CDE/JCasGen does not check for multiple definitions of same feature on type

[Thilo Goetz <tw...@gmx.de>]

Subject line says it all.  I manually created a new
AE descriptor, using copy/paste to create lots of
feature definitions.  As a result, I had the same
feature more than once.  The CDE edits such a file
without complaints.  JCasGen also generates Java files
without complaining, resulting in invalid Java code.

Are these bugs or missing features, or is CDE/JCasGen
simply not meant to be used this way?

(Note that I can't add a duplicate feature in CDE,
there checking does happen.)

--Thilo

Re: CDE/JCasGen does not check for multiple definitions of same feature on type

[Marshall Schor <ms...@schor.com>]

I will check when I get a chance, but my memory is that JCasGen does
work off of the internal representation - it uses the framework code to
read the type system, do all the imports and mergings that might be
needed, and then works off of that result.

So something else might be happening here...

-Marshall

Thilo Goetz wrote:
> Marshall Schor wrote:
>   
>> The CDE design was to check for errors when editing using the forms, but
>> to let the user do anything they want when editing the source XML directly.
>>
>> The checks done against the edited source are done, not by the CDE, but
>> by the base UIMA framework code.  My guess is that that code might be
>> missing checks for duplicates.
>>     
>
> Right, my bad.  Duplicate features are ok from a UIMA perspective.
>
>   
>> The JCasGen assumes the checking was done by the base UIMA framework code.
>>     
>
> Ok, but then JCasGen should work off of the internal representation.
> Seems to me JCasGen uses the XML as input, and that can produce
> invalid Java.
>
>   
>> -Marshall
>>
>> Thilo Goetz wrote:
>>     
>>> Subject line says it all.  I manually created a new
>>> AE descriptor, using copy/paste to create lots of
>>> feature definitions.  As a result, I had the same
>>> feature more than once.  The CDE edits such a file
>>> without complaints.  JCasGen also generates Java files
>>> without complaining, resulting in invalid Java code.
>>>
>>> Are these bugs or missing features, or is CDE/JCasGen
>>> simply not meant to be used this way?
>>>
>>> (Note that I can't add a duplicate feature in CDE,
>>> there checking does happen.)
>>>
>>> --Thilo
>>>
>>>
>>>   
>>>       
>
>
>   

Re: CDE/JCasGen does not check for multiple definitions of same feature on type

[Thilo Goetz <tw...@gmx.de>]

Marshall Schor wrote:
> The CDE design was to check for errors when editing using the forms, but
> to let the user do anything they want when editing the source XML directly.
> 
> The checks done against the edited source are done, not by the CDE, but
> by the base UIMA framework code.  My guess is that that code might be
> missing checks for duplicates.

Right, my bad.  Duplicate features are ok from a UIMA perspective.

> 
> The JCasGen assumes the checking was done by the base UIMA framework code.

Ok, but then JCasGen should work off of the internal representation.
Seems to me JCasGen uses the XML as input, and that can produce
invalid Java.

> 
> -Marshall
> 
> Thilo Goetz wrote:
>> Subject line says it all.  I manually created a new
>> AE descriptor, using copy/paste to create lots of
>> feature definitions.  As a result, I had the same
>> feature more than once.  The CDE edits such a file
>> without complaints.  JCasGen also generates Java files
>> without complaining, resulting in invalid Java code.
>>
>> Are these bugs or missing features, or is CDE/JCasGen
>> simply not meant to be used this way?
>>
>> (Note that I can't add a duplicate feature in CDE,
>> there checking does happen.)
>>
>> --Thilo
>>
>>
>>   

Re: CDE/JCasGen does not check for multiple definitions of same feature on type

[Marshall Schor <ms...@schor.com>]

The CDE design was to check for errors when editing using the forms, but
to let the user do anything they want when editing the source XML directly.

The checks done against the edited source are done, not by the CDE, but
by the base UIMA framework code.  My guess is that that code might be
missing checks for duplicates.

The JCasGen assumes the checking was done by the base UIMA framework code.

-Marshall

Thilo Goetz wrote:
> Subject line says it all.  I manually created a new
> AE descriptor, using copy/paste to create lots of
> feature definitions.  As a result, I had the same
> feature more than once.  The CDE edits such a file
> without complaints.  JCasGen also generates Java files
> without complaining, resulting in invalid Java code.
>
> Are these bugs or missing features, or is CDE/JCasGen
> simply not meant to be used this way?
>
> (Note that I can't add a duplicate feature in CDE,
> there checking does happen.)
>
> --Thilo
>
>
>