Eclipse Community Forums
Forum Search:

Search      Help    Register    Login    Home
Home » Modeling » ATL » [Theoretical?] Transformation problems not solvable with declarative languages?
[Theoretical?] Transformation problems not solvable with declarative languages? [message #1184111] Wed, 13 November 2013 08:21 Go to next message
Daniele G is currently offline Daniele G
Messages: 6
Registered: October 2013
Junior Member
There is a claim which states that "there are some transformation problems not solvable using a uniquely declarative transformation approach".

I've done some search on the web, but didn't find anything useful.
Could someone here point me to an example?

Cheers,
gd
Re: [Theoretical?] Transformation problems not solvable with declarative languages? [message #1184692 is a reply to message #1184111] Wed, 13 November 2013 16:23 Go to previous messageGo to next message
Massimo Tisi is currently offline Massimo Tisi
Messages: 5
Registered: February 2010
Junior Member
In very generic terms (*any* declarative approach) the claim looks clearly false to me.

If we limit the claim to the declarative subset of ATL, we believe (but nobody proved it formally) that we can represent as a declarative transformation any computable function among models.

The set of problems that you indicate with "transformation problems" may be however wider than the set of "computable functions among models". For instance they may include interactive transformations with user inputs during the execution (ATL does not address this case).

As a more interesting example, you may have specific requirements about the transformation that go beyond its functional meaning, for instance requirements about the identity of model elements (e.g., this source element and this target element must be the same element) or about the structure of the set of trace links. Because of language constraints (e.g., multiple trace links starting from the same source element are not allowed in ATL) not all these problems are solvable in ATL. However more research is needed to formally define the set of these problems.
Re: [Theoretical?] Transformation problems not solvable with declarative languages? [message #1186227 is a reply to message #1184692] Thu, 14 November 2013 15:11 Go to previous message
Daniele G is currently offline Daniele G
Messages: 6
Registered: October 2013
Junior Member
Very comprehensive response, Thank you.
Previous Topic:Empty output model with ATL+ANT and metamodel from xsd file
Next Topic:ANT+ATL with inter-model references
Goto Forum:
  


Current Time: Thu Oct 02 00:20:23 GMT 2014

Powered by FUDForum. Page generated in 0.05397 seconds
.:: Contact :: Home ::.

Powered by: FUDforum 3.0.2.
Copyright ©2001-2010 FUDforum Bulletin Board Software