uima-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mario Gazzo <mario.ga...@gmail.com>
Subject Re: Injecting nested resources using UIMA fit
Date Wed, 24 Jun 2015 07:41:22 GMT
Thanks Richard,

No hurry, I have a temporary workaround for it but I probably need to look into this soon.
I’ll keep you posted when I dig up some new findings.


> On 24 Jun 2015, at 09:29 , Richard Eckart de Castilho <rec@apache.org> wrote:
> Hi Mario,
> I'm not sure if the resource injection has a proper two-phase initialization
> (first instantiation, then injection) or if it behaves more like a constructor
> injection (injection during instantiation). I think it is the latter which
> requires that there is no circular injection - but that needs to be checked.
> I may not be able to get my hands into this in the next few days, so if have
> a need and want to dig into it more, feel free.
> Cheers,
> -- Richard
> On 19.06.2015, at 22:24, Mario Gazzo <mario.gazzo@gmail.com> wrote:
>> Hi Richard,
>> Seems something still spooky with my nested resources. This time I added an additional
resource, which depends on the same shared resource as another. I now have two different resources
dependent on the same external resource but during initialisation I get the error shown below.
I can use both resources independently of each other but just not together. I have been able
to replicate it by extending the earlier simple example (see further down).
>> You will notice in the example that I bind every resource to every description specified
even though there is no declared dependency. Reason is because our actual application has
a small DSL that aggregates the descriptions and then tries to bind resources on all possible
descriptions just before execution. However, I just did a naive approach where I bind every
resource to every possible description that was added but I didn’t do code that first inspects
whether a description has actually declared the dependency before attempting to bind. Yes,
its inefficient for a very large amount of resources but we don’t have that. My naive assumption
was that the UIMAfit binding process would figure this out way better than I could myself
and I just sticked with this since it worked fine until now and it was simple to program.
I don’t know whether this actually is the direct cause of the problem or whether my use
just triggered a bug. It works if I don’t do these superfluous bindings. If this use case
cannot be supported then I would like to know what would alternatively be the simplest way
to make these inspections on the descriptors before making a binding attempt?
>> Thanks a lot for your help.
>> Cheers
>> Mario

View raw message