Home » Modeling » EMF "Technology" (Ecore Tools, EMFatic, etc) » Workflow Engine ???
|
| Re: Workflow Engine ??? [message #116785 is a reply to message #116572] |
Thu, 03 April 2008 20:22   |
 Bernd Kolb
Messages: 57
Registered: July 2009 |
Member |
|
|
Hi Bryan,
Yes, this is the right place, and yes the project is still active.
The workflow engine is supposed to be an integration project for EMP. It
grew out of openArchitectureWare and has been used there for a few years
now. The idea is that a user of MWE can describe a sequence of steps
which should be executed in order to process a model. (E.g. read an UML2
Model -> Validate the model -> Transform it into an Ecore Model ->
delete a directory on your hard drive -> run a generator to generate
code from that model). As you noticed, you can create your own Workflow
Components. In there you'll find two methods: One which is executed
before the engine starts to run (checkConfiguration) and another one
which is called by the engine and supposed to do the real job (invoke)
At the moment wf-components can communicate using the wf-context. It is
basically a map where a component can read data from and put data in.
The wf-engine also implements the Eclipse debug framework. Additionally
it provides you with extension points so your components can plug into
that and thus make you component debuggable (we are using that to
provide a debugger for the M2T's Xpand languages).
For the time being it is not EMF based (for mostly historical reasons)
however this is one thing we'd like to change in the future. Also, we'd
like to have better support for static analysis. Another thing is that
we might change the language from XML to something more concise, but
this is not yet decided. (However if we do we will continue to support
the xml for a while). At the moment we are refactoring and rewriting our
Xpand contribution and thus the focus has been shifted a bit away from
MWE. However, MWE is fully usable and has used (with a different
namespace ;-)) in many projects in the past. Your help would
definitively be welcome!
From what I can see in your last sentence, this is what you have been
looking for :-). For a simple example, you might want to have a look at
the DirectoryCleaner shipped with MWE.
Whenever you have other questions, please don't hesitate to ask!
HTH,
Bernd
Bryan Hunt schrieb:
> Is this the right place to discuss / ask questions about MWE?
>
> Is the project active (there haven't been any builds since January)?
> Is there any architectural design documentation / notes that I can look at?
> What exactly does this "workflow engine" do?
> The wiki seems to indicate it's for "generators" - generation of what?
> I'm curious as to why the domain model is not based on EMF?
> What would this project like to be when it grows up?
> Is the project looking for help?
>
> I've been experimenting with my own "workflow engine" so I'm curious as
> to whether your concept of a workflow engine is the same as mine beyond
> the buzzwords. Looking a bit deeper at the wiki, it appears that I
> could define a WorkflowComponent that did anything (say running command
> line programs like /bin/ls), and the engine would stitch the components
> together and execute them in the proper order. If that's the case, then
> this might be what I'm looking for.
>
> Bryan
>
|
|
| |
| Re: Workflow Engine ??? [message #116832 is a reply to message #116809] |
Fri, 04 April 2008 11:08 |
Eclipse User |
|
|
|
Originally posted by: merks.ca.ibm.com
Bryan,
When reviewing the MWE work with Bernd, the fact that the MWE model
wasn't an EMF model struck me immediately. Shocking I'd noticed that
hey? :-P We had discussions about the fact that the engine needs to
work stand alone, which lead to further discussions about the fact that
EMF models do work stand alone.
I think a community works best when it's diverse with lots of people
contributing, so I imagine that working on a prototype EMF model that
corresponds to the hand written "model" in MWE today would be a very
useful contribution and an excellent way to get involved in the project
even as a committer.
Bryan Hunt wrote:
> Bernd, thanks for replying. If I were to come up with an example EMF
> model, would that model have to conform to the existing code base, or
> would I have the freedom to start with a fresh, simple design that
> could be built upon? I've poked around a bit at the source code and
> my ideas on this topic appear very similar to yours, but there are
> some things in the source that I don't understand, so I'd like to
> start with a simple model and go from there. I think better when I
> can see the model :)
>
> Bryan
>
> On 2008-04-03 15:22:00 -0500, Bernd Kolb <b.kolb@kolbware.de> said:
>
>> For the time being it is not EMF based (for mostly historical reasons)
>> however this is one thing we'd like to change in the future.
>
|
|
| |
| Re: Workflow Engine ??? [message #116856 is a reply to message #116832] |
Fri, 04 April 2008 14:23 |
Bernd Kolb
Messages: 57
Registered: July 2009 |
Member |
|
|
In fact as mentioned in my other note, I think in a first step is has
not to be compatible with the current one. One thing to keep in mind
though is we have basically 3 meta classes at the moment. The benefit of
"just" replacing them with Ecore classes is not too big ;-) So when
doing that, we also want to have a clean up (meaning simplify some of
the stuff) in the code.
Bernd
Ed Merks schrieb:
> Bryan,
>
> When reviewing the MWE work with Bernd, the fact that the MWE model
> wasn't an EMF model struck me immediately. Shocking I'd noticed that
> hey? :-P We had discussions about the fact that the engine needs to
> work stand alone, which lead to further discussions about the fact that
> EMF models do work stand alone.
> I think a community works best when it's diverse with lots of people
> contributing, so I imagine that working on a prototype EMF model that
> corresponds to the hand written "model" in MWE today would be a very
> useful contribution and an excellent way to get involved in the project
> even as a committer.
>
>
> Bryan Hunt wrote:
>> Bernd, thanks for replying. If I were to come up with an example EMF
>> model, would that model have to conform to the existing code base, or
>> would I have the freedom to start with a fresh, simple design that
>> could be built upon? I've poked around a bit at the source code and
>> my ideas on this topic appear very similar to yours, but there are
>> some things in the source that I don't understand, so I'd like to
>> start with a simple model and go from there. I think better when I
>> can see the model :)
>>
>> Bryan
>>
>> On 2008-04-03 15:22:00 -0500, Bernd Kolb <b.kolb@kolbware.de> said:
>>
>>> For the time being it is not EMF based (for mostly historical reasons)
>>> however this is one thing we'd like to change in the future.
>>
|
|
|
| Re: Workflow Engine ??? [message #116870 is a reply to message #116856] |
Fri, 04 April 2008 15:50 |
| Eclipse User |
|
|
|
Originally posted by: merks.ca.ibm.com
Bernd,
Blasphemy! A real model always has value.
Bernd Kolb wrote:
> In fact as mentioned in my other note, I think in a first step is has
> not to be compatible with the current one. One thing to keep in mind
> though is we have basically 3 meta classes at the moment. The benefit
> of "just" replacing them with Ecore classes is not too big ;-) So when
> doing that, we also want to have a clean up (meaning simplify some of
> the stuff) in the code.
>
> Bernd
>
>
> Ed Merks schrieb:
>> Bryan,
>>
>> When reviewing the MWE work with Bernd, the fact that the MWE model
>> wasn't an EMF model struck me immediately. Shocking I'd noticed that
>> hey? :-P We had discussions about the fact that the engine needs to
>> work stand alone, which lead to further discussions about the fact
>> that EMF models do work stand alone.
>> I think a community works best when it's diverse with lots of people
>> contributing, so I imagine that working on a prototype EMF model that
>> corresponds to the hand written "model" in MWE today would be a very
>> useful contribution and an excellent way to get involved in the
>> project even as a committer.
>>
>>
>> Bryan Hunt wrote:
>>> Bernd, thanks for replying. If I were to come up with an example
>>> EMF model, would that model have to conform to the existing code
>>> base, or would I have the freedom to start with a fresh, simple
>>> design that could be built upon? I've poked around a bit at the
>>> source code and my ideas on this topic appear very similar to yours,
>>> but there are some things in the source that I don't understand, so
>>> I'd like to start with a simple model and go from there. I think
>>> better when I can see the model :)
>>>
>>> Bryan
>>>
>>> On 2008-04-03 15:22:00 -0500, Bernd Kolb <b.kolb@kolbware.de> said:
>>>
>>>> For the time being it is not EMF based (for mostly historical reasons)
>>>> however this is one thing we'd like to change in the future.
>>>
|
|
| | | | |
| Re: Workflow Engine ??? [message #616068 is a reply to message #116572] |
Thu, 03 April 2008 20:22 |
Bernd Kolb
Messages: 57
Registered: July 2009 |
Member |
|
|
Hi Bryan,
Yes, this is the right place, and yes the project is still active.
The workflow engine is supposed to be an integration project for EMP. It
grew out of openArchitectureWare and has been used there for a few years
now. The idea is that a user of MWE can describe a sequence of steps
which should be executed in order to process a model. (E.g. read an UML2
Model -> Validate the model -> Transform it into an Ecore Model ->
delete a directory on your hard drive -> run a generator to generate
code from that model). As you noticed, you can create your own Workflow
Components. In there you'll find two methods: One which is executed
before the engine starts to run (checkConfiguration) and another one
which is called by the engine and supposed to do the real job (invoke)
At the moment wf-components can communicate using the wf-context. It is
basically a map where a component can read data from and put data in.
The wf-engine also implements the Eclipse debug framework. Additionally
it provides you with extension points so your components can plug into
that and thus make you component debuggable (we are using that to
provide a debugger for the M2T's Xpand languages).
For the time being it is not EMF based (for mostly historical reasons)
however this is one thing we'd like to change in the future. Also, we'd
like to have better support for static analysis. Another thing is that
we might change the language from XML to something more concise, but
this is not yet decided. (However if we do we will continue to support
the xml for a while). At the moment we are refactoring and rewriting our
Xpand contribution and thus the focus has been shifted a bit away from
MWE. However, MWE is fully usable and has used (with a different
namespace ;-)) in many projects in the past. Your help would
definitively be welcome!
From what I can see in your last sentence, this is what you have been
looking for :-). For a simple example, you might want to have a look at
the DirectoryCleaner shipped with MWE.
Whenever you have other questions, please don't hesitate to ask!
HTH,
Bernd
Bryan Hunt schrieb:
> Is this the right place to discuss / ask questions about MWE?
>
> Is the project active (there haven't been any builds since January)?
> Is there any architectural design documentation / notes that I can look at?
> What exactly does this "workflow engine" do?
> The wiki seems to indicate it's for "generators" - generation of what?
> I'm curious as to why the domain model is not based on EMF?
> What would this project like to be when it grows up?
> Is the project looking for help?
>
> I've been experimenting with my own "workflow engine" so I'm curious as
> to whether your concept of a workflow engine is the same as mine beyond
> the buzzwords. Looking a bit deeper at the wiki, it appears that I
> could define a WorkflowComponent that did anything (say running command
> line programs like /bin/ls), and the engine would stitch the components
> together and execute them in the proper order. If that's the case, then
> this might be what I'm looking for.
>
> Bryan
>
|
|
| |
| Re: Workflow Engine ??? [message #616072 is a reply to message #116809] |
Fri, 04 April 2008 11:08 |
Ed Merks
Messages: 33140
Registered: July 2009 |
Senior Member |
|
|
Bryan,
When reviewing the MWE work with Bernd, the fact that the MWE model
wasn't an EMF model struck me immediately. Shocking I'd noticed that
hey? :-P We had discussions about the fact that the engine needs to
work stand alone, which lead to further discussions about the fact that
EMF models do work stand alone.
I think a community works best when it's diverse with lots of people
contributing, so I imagine that working on a prototype EMF model that
corresponds to the hand written "model" in MWE today would be a very
useful contribution and an excellent way to get involved in the project
even as a committer.
Bryan Hunt wrote:
> Bernd, thanks for replying. If I were to come up with an example EMF
> model, would that model have to conform to the existing code base, or
> would I have the freedom to start with a fresh, simple design that
> could be built upon? I've poked around a bit at the source code and
> my ideas on this topic appear very similar to yours, but there are
> some things in the source that I don't understand, so I'd like to
> start with a simple model and go from there. I think better when I
> can see the model :)
>
> Bryan
>
> On 2008-04-03 15:22:00 -0500, Bernd Kolb <b.kolb@kolbware.de> said:
>
>> For the time being it is not EMF based (for mostly historical reasons)
>> however this is one thing we'd like to change in the future.
>
Ed Merks
Professional Support: https://www.macromodeling.com/
|
|
| |
| Re: Workflow Engine ??? [message #616074 is a reply to message #116832] |
Fri, 04 April 2008 14:23 |
Bernd Kolb
Messages: 57
Registered: July 2009 |
Member |
|
|
In fact as mentioned in my other note, I think in a first step is has
not to be compatible with the current one. One thing to keep in mind
though is we have basically 3 meta classes at the moment. The benefit of
"just" replacing them with Ecore classes is not too big ;-) So when
doing that, we also want to have a clean up (meaning simplify some of
the stuff) in the code.
Bernd
Ed Merks schrieb:
> Bryan,
>
> When reviewing the MWE work with Bernd, the fact that the MWE model
> wasn't an EMF model struck me immediately. Shocking I'd noticed that
> hey? :-P We had discussions about the fact that the engine needs to
> work stand alone, which lead to further discussions about the fact that
> EMF models do work stand alone.
> I think a community works best when it's diverse with lots of people
> contributing, so I imagine that working on a prototype EMF model that
> corresponds to the hand written "model" in MWE today would be a very
> useful contribution and an excellent way to get involved in the project
> even as a committer.
>
>
> Bryan Hunt wrote:
>> Bernd, thanks for replying. If I were to come up with an example EMF
>> model, would that model have to conform to the existing code base, or
>> would I have the freedom to start with a fresh, simple design that
>> could be built upon? I've poked around a bit at the source code and
>> my ideas on this topic appear very similar to yours, but there are
>> some things in the source that I don't understand, so I'd like to
>> start with a simple model and go from there. I think better when I
>> can see the model :)
>>
>> Bryan
>>
>> On 2008-04-03 15:22:00 -0500, Bernd Kolb <b.kolb@kolbware.de> said:
>>
>>> For the time being it is not EMF based (for mostly historical reasons)
>>> however this is one thing we'd like to change in the future.
>>
|
|
|
| Re: Workflow Engine ??? [message #616075 is a reply to message #116856] |
Fri, 04 April 2008 15:50 |
Ed Merks
Messages: 33140
Registered: July 2009 |
Senior Member |
|
|
Bernd,
Blasphemy! A real model always has value.
Bernd Kolb wrote:
> In fact as mentioned in my other note, I think in a first step is has
> not to be compatible with the current one. One thing to keep in mind
> though is we have basically 3 meta classes at the moment. The benefit
> of "just" replacing them with Ecore classes is not too big ;-) So when
> doing that, we also want to have a clean up (meaning simplify some of
> the stuff) in the code.
>
> Bernd
>
>
> Ed Merks schrieb:
>> Bryan,
>>
>> When reviewing the MWE work with Bernd, the fact that the MWE model
>> wasn't an EMF model struck me immediately. Shocking I'd noticed that
>> hey? :-P We had discussions about the fact that the engine needs to
>> work stand alone, which lead to further discussions about the fact
>> that EMF models do work stand alone.
>> I think a community works best when it's diverse with lots of people
>> contributing, so I imagine that working on a prototype EMF model that
>> corresponds to the hand written "model" in MWE today would be a very
>> useful contribution and an excellent way to get involved in the
>> project even as a committer.
>>
>>
>> Bryan Hunt wrote:
>>> Bernd, thanks for replying. If I were to come up with an example
>>> EMF model, would that model have to conform to the existing code
>>> base, or would I have the freedom to start with a fresh, simple
>>> design that could be built upon? I've poked around a bit at the
>>> source code and my ideas on this topic appear very similar to yours,
>>> but there are some things in the source that I don't understand, so
>>> I'd like to start with a simple model and go from there. I think
>>> better when I can see the model :)
>>>
>>> Bryan
>>>
>>> On 2008-04-03 15:22:00 -0500, Bernd Kolb <b.kolb@kolbware.de> said:
>>>
>>>> For the time being it is not EMF based (for mostly historical reasons)
>>>> however this is one thing we'd like to change in the future.
>>>
Ed Merks
Professional Support: https://www.macromodeling.com/
|
|
| | | | |
Goto Forum:
Current Time: Wed Apr 24 19:12:09 GMT 2024
Powered by FUDForum. Page generated in 0.03718 seconds |
] 
Search
Help
Register
Login
Home
