Skip to main content
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [polarsys-iwg] UML Generators Proposal Question
Originally I wanted just to ask about Xtext/Xpand for code generation and maybe the reverse engineering code. 
There isn't any XText implementation for Java and C.
So, we plan to use directly CDT and JDT.
Regarding Sphinx and CDO as backend, it is a planning for Artop/EATOP afaik, and AUTOSAR models can really grow. Then I thought, that UML Models could also grow and it was rather a question for performance reason if Papyrus wouldn't benefit from this too. The thing is, that I read somewhere that Sphinx will use CDO only at runtime, loading the XML files into it, and saving it back to it, instead of only reading/writing to XML files.
I understand. But, we can manage very large models with EMF itself. I don't see the add-in value of Sphinx if EMF is well used. For exemple, EMF Compare success to manage 1,5GB of models, and it scales very well. 


* Regarding "Scripting"  EASE - http://wiki.eclipse.org/E4/Scripting


What is your question?
Well, that was also just one of the additonal questions popping up after the one about Xtext/Xpand for code generators. It was rather again about another contribution to Eclipse, or using/contributing to already existing Eclipse projects. How many different Eclipse Plugins does one have to install and learn to work with to see them die after 1 or 2 years, learning then a new Eclipse contribution? We have some AUTOSAR configuration and generation tools based on Eclipse+Artop, some use JET to get some code out, others use Java Plugins for generation, here comes Acceleo, then some use Xtext/Xpand. My job is not to develop model transformers, I'm an embedded SW developer. Java, JET, Acceleo, Xtext/Xpand are not really what I need to get my ECU running. 
They are just tools that should help me, not keeping me from working.

You are right.
And that why I prefer to keep a consistency of the technology used to execute code generators of UML Generators. And, as initial contributions are all created with Acceleo, as it uses a standard syntax, as this syntax is easy to understand for somebody who already know OCL (which is the case of UML users), as 100% of initial commiters know Acceleo and 0% know Xpand, it is really better to stay on Acceleo. Your proposition to use XPand going in the wrong way: it would complexified the architecture by adding a new dependance and a new syntax.
For Artop, Eclipse community can't help you as this isn't an OpenSource project, and it isn't in Eclipse Foundation. We can't be responsible of external projects's choices.
I currently have here an Eclipse installation, which idles with some open editors at 800MB RAM usage. At work I have additonal tools, that each want between 300 .. 700MB RAM, one tool (CANoe also requires after 2..3 hours simulation time at 2..4 GB RAM.
That's exactly why I prefer to not add non-required dependencies, and keep the architecture as simple as possible. 
The reason I ask (at least for Xtext/Xpand and Sphinx)?
* Several other projects like EATOP and ARTOP are going there
EATOP and ARTOP doesn't rely on UML. So, I don't really understand your need to integrate them with UML Generators.
Well, EATOP is about EAST-ADL, and there is an UML Profile for it in Papyrus.
OK, I understand. It already worked with EAST-ADL, but it was a pure Ecore metamodels. If you are engage in this project, perhaps it can make sense to propose them to be consistant with UML Generators by just proposed extension to the generic templates. 
* AFAIK, Papyrus is already using it for their editors.
* I wonder, if Xtext DSLs can be written for each language and maybe also easier extended, e.g. different compiler addons for embedded compilers. I just wanted to remind you for reverse engineering to the different dialects/extension regarding void _interrupt MyISR(void) vs. interrupt void MyISR() vs. void MyIsr(void) _interrupt 
        int @far / @near myVar;
        int myVar = 0 _at(0x0100);
        _inline vs. _inline_  vs. inline
        AUTOSAR like Compiler/Memory Abstractions
Thanks to template overriding feature of Acceleo, it's very easy to manage these different generation strategies with a simple property. Creating an XText DSL just for this is an heavy solution for a simple problem. And I prefer to use simple solution until possible. http://wiki.eclipse.org/Acceleo/Views_Perspective#Overrides_View
Ok, I'll take a look on it. Even though, I currently don't know, where to start to choose from Generation Tools.
But perhaps you have some complex use cases and this approach can make sense. I propose to wait to use the future official newsgroup for technical discussion.
* And last but not least, maybe there is one times a less spread amount of projects doing the same over and over. A lot of subprojects exist now on Eclipse-Foundation, and several have gone with the wind, but other pop up again. I'm personally overwhelmed by all the different kinds of subprojects and solutions. 
I agree.
And that exactly why we are launching this project.
Any contribution is welcomed!
If you want to be list as an interested party, you can ask (I just need to know if it's personal, of professional (in this case, I need the name of your company)).
Well, at the moment, I personally just try to follow/track/evaluate the current projects and development about UML and SysML, Code Generators and their applicability to the Automotive Field I'm working in. We are currently using Rhapsody at work, which is a bit tedious to use in a C environment. Also, sometimes I miss a tool/feature to express my ideas, like Rhasody missing the timing diagram, or when using just UML, the missing difference between Ports and Flow Ports of SysML, e.g. when specifing Client-Server Interfaces vs. Sender-Receiver Interfaces. The AUTOSAR tools we currently use are, well, somehow hard to use. I'm not a fan of editing an AUTOSAR configuration or define SW-Component Interfaces 
just in tree-based editors.
So, in my current projects we have about 5..8 different tools to change between and exchange data between. And given the different projects and versions, comparing different versions of the file types is getting harder and harder. Like merging Rhapsody versions with their DiffMerge tool, or AUTOSAR xml files on XML level, this isn't much fun.
Maybe this also explains my questions about the different Eclipse Plugins above. :o)
If you want to have an easy-to-use and easy-to-customize tool for Autosar, just try Sirius : http://www.eclipse.org/sirius With this Eclispse project, you can rely directly on the XML files of Artop. And when some concepts are missing (like port, interfaces, like with others models, ...), you can by yourself add them without the need to create java code. We did several years ago a prototype for Artop, and it worked very well : http://marketplace.obeonetwork.com/module/autosar 



Regards, kessel
_______________________________________________
polarsys-iwg mailing list
polarsys-iwg@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/polarsys-iwg




begin:vcard
fn:Etienne Juliot
n:Juliot;Etienne
org:Obeo
adr;quoted-printable:;;7 boulevard Amp=C3=A8re;Carquefou;;44481;France
email;internet:etienne.juliot@xxxxxxx
title:Vice President
tel;work:+33 2 51 13 55 94
tel;cell:+33 6 86 78 14 82
url:http://www.obeo.fr
version:2.1
end:vcard

Back to the top