|Re: Use of ATL [message #1016838 is a reply to message #1016754]
||Thu, 07 March 2013 20:47
|| Dennis Wagelaar
Registered: September 2012
Op 07-03-13 16:19, Federico Toledo schreef:|
> I can say that I'm happy using it. I've tried a couple of model-to-model
> languages before, and till the moment this is the best for me. I didnt try
> with "huge".. but I did with what I consider pretty big UML files, processing
> not only class diagrams, but also activity diagrams and sequence diagrams,
> using profiles (with stereotypes...). My script can process around a hundred
> of concepts in less than 2 seconds. Of course, sometimes you have to deal with
> some "situations"... but you can find useful help in this forum.
> Hope you find this info helpful
ATL typically works with any EMF models you can fit in your computer's memory.
CDO models that are accessed over the network tend to work less well, as ATL
should work "close to the data"; ATL will iterate over the entire input model
a couple of times (typically once for every rule). Obviously, your metamodel
needs to fit into your memory; this is an EMF requirement.
Depending on whether and what "manual" Java code you embedded in your
metamodel implementation (e.g. UML2 has quite some manual Java code as part of
its metamodel), ATL can run into problems. For UML2, these problems can be
worked around. YMMV. Metamodels without custom code should always work with ATL.
In practical terms, I've worked with +-15 MB UML models (Java 1.6 API), and
managed to transform those into OWL ontologies in a time varying from 2 to 20
minutes. Model comparison between a large (15MB) a small (0.5MB) UML model can
be done in a matter of seconds.
Powered by FUDForum
. Page generated in 0.01687 seconds