[news.eclipse.tools.emf] Re: EMF C++ runtime support?

["Jose Hernandez" <jose.hernandez@xxxxxxxxxx>]

Hi Reinhold,

Your EMF for .net project looks interesting.  I must admit your that your 
list of challenges is not what I expected.  I thought that the real 
complexity in a multi-platform EMF system would be in the code that binds 
all these implementations together rather than the C++ specific issues that 
you mentioned.

Coming to Java from the C++ world, I am not inclined to think the issues 
that you mentioned are insurmountable but as you well point out, the 
difficulty often lies in the functionally that has no direct correspondence 
between Java and C++.

I worked for a number of years on cross platform Solaris/Windows products 
written in C and C++ and in my personal experience C++ portability isn't a 
real worry once you understand the limitations of the compilers you are 
working with.  I don't think it would be particularly difficult or 
complicated to extend the STL map and list templates to include EMF support. 
Introspection might be an issue though.

I guess I will need to take a look at your EMF for .net project to see what 
is involved.

Thank you all for your replies.

Regards,
Jose

-----

"Reinhold Bihler" <reinhold.bihler@xxxxxx> wrote in message 
news:fo0rvt$k18$1@xxxxxxxxxxxxxxxxxxxx
Jose,

C++ code generation and runtime support for EMF would be very cool! In the 
company
I am working for we use C++, Java and C# for the development of our 
products. There is
often the need to exchange modeled data between software components written 
in
different languages. That was one reason why I started the EMF4Net project 
Ed mentioned
(see also www.emf4net.org and wiki.eclipse.org/EMF4Net_Proposal for more 
information)
which basically tries to translate EMF's Java runtime to C#.

Translating EMF's runtime to C++ I see at least the following challenges:

- C++ lacks the concept of a garbage collector for automatic memory 
management. This
   could be solved by using a smart pointer concept (see e.g. 
http://www.boost.org/libs/smart_ptr/smart_ptr.htm )

- C++ lacks imho a "real" standard library. I know there is the STL but in 
comparison to the
   Java SDK API or the .Net SDK libraries the provided functionality is 
very limited. The EMF
   runtime uses different list and map implementations. This is done by 
extending basic list and
   map implementations. I am not sure how this could be done by extending 
the STL list and map
   implementations (I am by no means a C++ guru but guess a real guru would 
find a way ;-).
   Probably there are open source C++ libraries that could help to overcome 
C++ lack of standard
   libraries. Apache has some very good C++ libraries like the C++ 
implementation of Xerces
   and there is the boost project (http://www.boost.org/).

- <rant on>In the C/C++ world every compiler seems to follow its own 
standard so providing
   a portable C++ library is a challenge on its own</rant off>. However, 
for a first step a
   implementation for a wide spread compiler like gcc, Intel's or 
Microsoft's C++ compilers
   would do. The libary could then be ported to other compilers later on.

The C++ code generation part is not a real problem as the EMF code 
generation and Java
code templates can be extended and modified easily.

So my conclusion is that a C++ implementation of EMF's runtime could be done 
especially as
the eclipse community and the EMF guys in special are very helpful. And 
there is also the
eclipse CDT project which might be used to develop and integrate a EMF4C++ 
implementation.

It is definitively a challenge but more a matter of the willingness to 
invest real efforts. It would
make a cool sub project to EMF and a wonderful contribution to the 
community. Hint, hint.. ;-)

Best regards,

Reinhold
"Ed Merks" <merks@xxxxxxxxxx> wrote in message 
news:fnvk0a$7oe$1@xxxxxxxxxxxxxxxxxxxx
Jose,

The EMF4Net guys did it for C# but that's a lot easier given how much C# is 
like Java:

http://www.eclipse.org/modeling/emft/?project=emf4net

I hope someone response to you query as well.  I'd be very interested!


Jose Hernandez wrote:
Thanks Ed.

I will be interested to know if anyone reading this thread can offer any
hints as to how one might go about creating an C++ EMF runtime.

Jose

"Ed Merks" <merks@xxxxxxxxxx> wrote in message
news:fnvdsr$nga$1@xxxxxxxxxxxxxxxxxxxx

I'm not aware of specifics, but I do recall someone telling me a few years
back about their organizations's effort in this regard.

Jose Hernandez wrote:

Does anyone know if there is any such thing as a version of the EMF
runtime for C++?