> On April 13, 2017, 12:37 p.m., David Radley wrote: > > My understanding of Rest calls is that POST should be for create and PUT > > for update - to mirror the http spec - I would stick withj this standard. > > I like the separation between create and upate - especially for types - so > > we can explicitly understand the intent of the operation as create or > > update for governance. > > Apoorv Naik wrote: > Yes that's correct. Let me revisit this and see if it makes more sense to > introduce a new endpoint for this functionality. Let's say POST > /types/typedefs/createOrUpdate > > Apoorv Naik wrote: > Hey David, do you any comments on introducing the new endpoint ? > > David Radley wrote: > Looking at https://en.wikipedia.org/wiki/Representational_state_transfer > this seems to be a reasonable description of the REST standard. I notice the > table "Relationship between URL and HTTP methods". so it seems to me : > - post without the new resournce name in the URL is create only. This is > the creation mechanism which is very widespread ; I have only seen this way > of creating. > - put with the new resource name in the URL is update or create. I have > not seen this widely used for create. > > I think if we expose the post and put as above - then the user can choose > between create only or "update or create". > > David Radley wrote: > With the new proposed endpoint - createOrUpdate would look like the name > of a resource. So this would be non-standard Rest; I suggest we keep as > standard as possible on the API to facilitate easy integration.
In this specific case, the API deals with update to bunch of types as whole. Some of the updates could refer to new types, which need to be created in the same transaction that performs type-update. Hence this patch. In many cases, use of createOrUpdate semantics can be useful/efficient - like when the caller doesn't know/care if the object exists or not. - Madhan ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/58333/#review171862 ----------------------------------------------------------- On April 12, 2017, 8:35 p.m., Apoorv Naik wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/58333/ > ----------------------------------------------------------- > > (Updated April 12, 2017, 8:35 p.m.) > > > Review request for atlas, Ashutosh Mestry, Madhan Neethiraj, Nixon Rodrigues, > Sarath Subramanian, and Suma Shivaprasad. > > > Repository: atlas > > > Description > ------- > > TypesREST to support creation in an update call > > > Diffs > ----- > > intg/src/main/java/org/apache/atlas/store/AtlasTypeDefStore.java > 198bd8fe515a96e654b24de3af92b6edfac3a6ae > > repository/src/main/java/org/apache/atlas/repository/store/graph/AtlasTypeDefGraphStore.java > 82465bfcbe7705e5493481a5aa080c0390a1cfe4 > webapp/src/main/java/org/apache/atlas/web/resources/TypesResource.java > 424ca58ce0bdeb71a6ddc829f98f303988216c99 > webapp/src/main/java/org/apache/atlas/web/rest/TypesREST.java > 5d846cf317ae46fa88ef7877baad129638e38e03 > > > Diff: https://reviews.apache.org/r/58333/diff/3/ > > > Testing > ------- > > mvn clean package executes successfully > > > Thanks, > > Apoorv Naik > >
