Mailing-List: contact uima-user-help@incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: uima-user@incubator.apache.org
Received-SPF: pass (athena.apache.org: domain of msa@schor.com designates
 67.18.65.20 as permitted sender)
Message-ID: <4A8EB631.3030805@schor.com>
Date: Fri, 21 Aug 2009 10:58:57 -0400
From: Marshall Schor <msa@schor.com>
User-Agent: Thunderbird 2.0.0.23 (Windows/20090812)
MIME-Version: 1.0
To: uima-user@incubator.apache.org
Subject: Re: Creating an aggregate analysis engine on the fly
References: <6dfceae40908210147w5eb34281l44659a9318cfa915@mail.gmail.com>
In-Reply-To: <6dfceae40908210147w5eb34281l44659a9318cfa915@mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit

A good source to explore to see how this can be done is the test cases
for uima, which you can find in the uima source code; for instance, look
in the project uimaj-core, in the folder src/test/java.  One test which
might have examples to look at is the class AnalysisEngine_implTest. 

Of course, that is for Analysis Engines, and not for CPE's.

For that, you could look at the tests in the project uimaj-cpe, in the
folder src/test/java.  In the class CpmProcessingTest, there is a method
"setupCpm" that builds a cpe descriptor.

HTH.

-Marshall

Fabio Mancinelli wrote:
> Hi everybody,
>
> I've looked around but I didn't find anything about this, so I am
> asking to more expert people.
> The use case is the following:
>
> Imagine you have a set of analysis engines.
> By looking at their descriptors you want to synthesize at runtime an
> aggregate analysis engine with a custom control flow that takes a
> subset of those analysis engines and that is put into a CPE for
> execution.
>
> Now my problem is that since I don't know the analysis engines
> beforehand (I discover them at runtime) and the set could also change
> during the time, I cannot write an XML descriptor that specify the
> aggregate analysis engine.
>
> One solution to this problem is to have a "template XML" somewhere,
> fill it with the references to the analysis engines that will be
> chosen for composing the aggregate, write it somewhere and tell the
> CPE to use that descriptor for its cas processor . But this solution
> look quite clumsy and complicated.
>
> Is there another way to "manufacture" programmatically an aggregate
> analysis engine at runtime in order to realize this use case?
>
> Thanks,
> Fabio
>
>
>