Messages

1

General Board / Re: e-Book for Enterprise Architect available

« on: March 08, 2012, 08:31:25 pm »

not really, it's just a question of some days, then it's available again.
as far as i know the reason is the increased demand in the last time

2

General Board / Re: e-Book for Enterprise Architect available

« on: March 08, 2012, 06:55:43 am »

the book is not available in an electronic format:  
http://www.sparxsystems.eu/resources/ea-training-book/

3

General Board / Re: e-Book for Enterprise Architect available

« on: March 08, 2012, 01:54:03 am »

"Probably the only book about Enterprise Architect in the world"
Definitely not true
Sparxsystems Central Europe offers a book titled "Project Management with UML and Enterprise Architect"

here the table of content for the English edition:http://www.sparxsystems.eu/fileadmin/user_upload/pdfs/Buchcover/Table_of_content_8_eng.pdf
or German edition:  
http://www.sparxsystems.de/fileadmin/user_upload/pdfs/UML_mit_Enterprise_Architect_Inhaltsverzeichnis_9.1.pdf

and it's really a good one

4

General Board / Free edition of AMUSE Model debugger and simulator

« on: August 09, 2012, 11:24:47 pm »

http://www.lieberlieber.com/en/our-offering/amuse-22.html

Edition Comparison

AMUSE Feautes	Free Edition	Pro Edition
UML-Model execution in Enterprise Architect	•	•
Event simulation	•	•
Monitoring attribute values	•	•
Validation and auto-correction rules	•	•
Trace Viewer for simulation events	•	•
Automation of UML-Models using external ressorces	•	•
Automation external Components using UML-Models	•	•
Parallel execution and visualization of multiple instances		•
Message broadcasting between multiple instances		•
Editing of attribute values at runtime		•
Code Generation: UML => Embedded C		•

5

General Board / Re: How to mark a different sources of requirement

« on: March 09, 2012, 08:52:09 am »

what about exporting relationship matrix, where the package with the requirements is the source and the target?

6

General Board / proved numbers: MDD saves time and money

« on: March 08, 2012, 02:01:24 am »

Summary of the dissertation “Model Based Development of Embedded Software Systems in the Automotive – Costs and Benefits” Author: Sascha Kirstan; Academic Advisor: Prof. Dr. Dr. h.c. Manfred Broy; Technical University Munich

The dissertation defined and verified following 26 hypotheses – enjoy!

Hypothesis 1	Usage of MBSE saves at least 30% of time during development lifecycle	[highlight]Fully confirmed (up to 36%)[/highlight]
Hypothesis 2	Usage of MBSE leads to productivity improvement	Fully confirmed

Here the full story: http://blog.lieberlieber.com/2012/03/06/dissertation-summary-costs-and-benefits-of-model-based-development/

7

Uml Process / Re: Code Generation

« on: April 13, 2011, 12:56:05 am »

How ever , I repeat my question: why round-trip engineering at all?

Manufacturer should focus on usable code generation solution (let us start with PSM to PSI), which allows me to keep the same abstraction level at every development step (develop, debug, test).

Do you know even one?

8

Uml Process / Re: Code Generation

« on: April 13, 2011, 12:36:19 am »

...  
Then we're talking about taking a (in OMG terms) a PSM, generating an artifact into the real world then reverse engineering the artifact into a PSI (my invention - the Platform Specific Implementation.
...

Oh, it's basically a good idea to talk about it in OMG terms of Model Driven Architecture (MDA).
1. The term "problem domain" I described can be easily mapped to Business Model or Computation Independent Model (CIM)
2. The term "solution domain" is more complex and has to mapped to
 a) Platform Independent Model (PIM) is a technology-neutral system specification that conforms to a business model
 b) Platform Specific Model (PSM) is a technology specification that conforms to the PIM system specification
 c) And finally - Platform Specific Implementation (PSI)

The "easiest" part of code generation is conversion between PSM and PSI. It can be more or less complex depends on technologies and the decision what is the master. In case the code is the master, means, the user is allowed to modify the code after generation and he/she is allowed to use every feature of the target programming language also, then even that "easy" part becomes very-very complex.

The next level is the conversion from PIM to PSI. It is only possible if you do it in two steps: first one - modeling PIM, next one enriching it by the platform specific part but keeping it separately, so the user can change PSM without touching PIM later on. You can also make platform specific part exchangeable / configurable. In both cases it is not a trivial job.
I believe it is possible, but there are a number of challenges to be solved.

So, we are working on that  

9

Uml Process / Re: Code Generation

« on: April 08, 2011, 05:43:13 pm »

IOW: thanks a lot for a hint - now i realized a fact i basically knew all the time but wasn't really aware of it.

It's on of the biggest advantage of UML and obviously an issue for the most UML user at the same time:

UML uses nearly the same constructs / elements for describing
    the problem domain and the solution domain.

But in case of use cases it's clear, that type of diagrams shall be only used for describing the problem domain. And in general (99%) the problem has only indirectly to do with the solution.

How that?

How many solutions do you can provide for one simple use case (problem to solve): user wants from point A to point B??? I guess thousands. The job of the architect is to consider the non-functional requirements (and the customer is typically not aware of them) to provide an solution optimal for the customer.
By the way what is the difference between an architect and a key developer? You find the answer in the explanation above.

OK, back to the thread topic
In case we seriously intend to discuss the roundtrip engineering (it's basically the topic) we have only to consider the solution part, that means somebody understood the problem and designed a system intended to be executable.

10

Uml Process / Re: Code Generation

« on: April 01, 2011, 10:19:39 pm »

It depends on the platform and on the language basically. In general as long as UML-editor and IDE only share the code and not the model in behind, i agree - the complexity is too high. But for C# or Java an integrated solution is possible i guess.

But in my opinion the roundtrip engineering basically misses the point. The real profit of the code generation is the usage of higher abstraction level of UML/Model. As far as I start reworking the generated code i lose the advantage of UML/Modeling. That's the reason why roundtrip engineering will never be mainstream.

The basic problem of current tool chains is not the missing of reliable and usable roundtrip solutions but rather inhomogeneous abstraction level during development life cycle, or with other words: nobody would use C# if he has to debug assembler code after that. The gab between these two worlds overbalances the advantages of MDD.

As simple as that

11

Automation Interface, Add-Ins and Tools / Re: EA State Machine C code generator in 600 LOC

« on: November 06, 2012, 03:50:33 am »

try to hover your mouse over the state chart or generated code!

Chrom and Firefox renders the interactive SVG picture correctly and you can visually track the mapping between the parts of the UML state machine and the generated code out of them.

Excepting you use Internet Explorer, then you just see a static picture and have to open SVG picture separately via the link under picture...

12

Automation Interface, Add-Ins and Tools / EA State Machine C code generator in 600 LOC

« on: November 06, 2012, 03:32:29 am »

Alternative approach for EA code-generation-templates. As you know the EA does not generate any C-code for statemachines and the extending of templates for uml behaviors like state machines on your own is very tricky stuff.

Following article describes an alternative approach:
http://blog.lieberlieber.com/2012/11/05/a-custom-uml-state-machine-c-code-generator-in-600-lines/

The article provides very good and very short overview for
- Addin-development
- state machine -> code transformation

The described example needs just 600 lines of C# code. As it is a .NET code you can debug it, trace it, what ever you want and not can using EA templates. The described example supports:
- Primitive state machines
- Signal Triggers
- Composite States
- Entry- and Exit-Actions
- Transition Effects
- Decision Nodes

13

Automation Interface, Add-Ins and Tools / Re: Is it possible to call Addin from shape script

« on: April 04, 2012, 11:06:03 pm »

great, as usual - simple if you know how it works
thanks!

14

Automation Interface, Add-Ins and Tools / Is it possible to call Addin from shape script?

« on: April 04, 2012, 10:09:22 pm »

need to display some values not available via standard hasproperty method. Would be great if add-in calls could be used. Help says:

Properties for Element Shape Scripts
· addin (value returned from an Add-In)

Chapter "Display Element/Connector Properties"

but how??? I tried the call-syntax used in the code-generation templates - without success

15

Automation Interface, Add-Ins and Tools / Re: If you want understand the EA layer architectu

« on: March 13, 2012, 05:52:45 am »

Thanks for your question

No doubts - there are some highly important features could / should be implemented. But much more important: EA is a very affordable stable, fast, easy-to-handle and over years backward compatible visual modeling platform. If one of these criteria get overturned - it could be the end of the success story - as simple as that   The other point - Sparx Systems has only two dozen developers and it is good so. Thus I  well understand the reasons why some features take time.

However - the message is: EA is a platform. By the way, as far as I know it's possible to suppress the standard property dialog and replace it by your own. But i wouldn't recommend to go that way. Better you implement your own docked add-in window (look for Repository Class in the EA help especially for AddWindow method) and visualize there the properties in the way you or you customers need.

The only issue (depends on your needs) could be the EA data model. It's bit different to UML meta model, e.g. there is no basis class for all types of model elements. It will take time to learn it. That's why we build our own access layer over the EA interface. It makes the learning curve dramatically shorter. Currently it's reserved for internal use only, but we are discussing about alternatives. If you need more details, pls let me know.