[emmanuel]

Hello all,

From the forum questions/discussions, it seems MyEclipseIde is at inception of providing a UML extension/plugin.

I would like such an extension. However, it seems like a massive task which could potentially sink the organisation tackling it.

From my experience working in Rose, XDE, Omondo, TogetherSoft and others, no one such tool provides comprehensive UML features per diagram. Each seems to have a niche; and by example, for the price tag of XDE, it seems to be immature; and each of these tools have pros and cons.

Seeing as MyEclipseIde seems to be at inception in this regard, would research be undertaken to analyse what is required from such an extension before commencing?

I propose that the community provide requests to be collated into requirements; and then that the community review the requirements and propose an implementation order; from this feedback, MyEclipseIde could plan a roadmap for implementing the features; or possibly even drop the idea altogether – due to its project size.

Kind regards
Emmanuel

[support-michael]

Emmanuel,

Thanks for the feedback.

<opinion>
Our current UML requirements are very basic and geared to Agile Modeller methods. That is, our initial focus to provide basic UML model “drawing” capability. If you can’t do that well then nothing else matters. I have used a number of the UML tools and my impression of the code-sync modelling tools is that they get in the way more than they help. I like reverse engineering tools for inspecting new codebases but that is where it stops for me. If I have to model at the code level I should just write the code directly since that is the objective. I respect a description I once heard for the word “model” – “less than the real thing.” That is my $0.02 USD on the topic. Management works out the strategy for how we deliver functionality without breaking the bank.
</opinion>

Regards,

Michael
MyEclipse Support Dude

[Mark Walker]

@support wrote:

Emmanuel,

Thanks for the feedback.

<opinion>
Our current UML requirements are very basic and geared to Agile Modeller methods. That is, our initial focus to provide basic UML model “drawing” capability. If you can’t do that well then nothing else matters. I have used a number of the UML tools and my impression of the code-sync modelling tools is that they get in the way more than they help. I like reverse engineering tools for inspecting new codebases but that is where it stops for me. If I have to model at the code level I should just write the code directly since that is the objective. I respect a description I once heard for the word “model” – “less than the real thing.” That is my $0.02 USD on the topic. Management works out the strategy for how we deliver functionality without breaking the bank.
</opinion>

Regards,

Michael
MyEclipse Support Dude

I have used a number of the UML tools and my impression of the code-sync modelling tools is that they get in the way more than they help. I like reverse engineering tools for inspecting new codebases but that is where it stops for me. If I have to model at the code level I should just write the code directly since that is the objective. I respect a description I once heard for the word “model” – “less than the real thing.”

Yeah,

what he said. Having the model rigourously and directly tied to code is a pain in the a$%$ especially when trying to do domain and analysis level modelling. Plus most of the so-called UML tools that do code sync don’t even support the full UML specification (I include both Rose and Together in this). I presume (and I could be wrong) that this is because it would make code/model synchronization more difficult. If therefore, the UML side of things is being limited by the code gen side of things, then I’d rather have the UML side implemented over and above code sync.

Ciao,

Mark Walker.

[drewmca]

I also agree. I like to have complete control over my model and I certainly like to have complete control over my code. I have found these to often be mutually exclusive. Most round-trip UML tools write code that doesn’t look the way I’d like. I think that UML models are great for analysis and as a starting point for forward-engineered code (provided you have complete control over the output; I use AndroMDA for this). But round-tripping just isn’t worth the hassle for me. I haven’t seen an implementation yet that worked the way I want. I certainly don’t want a bunch of lousy-looking ID codes littered throughout my code; what good is that to someone who doesn’t use the same tool I use? I think that modeling and coding should be loosely coupled. Make them work together when necessary but for the most part don’t make it required.

[Sjoerd van Leent]

Indeed most tools such as TogetherSoft and Rose have code sync. However, I think it’s much better to have an option: “Refactor/Implement class” in a popup menu and/or with a shortcut on the selected class in for example a class diagram. Furthermore, looking at UML it needs more than only a class diagram. Things as Sequence Diagrams, Deployment diagrams, Statechart diagrams and less used Use Case diagrams should be included as well.

Maybe it’s an idea to provide an extended class diagram which can have support for JSP/Servlet/JSF/Struts design as well. 😉

[joegaber]

I just became a registered MyEclipse member and recent NetBeans convert. I am very excited about the future direction of both tools. I use UML and modeling all the time. Most of my duties, these days, deals with application architecture and design more than coding; however, I’ve done coding for over 20 years. I have used Rose enough to know I don’t like it. I have been using a product (commercial) called MagicDraw for over 5 years now. It is a superb UML tool. It is far easier to learn than Rose, it is based on a plug-in extensibility model, and fully integrates for round-trip engineering in Eclipse (as well as others). I am not familiar with an open source UML tool as good; however, I would suggest looking at integrating one (if a good enough one exists) into MyEclipse rather than creating one yourself. I believe it would be a very poor business decision on the part of Genuitec to do so. I believe this from my experiences in creating a software development company in 1991 and selling to a public company in 1998. I believe you should focus your resources on what you do really well, and based on the comments made in this forum about UML in general, I wouldn’t think the company, or the developers, would have your hearts into it. I love what you’ve done with MyEclipse. I hope you keep doing more of the same – productivity for programming and integration of framework technologies!!

Best regards,

[Author]

Posts