Registered: Apr 2006
There is a large overlapping between ATL and QVT.
Both are rule based and use OCL.
Both are defined by a precise MOF metamodel.
Both accept XMI as input and output.
Apparently, at this point in time, ATL is one of the
most QVT-compliant open source transformation language.
Also due to the very modular nature of its model
transformation virtual machine, ATL has much potential
to be easily extended.
The paper mentioned signal some points that were
not found of immediate practical importance by the
ATL team and that corresponds to non implemented features. However, in most cases there are turn-around suggested possibilities. One example is bidirectional
transformations that could be rendered by a couple of
unidirectional transformation in thee AMMA context,
without any loss of practicality.
ATL should be considered in relation to other
open source tools of the AMMA platform like AMW,
AM3, TCS (Textual Concrete Syntax), etc. All these
tools extend the capabilities of ATL well beyond
the present characteristics of QVT.
The AMMA platform provides a set of Domain
Specific Transformation Languages, including
ATL, TCS, BCS, XCS, etc. As a consequence there
is less mismatch between :
- model to model transformation
- model to text transformation
- text to model transformation
The paper points out that ATL has been defined to solve
broader problems than QVT which is just focusing
on PIM to PSM. For example it should be possible
at some later time in the AMMA platform, to deal with binary file transformations.
This is an important difference with QVT which is
restricted to "XMI to XMI transformations", with
limited extensibility. We know very well that the
amount of real life native data encoded in XMI
is near zero. This is a serious limit to the generalization
of QVT. It is not very likely that the percentage
of XMI encoded data will increase much in the
All the best
Report this post to a moderator | IP: Logged