >> >> Best regards, >> >> >> Martin >> >> >> >> >> >> > >> >> >> > >> >> >> > >> >> >> >> >> >> >> >> So from my point of view the API with the __or__ is fine, maybe we >> >> >> >> should make the implementation a bit more clear. >> >> >> >> >> >> >> >> >> >> >> > >> >> >> > >> >> >> > Thanks, >> >> >> > Nina >> >> >> > >> >> >> > >> >> >> >> Regards, >> >> >> >> Martin >> >> >> >> >> >> >> >> >> >> >> >> On Mon, Nov 1, 2010 at 6:47 PM, Nina Jeliazkova >> >> >> >> <jeliazkova.nina at gmail.com> wrote: >> >> >> >> > Dear Christoph, >> >> >> >> > >> >> >> >> > >> >> >> >> > On 1 November 2010 19:22, Christoph Helma <helma at in-silico.ch> >> >> wrote: >> >> >> >> > >> >> >> >> >> Dear Nina, All, >> >> >> >> >> >> >> >> >> >> Excerpts from Nina Jeliazkova's message of Sat Oct 30 09:53:02 >> >> +0200 >> >> >> >> 2010: >> >> >> >> >> > Hello All, >> >> >> >> >> > >> >> >> >> >> > We have found somewhat inconsistent documentation, regarding >> the >> >> >> usage >> >> >> >> of >> >> >> >> >> > tasks. The API wiki says it is either the URL of the result >> >> >> returned, >> >> >> >> or >> >> >> >> >> > the task URL. >> >> >> >> >> > >> >> >> >> >> > In practice, if the (default) "Accept:application/rdf+xml" >> >> header >> >> >> is >> >> >> >> >> > requested, a task URL is always returned and the task >> contains >> >> >> >> >> information >> >> >> >> >> > if it is completed or not, and if it is completed, then the >> RDF >> >> >> >> >> > representation of the task contains the URI of result. >> >> >> >> >> >> >> >> >> >> I always had the impression that >> >> >> >> >> - POST operations should return the URI of the created object >> >> _or_ a >> >> >> >> task >> >> >> >> >> URI (i.e. Accept:text/uri-list) >> >> >> >> >> - GET operations should return the object representation >> in >> >> >> >> OWL-DL >> >> >> >> >> (by default in RDF/XML format) (i.e. >> Accept/application/rdf+xml) >> >> >> >> >> >> >> >> >> >> I would expect the following workflow for services that require >> >> tasks >> >> >> >> >> (thats how it was implemented in ALU/IST services): >> >> >> >> >> >> >> >> >> >> POST to the service - returns task URI >> >> >> >> >> GET the task URI - returns task rdfxml with status (and URI of >> the >> >> >> >> created >> >> >> >> >> object if status is Completed) >> >> >> >> >> If status is Completed: GET URI of result object - returns >> object >> >> >> >> >> representation (rdfxml by default) >> >> >> >> >> >> >> >> >> >> What would be the advantage of getting RDF/XML for a newly >> created >> >> >> task >> >> >> >> >> (status would be Created by definition)? >> >> >> >> >> >> >> >> >> > >> >> >> >> > Not necessarily, for a quick operation the task might already be >> >> >> >> completed >> >> >> >> > and the result returned. What we could gain is one HTTP call >> >> less. >> >> >> >> > >> >> >> >> > The confusing point is actually "_or_" in the API - "URI of the >> >> >> created >> >> >> >> > object _or_ a task URI (i.e. Accept:text/uri-list) " - it >> implies >> >> one >> >> >> >> could >> >> >> >> > receive URI of the result , not the task. I would suggest to >> >> specify >> >> >> >> that >> >> >> >> > POST to a service always return a task (URI or representation) , >> >> not >> >> >> the >> >> >> >> URI >> >> >> >> > of the object. >> >> >> >> > >> >> >> >> > IDEA implementation is as follows: >> >> >> >> > - POST to the service - return task URI (if text/uri-list) or >> >> RDF/XML >> >> >> of >> >> >> >> a >> >> >> >> > task (if Accept:application/rdf+xml is requested). >> >> >> >> > >> >> >> >> > - GET the task URI - returns task content in the requested >> >> >> representation >> >> >> >> > always (e.t. RDF/XML is requested, text-uri-list if requested). >> >> >> >> > In case of text/uri-list it returns task URI, but if the task is >> >> >> >> completed, >> >> >> >> > returns the object URI (with uri-list there is no really other >> >> options >> >> >> >> how >> >> >> >> > to return the result) >> >> >> >> > >> >> >> >> > - GET task URI never returns object representation, the result >> URI >> >> is >> >> >> >> > available in the task representation. The reason is the object >> >> might >> >> >> be >> >> >> >> a >> >> >> >> > huge, and the client application might not necessarily want to >> >> receive >> >> >> >> > entirely at this moment (e.g. might prefer to show it page by >> page >> >> ). >> >> >> >> > >> >> >> >> > Best regards, >> >> >> >> > Nina >> >> >> >> > >> >> >> >> > >> >> >> >> >> >> >> >> >> >> Best regards, >> >> >> >> >> Christoph >> >> >> >> >> >> >> >> >> >> > >> >> >> >> >> > However, if the "Accept:text/uri-list" is requested instead >> of >> >> the >> >> >> >> >> default >> >> >> >> >> > RDF, the URL returned by completion will be the one of the >> >> result >> >> >> >> itself >> >> >> >> >> > (e.g. URL of a dataset with predicted results). >> >> >> >> >> > >> >> >> >> >> > Should we change anything, or just try to make the >> documentation >> >> >> >> clear? >> >> >> >> >> > Perhaps make explicit what is being returned if different >> >> "Accept:" >> >> >> >> >> headers >> >> >> >> >> > are used? >> >> >> >> >> > Are there implementations that handles these cases >> differently? >> >> >> >> >> > >> >> >> >> >> > 1)Algorithm http://opentox.org/dev/apis/api-1.1/Algorithm >> >> >> >> >> > Result: *model URI ,dataset URI,featureURI* , Redirect to >> task >> >> URI >> >> >> for >> >> >> >> >> time >> >> >> >> >> > consuming computations >> >> >> >> >> > >> >> >> >> >> > Note: Redirect was removed from the API but apparently this >> page >> >> >> was >> >> >> >> not >> >> >> >> >> > updated, there is still reference to 303 status codes. >> >> >> >> >> > >> >> >> >> >> > 2)Model http://opentox.org/dev/apis/api-1.1/Model >> >> >> >> >> > Result: URI of created prediction dataset (predictions are >> >> >> features), >> >> >> >> >> task >> >> >> >> >> > URI for time consuming computations >> >> >> >> >> > >> >> >> >> >> > Best regards, >> >> >> >> >> > Nina >> >> >> >> >> _______________________________________________ >> >> >> >> >> Development mailing list >> >> >> >> >> Development at opentox.org >> >> >> >> >> http://www.opentox.org/mailman/listinfo/development >> >> >> >> >> >> >> >> >> > _______________________________________________ >> >> >> >> > Development mailing list >> >> >> >> > Development at opentox.org >> >> >> >> > http://www.opentox.org/mailman/listinfo/development >> >> >> >> > >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> -- >> >> >> >> Dipl-Inf. Martin Gütlein >> >> >> >> Phone: >> >> >> >> +49 (0)761 203 8442 (office) >> >> >> >> +49 (0)177 623 9499 (mobile) >> >> >> >> Email: >> >> >> >> guetlein at informatik.uni-freiburg.de >> >> >> >> _______________________________________________ >> >> >> >> Development mailing list >> >> >> >> Development at opentox.org >> >> >> >> http://www.opentox.org/mailman/listinfo/development >> >> >> >> >> >> >> > _______________________________________________ >> >> >> > Development mailing list >> >> >> > Development at opentox.org >> >> >> > http://www.opentox.org/mailman/listinfo/development >> >> >> > >> >> >> >> >> >> >> >> >> >> >> >> -- >> >> >> Dipl-Inf. Martin Gütlein >> >> >> Phone: >> >> >> +49 (0)761 203 8442 (office) >> >> >> +49 (0)177 623 9499 (mobile) >> >> >> Email: >> >> >> guetlein at informatik.uni-freiburg.de >> >> >> _______________________________________________ >> >> >> Development mailing list >> >> >> Development at opentox.org >> >> >> http://www.opentox.org/mailman/listinfo/development >> >> >> >> >> > _______________________________________________ >> >> > Development mailing list >> >> > Development at opentox.org >> >> > http://www.opentox.org/mailman/listinfo/development >> >> > >> >> >> >> >> >> >> >> -- >> >> Dipl-Inf. Martin Gütlein >> >> Phone: >> >> +49 (0)761 203 8442 (office) >> >> +49 (0)177 623 9499 (mobile) >> >> Email: >> >> guetlein at informatik.uni-freiburg.de >> >> _______________________________________________ >> >> Development mailing list >> >> Development at opentox.org >> >> http://www.opentox.org/mailman/listinfo/development >> >> >> > _______________________________________________ >> > Development mailing list >> > Development at opentox.org >> > http://www.opentox.org/mailman/listinfo/development >> > >> >> >> >> -- >> Dipl-Inf. Martin Gütlein >> Phone: >> +49 (0)761 203 8442 (office) >> +49 (0)177 623 9499 (mobile) >> Email: >> guetlein at informatik.uni-freiburg.de >> _______________________________________________ >> Development mailing list >> Development at opentox.org >> http://www.opentox.org/mailman/listinfo/development >> > _______________________________________________ > Development mailing list > Development at opentox.org > http://www.opentox.org/mailman/listinfo/development > -- Dipl-Inf. Martin Gütlein Phone: +49 (0)761 203 8442 (office) +49 (0)177 623 9499 (mobile) Email: guetlein at informatik.uni-freiburg.de
- Previous message: [OTDev] Algorithm/Model/Task related API documentation
- Next message: [OTDev] Algorithm/Model/Task related API documentation
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Development mailing list