| [Epsilon] Re: Comparative study between ATL and ETL [message #376864] |
Wed, 18 April 2007 11:41  |
 Dimitrios Kolovos
Messages: 1776
Registered: July 2009 |
Senior Member |
|
|
Hi Laurent,
Please take my answer with a grain of salt since I am the principal
developer of ETL and I may be a little biased :) Perhaps Frederic or
Mikael can correct any inaccuracies/omissions.
In general, ETL is more imperative than ATL. For example, in ETL
operations (the equivalent of ATL helpers) (as well as in ETL rules) you
can write sequences of statements that query/modify any of the involved
models (source or target). Although ATL also supports imperative
statements (in the do block), it has a number of limitations such as
that you cannot define new variables in an imperative block (see p. 36
in the ATL manual for this and other limitations of ATL imperative
blocks). Finally, in the latest version of ETL you can define and access
native Java objects (through the Native datatype: e.g. var f : new
Native('javax.swing.JFrame');)
On the other hand, ATL is more mature (an official m2m component) than
ETL and supports additional features such as multiple iterators in
collection-iterating operations. Also, each ATL transformation is a
model itself (ATL has a proper metamodel and tools for
injecting-extracting textual specifications into/from ATL models) and
this enables users to write higher-order transformations (this is not
implemented for ETL yet).
I think it would be a very interesting contribution if you could come up
with exemplar transformations expressed both in ATL and ETL
(particularly between metamodels that are not structurally close to each
other) and report on the amount/complexity of the required code in each
language.
Cheers,
Dimitrios
P.S. If you have any further questions on ETL, please add the GMT
newsgroup (eclipse.modeling.gmt) in your "to" list, prefixing your
message with [Epsilon].
Laurent Merlet wrote:
> Hi all,
>
> I'm currently working on a comparative study between ATL and ETL
> (Epsilon Transformation Language).
> I have already test some examples using ATL & ETL plugins for eclipse.
> But I dont't see exactly what kind of differences exists between this 2
> transformation language.
>
> Is someone can help me...
>
> Best regards,
>
> Laurent MERLET
|
|
|
| Re: [Epsilon] Re: Comparative study between ATL and ETL [message #376866 is a reply to message #376864] |
Wed, 18 April 2007 19:10  |
Eclipse User |
|
|
|
Originally posted by: merlet.l.free.fr
Hi Dimitrios,
Thank you for your answer. It gives me a great clarification on main
differences between those languages.
I am sure that you are not corrupted by your core participation in ETL ;-)
I wish I would have enought time in my study to build a report on an
exemple of metamodel transformation like you mentionned.
Best regards,
Laurent MERLET
Dimitrios Kolovos a écrit :
> Hi Laurent,
>
> Please take my answer with a grain of salt since I am the principal
> developer of ETL and I may be a little biased :) Perhaps Frederic or
> Mikael can correct any inaccuracies/omissions.
>
> In general, ETL is more imperative than ATL. For example, in ETL
> operations (the equivalent of ATL helpers) (as well as in ETL rules) you
> can write sequences of statements that query/modify any of the involved
> models (source or target). Although ATL also supports imperative
> statements (in the do block), it has a number of limitations such as
> that you cannot define new variables in an imperative block (see p. 36
> in the ATL manual for this and other limitations of ATL imperative
> blocks). Finally, in the latest version of ETL you can define and access
> native Java objects (through the Native datatype: e.g. var f : new
> Native('javax.swing.JFrame');)
>
> On the other hand, ATL is more mature (an official m2m component) than
> ETL and supports additional features such as multiple iterators in
> collection-iterating operations. Also, each ATL transformation is a
> model itself (ATL has a proper metamodel and tools for
> injecting-extracting textual specifications into/from ATL models) and
> this enables users to write higher-order transformations (this is not
> implemented for ETL yet).
>
> I think it would be a very interesting contribution if you could come up
> with exemplar transformations expressed both in ATL and ETL
> (particularly between metamodels that are not structurally close to each
> other) and report on the amount/complexity of the required code in each
> language.
>
> Cheers,
> Dimitrios
>
> P.S. If you have any further questions on ETL, please add the GMT
> newsgroup (eclipse.modeling.gmt) in your "to" list, prefixing your
> message with [Epsilon].
>
> Laurent Merlet wrote:
>> Hi all,
>>
>> I'm currently working on a comparative study between ATL and ETL
>> (Epsilon Transformation Language).
>> I have already test some examples using ATL & ETL plugins for eclipse.
>> But I dont't see exactly what kind of differences exists between this 2
>> transformation language.
>>
>> Is someone can help me...
>>
>> Best regards,
>>
>> Laurent MERLET
|
|
|
| Re: [Epsilon] Re: Comparative study between ATL and ETL [message #580264 is a reply to message #376864] |
Wed, 18 April 2007 19:10 |
| Eclipse User |
|
|
|
Originally posted by: merlet.l.free.fr
Hi Dimitrios,
Thank you for your answer. It gives me a great clarification on main
differences between those languages.
I am sure that you are not corrupted by your core participation in ETL ;-)
I wish I would have enought time in my study to build a report on an
exemple of metamodel transformation like you mentionned.
Best regards,
Laurent MERLET
Dimitrios Kolovos a écrit :
> Hi Laurent,
>
> Please take my answer with a grain of salt since I am the principal
> developer of ETL and I may be a little biased :) Perhaps Frederic or
> Mikael can correct any inaccuracies/omissions.
>
> In general, ETL is more imperative than ATL. For example, in ETL
> operations (the equivalent of ATL helpers) (as well as in ETL rules) you
> can write sequences of statements that query/modify any of the involved
> models (source or target). Although ATL also supports imperative
> statements (in the do block), it has a number of limitations such as
> that you cannot define new variables in an imperative block (see p. 36
> in the ATL manual for this and other limitations of ATL imperative
> blocks). Finally, in the latest version of ETL you can define and access
> native Java objects (through the Native datatype: e.g. var f : new
> Native('javax.swing.JFrame');)
>
> On the other hand, ATL is more mature (an official m2m component) than
> ETL and supports additional features such as multiple iterators in
> collection-iterating operations. Also, each ATL transformation is a
> model itself (ATL has a proper metamodel and tools for
> injecting-extracting textual specifications into/from ATL models) and
> this enables users to write higher-order transformations (this is not
> implemented for ETL yet).
>
> I think it would be a very interesting contribution if you could come up
> with exemplar transformations expressed both in ATL and ETL
> (particularly between metamodels that are not structurally close to each
> other) and report on the amount/complexity of the required code in each
> language.
>
> Cheers,
> Dimitrios
>
> P.S. If you have any further questions on ETL, please add the GMT
> newsgroup (eclipse.modeling.gmt) in your "to" list, prefixing your
> message with [Epsilon].
>
> Laurent Merlet wrote:
>> Hi all,
>>
>> I'm currently working on a comparative study between ATL and ETL
>> (Epsilon Transformation Language).
>> I have already test some examples using ATL & ETL plugins for eclipse.
>> But I dont't see exactly what kind of differences exists between this 2
>> transformation language.
>>
>> Is someone can help me...
>>
>> Best regards,
>>
>> Laurent MERLET
|
|
|
Powered by
FUDForum. Page generated in 0.02011 seconds
] 
Search
Help
Register
Login
Home
