OK: isSubtypeOf is an action on a type, not on instances of a type. I should have thought twice. I like this solution a lot. Thank you!
Johan. On Mon, Nov 7, 2011 at 9:44 AM, Dan Haywood <[email protected]> wrote: > > > On 04/11/2011 08:55, Johan Andries wrote: >> >> It is my goal to put a first version (just a zip file) of the >> javascript application online by the end of next week. After that I'll >> create a Google code project to host it more officially. > > Good stuff. > > >> >> Two remarks about the polymorphism question: >> >> 1. >> http://~/domainTypes/org.apache.isis.example.claims.dom.employee.Employee >> would have to list all super classes (and interfaces it implements), >> not all sub classes as you mention. In this example that would be the >> two interfaces Claimant and Approver. I am not sure that would not >> scale. > > You are right that this is a bounded set and so ought to scale reasonably > well. But some of the lists could be quite large (especially if we included > transitive dependencies so that the client didn't have to walk all the way > up the hierarchy). So this implementation is going to be ugly. > > I'm not going to rule out this approach for good, but I think the need can > be addressed by (2) below - let me expand on it there. > > >> >> 2. isAssignableFrom is a technical action without real business >> meaning, so it must not be offered to end-users in the UI. Isn't that >> a bit inconsistent? > > I guess I didn't explain this properly. I wasn't suggesting that this > action would be available on the DomainObject resource; rather it is > available as a brand-new resource linked from the DomainType resource. > > I've just gone ahead and implemented (and committed) this, though I haven't > yet updated the docs. Still, if you do a GET on the DomainType, you'll see > a new "typeActions" list, which lists the different type actions you can > invoke; eg: > > http://localhost:8080/domainTypes/org.apache.isis.tck.objstore.dflt.scalars.ApplibValuedEntityRepositoryDefault > > will return: > > "typeActions" : [ { > "id" : "isSubtypeOf", > "rel" : "typeaction", > "href" : > "http://localhost:8080/domainTypes/org.apache.isis.tck.objstore.dflt.scalars.ApplibValuedEntityRepositoryDefault/typeactions/isSubtypeOf/invoke";, > "method" : "GET", > "type" : > "application/json;profile=\"urn:org.restfulobjects/domaintypeissubtypeof\"", > "arguments" : { > "supertype" : { > "href" : null > } > } > } ], > > This can then be invoked in the usual way, ie passing the arguments map as a > URL-encoded map for a query string called args. > > Eg to check that the above type IS a subtype of itself, you can use: > > http://localhost:8080/domainTypes/org.apache.isis.tck.objstore.dflt.scalars.ApplibValuedEntityRepositoryDefault/typeactions/isSubtypeOf/invoke?args=%7B%20%22supertype%22%20%3A%20%7B%20%20%22href%22%20%3A%20%22http%3A%2F%2Flocalhost%3A8080%2FdomainTypes%2Forg.apache.isis.tck.objstore.dflt.scalars.ApplibValuedEntityRepositoryDefault%22%20%20%7D%20%7D > > ie, args is the URL encoded version of > > { "supertype" : { "href" : > "http://localhost:8080/domainTypes/org.apache.isis.tck.objstore.dflt.scalars.ApplibValuedEntityRepositoryDefault"; > } } > > (I'm using a little online utility > http://www.albionresearch.com/misc/urlencode.php) to encode/decode maps. > > > >> >> 3. The js application would have to call this isAssignableFrom action >> right at the moment that the user tries to drop an Employee object on >> a Claiment or Approver form field. > > That's true, it would. I don't see this as an issue. > > Of course I could cache previous >> >> calls to isAssignableFrom. > > Yes, though the representation also has a cache-control of 1 day set, so the > web browsers caching (or cache proxies) should take care of this rather than > being write something yourself. > > > >> >> Conclusion: I would like to ask you to reconsider my original >> suggestion (item1 in the list above)... > > Can you see how you get on with (2); if it's unworkable, then we can revisit > (1). > > Thx > Dan > > > >> >> Johan. >> >> On Fri, Nov 4, 2011 at 9:24 AM, Dan Haywood >> <[email protected]> wrote: >>> >>> Interesting question. >>> >>> Rather than listing all subtypes - which I think would not scale >>> particularly well - I think it would be better to have an action, >>> equivalent to java.lang.Class#isAssignableFrom(java.lang.Class). This >>> would allow a client to ask whether the Employee is compatible with the >>> Claimant interface. >>> >>> In other words: >>> >>> { >>> "href": "http:// >>> >>> ~/types/com.mycompany.myapp.claims.Claimant/actions/isAssignableFrom/invoke", >>> "arguments": { >>> "href": "http://~/types/com.mycompany.myapp.employee.Employee"; >>> } >>> } >>> >>> Thoughts? >>> >>> ~~~ >>> Question back to you: do you intend to open source your Javascript app? >>> I'd be very interested in seeing it, at any rate. >>> >>> Cheers >>> Dan >>> >>> >>> >>> >>> On 4 November 2011 07:13, Johan Andries<[email protected]> wrote: >>> >>>> Hi, >>>> >>>> in the claims example application the class >>>> >>>> org.apache.isis.example.claims.dom.claim.Claim >>>> >>>> is using the interfaces >>>> >>>> org.apache.isis.example.claims.dom.claim.Claimant >>>> org.apache.isis.example.claims.dom.claim.Approver >>>> >>>> both implemented by >>>> >>>> org.apache.isis.example.claims.dom.employee.Employee >>>> >>>> The problem is now that my javascript application does not know about >>>> the polymorphic relationship between the Employee class and the two >>>> interfaces, so in this case I have no way of enforcing type safety in >>>> the UI when invoking actions or updating properties. >>>> >>>> One way of solving this would be to specify all implemented interfaces >>>> and superclasses for a given domain type in chapter 20 (Domain Type >>>> Resource) of the restful objects spec (v0.50). For example, >>>> >>>> >>>> http://~/domainTypes/org.apache.isis.example.claims.dom.employee.Employee >>>> >>>> should list all other domain types it can act as. >>>> >>>> Or is there an other solution I am not aware of? >>>> >>>> Johan. >>>> >>> >> > >
