[ome-devel] OME-SRV-TESTS and consistency questions

[Harry Hochheiser]

Some recent examination of small ome-java/back-end bugs (as identified 
by Curtis) led me to a larger concern that merits some wider 
discussion.

Basically, there is an inconsistency between the OMe-JAVA tests in 
OME-SRV-TESTS and the data model that renders those tests useless for 
the newest data model (2.15). The problem lies in the requirement that 
MEXs not be null for STs.  As written, the OME-SRV-TESTs that create 
category groups and other objects do not create MEXs. Thus, these tests 
fail to meet the integrity constraints, and therefore fail outright.

I am aware that Ilya ran these tests last month, and they seemed to 
work. i cannot explain this result.

This is not a problem in Shoola, because the annotation manager is 
called to create annotations for objects before they are complete. This 
provides the necessary MEXs.

This inconsistency raises a couple of questions. On the one hand, the 
OME-SRV-TESTS should probably be as close to the Shoola code as 
possible - if Shoola's OMEDS layer is using the annotation manager to 
create annotations on newly created objects, the OME-SRV-TESTS should 
probably be doing the same thing. Divergence between the tests and the 
client presents an opportunity for further inconsistency and confusion.

The other, potentially deeper question involves clarity of the API. If 
the OME-JAVA API contains calls for creating objects that are not in 
and of themselves sufficient for complete creation of legal objects,  
we should consider extending the API to make these functions more clear 
and modular. As it stands now, the shoola data server (via the class 
OMEDSGateway) is tightly coupled to the OMe-JAVA layer in many ways. 
This makes the construction of clear code somewhat tricky.


-harry