Even though the JCas classes can be generated from the XML file, you are allowed
to add additional things to those source files, including

   - additional fields

   - additional methods

See
http://uima.apache.org/d/uimaj-2.10.4/references.html#ugr.ref.jcas.augmenting_generated_code

-Marshall

On 11/15/2019 9:50 AM, Alain Désilets wrote:
> On Thu, Nov 14, 2019 at 4:51 PM Richard Eckart de Castilho <[email protected]>
> wrote:
>
>> Sure. You generate the JCas classes once and then you add the methods you
>> want
>> to them. Cf. e.g.
>>
>>
>> https://github.com/dkpro/dkpro-core/blob/8043e10bf10a61fe47e21946ea609bda9f2278a0/dkpro-core-api-metadata-asl/src/main/java/de/tudarmstadt/ukp/dkpro/core/api/metadata/type/DocumentMetaData.java#L290-L447
>
> I know how to create a subclass of Annotation, RelationAnnotation in my
> case. The problem is that if I try to use this subclass in an Annotator,
> UIMA complains that RelationAnnotation is not in the UIMA type system, and
> it lists the available types. This list is essentially the list of types
> defined in some UIMA xml file. This tells me that only those annotation
> classes defined in the xml file can be used in an Annotator. Or at least,
> that I am missing a step for registering my RelationAnnotation class with
> the UIMA type system.
>
> On the other hand, if I define the RelationAnnotation in the xml file, I
> can use it in an Annotator but then I can't figure out how to add methods
> to it, since the Java source for that class is generated automatically (by
> some UIMA maven plugin I presume).
>
> But the question is: why do you want to add new methods? (and is it really
>> a good idea?)
>>
> Essentially, I want to add methods for "derived attributes", i.e.
> attributes whose values are computed from primitive attributes defined in
> the xml file.
>
> I guess I could make those attributes be primitve (i.e. defined in the xml
> file), but then, any annotator that creates a RelationAnnotation would have
> to make sure to set those other attributes correctly. I would much rather
> have the RelationAnnotation class compute those derived attributes itself,
> as it garantees that they will always be computed the same way.
>
> Alain
>

Reply via email to