Although your point has a certain logical consistency to it, here's what I was
thinking at the time:

In UIMA v2, JCas classes were optional, kind of an add-on.

In UIMA v3, instances of JCas classes are *the* representation of a feature
structure.  While it is true that users are not required to define any JCas
classes, or use the JCas interface, nevertheless, the built-in ones are there
and are used to represent instances of the FeatureStructures, even in the case
if a user never defines any JCas classes. 

So, I thought it would be better to have (for the new APIs) a signature that
indicated this.

-Marshall

On 4/16/2019 9:58 AM, Richard Eckart de Castilho wrote:
> Hi there,
>
> I wonder: should the CAS.select() signatures be changed to return <T extends 
> FeatureStructure> instead of <T extends TOP>?
>
> After all, TOP is a JCas class - seems a bit odd to return that in the CAS 
> API?
>
> Cheers,
>
> -- Richard

Reply via email to