Eclipse Community Forums: Visual Editor (VE)

Help

Home

Home » Archived » Visual Editor (VE) » Connections

Show: Today's Messages :: Show Polls :: Message Navigator

Connections [message #16831]

Mon, 26 January 2004 01:54

humour

Messages: 2
Registered: July 2009

Junior Member

I am very happy to see that the Visual Editor has been donated to Eclipse.
I was quite impressed when I tried WSAD 5.1 (especially after the version
5.0 that was quite buggy), and it is now good to have the same
possibilities in eclipse itself. Good stuff!

However, I am working on a large-scale Java/Swing project that has been
started with VAJ (as I used to work for IBM myself;-) ), and using a lot
of very useful Visual Programming especially connections between
properties, connection to code and the ability to have non Visual Beans.
And now that VAJ is not evolving anymore and that WSAD still not support
those features we are a bit stuck with our 800 something visual classes
(still developping with VAJ and its infamous JDK1.2.2!).

Visual programming is definitely a very powerful tool and I don't see any
pattern of GUI programming that can easily replace it especially for the
maintenability. Have you got any suggestions?

Is there any chance to have Visual Programming in a future version of the
Visual Editor (acording to what I read it should not be soon as there seem
to be other priorities :-()? What would you recommend to do in the
meantime?

Report message to a moderator

Re: Connections [message #16890 is a reply to message #16831]

Mon, 26 January 2004 09:18

Francesc RosÃ©s

Messages: 213
Registered: July 2009

Senior Member

Philippe,

I think the Visual Programming is not a goal for the
VE team. There are a lot of VAJ features no longer supported.
For example, the "Promote bean feature", the bean info editor,
etc. C'est dommage...

At this moment is the best VE for Swing in Eclipse. Konstantin
Scheglov (Instantiations) is working in a new VE and, I think,
Scott Stanchfield too.

Francesc

Philippe wrote:

> I am very happy to see that the Visual Editor has been donated to Eclipse.
> I was quite impressed when I tried WSAD 5.1 (especially after the version
> 5.0 that was quite buggy), and it is now good to have the same
> possibilities in eclipse itself. Good stuff!

> However, I am working on a large-scale Java/Swing project that has been
> started with VAJ (as I used to work for IBM myself;-) ), and using a lot
> of very useful Visual Programming especially connections between
> properties, connection to code and the ability to have non Visual Beans.
> And now that VAJ is not evolving anymore and that WSAD still not support
> those features we are a bit stuck with our 800 something visual classes
> (still developping with VAJ and its infamous JDK1.2.2!).

> Visual programming is definitely a very powerful tool and I don't see any
> pattern of GUI programming that can easily replace it especially for the
> maintenability. Have you got any suggestions?

> Is there any chance to have Visual Programming in a future version of the
> Visual Editor (acording to what I read it should not be soon as there seem
> to be other priorities :-()? What would you recommend to do in the
> meantime?

Report message to a moderator

Re: Connections [message #16920 is a reply to message #16831]

Mon, 26 January 2004 09:19

Francesc RosÃ©s

Messages: 213
Registered: July 2009

Senior Member

Report message to a moderator

Re: Connections [message #17034 is a reply to message #16920]

Mon, 26 January 2004 16:25

Scott Stanchfield

Messages: 263
Registered: July 2009

Senior Member

I might as well chime in...

I've been keeping Penumbra on hold to see where VE seems to be going.
While I think VE is a big plus for Eclipse, it doesn't really blend well
with my thoughts for what a Visual Designer should be, especially some
of the core concepts such as two-way synchronization, which I believe is
very restrictive to code generation. (In short, think about what would
happen if a C++ compiler had to be able to recreate C++ code from the
generated ASM code... Bye bye optimization and other cool stuff...)

I'm going to start working on Penumbra again. Hopefully I'll have some
time soon. I may decide to open source it and work with a small team of
folks who have more time than I to code -- not sure yet.

Later,
- Scott

In article <bv2m2d$6h9$1@eclipse.org>, froses@menta.net says...
> Philippe,
>
> I think the Visual Programming is not a goal for the
> VE team. There are a lot of VAJ features no longer supported.
> For example, the "Promote bean feature", the bean info editor,
> etc. C'est dommage...
>
> At this moment is the best VE for Swing in Eclipse. Konstantin
> Scheglov (Instantiations) is working in a new VE and, I think,
> Scott Stanchfield too.
>
> Francesc
>

Report message to a moderator

Re: Connections [message #17121 is a reply to message #17034]

Mon, 26 January 2004 22:12

Joe Winchester

Messages: 496
Registered: July 2009

Senior Member

Hi Scott,

> I've been keeping Penumbra on hold to see where VE seems to be going.
> While I think VE is a big plus for Eclipse, it doesn't really blend well
> with my thoughts for what a Visual Designer should be, especially some
> of the core concepts such as two-way synchronization, which I believe is
> very restrictive to code generation. (In short, think about what would
> happen if a C++ compiler had to be able to recreate C++ code from the
> generated ASM code... Bye bye optimization and other cool stuff...)

The VE is designed to be a framework as well as a finished GUI builder. I
don't want to argue the case on round tripping versus top down code
generation, basically because both points of view are correct. It's horses
for courses.

> I'm going to start working on Penumbra again. Hopefully I'll have some
> time soon. I may decide to open source it and work with a small team of
> folks who have more time than I to code -- not sure yet.

Don't give up on the VE just yet. It is designed into a series of plugins
that introduce functionality in well defined layers, and it has been used
successfully for AWT/Swing, SWT (in progress), Ultra Lightweight Client from
canoo (a J2EE rich client), and also internally within IBM for XML
serialization using XMLEncoding as well as AUIML visual builder (Application
user interface markup language for the i-Series). The VE layer that does
the code generation is separate from the one that does all the GUI stuff
with target VM creation, property sheet, layout management, etc... Right
now we just happen to generate code and to serialize this as our persistent
format because that's what the IBM marketing folks told us (and presumabely
what they found from talking to users and looking at the competition) was
what we should work on.

However ... one of the big reasons we wanted to become open-source was not
just to give away a GUI builder in Eclipse, but to try and foster the whole
idea of having an extensible framework for GUI builders. To do the kind of
high level VCE artifacts such as property connections or even higher than
that probably does require an XML model to capture this as metadata and top
down generation (just cause it's a lot harder to parse from source) but it
also requires a lot of GUI stuff that the VE does do in its lower layers.

What'd be really cool is you tried to leverage the VE to do this for
Penumbra (cool name by the way), and on the dev mailing list get a
discussion going about how to build an editor similar to what you have in
mind. Forget the VE as an end-user out of the box GUI builder with fixed
regeimes for how it generates and persists its model, but view it is instead
as a series of layers and which would you keep and which do you need to swop
out for your own ones.

Best regards,

Joe Winchester

Report message to a moderator

Re: Connections [message #17151 is a reply to message #17121]

Tue, 27 January 2004 11:42

Scott Stanchfield

Messages: 263
Registered: July 2009

Senior Member

In article <401590BE.CA787B6E@uk.ibm.com>, winchest@uk.ibm.com says...
> Hi Scott,

Hey Joe!

> > I've been keeping Penumbra on hold to see where VE seems to be going.
> > [...]
>
> The VE is designed to be a framework as well as a finished GUI builder. I
> don't want to argue the case on round tripping versus top down code
> generation, basically because both points of view are correct. It's horses
> for courses.

I'll agree it's not worth debating right now. It's really a matter of
whether one views "visual programming" as a higher-level representation
(3GL++) or simply another view on an existing language (3GL). (Note I
don't want to say 4GL or folks will start thinking of stuff like CSP
again. Not that a next-generation COBOL is a bad thing... Oh wait... ;)

>
> > I'm going to start working on Penumbra again. Hopefully I'll have some
> > time soon. I may decide to open source it and work with a small team of
> > folks who have more time than I to code -- not sure yet.
>
> Don't give up on the VE just yet. It is designed into a series of plugins
> that introduce functionality in well defined layers [...]

Understood, but I think that brings along more baggage than I'd like
right now. I'll be keeping a close eye on it, and if I feel it meets my
goals, I'll likely abandon Penumbra. I just don't think it does...

Ultimately the two tools may be able to leverage some of each other's
code, but of course that depends on licensing and how much tweaking is
necessary...

> However ... one of the big reasons we wanted to become open-source was not
> just to give away a GUI builder in Eclipse, but to try and foster the whole
> idea of having an extensible framework for GUI builders. To do the kind of
> high level VCE artifacts such as property connections or even higher than
> that probably does require an XML model to capture this as metadata and top
> down generation (just cause it's a lot harder to parse from source) but it
> also requires a lot of GUI stuff that the VE does do in its lower layers.

Agreed. There's a lot of basic plumbing that will be pretty common. Much
of this I see as GEF's job, but there are more advanced layers that can
be built on GEF. The trouble is, the more specialized those layers get,
the more a specific design philosphy gets embedded.

I haven't dug through the VE code yet; I've wanted to stay fairly clean-
room for now and listen to what's going on, what plans are being made,
etc. I like some of the concepts in VE, but I have a design plan that I
want to follow through on and see where it goes...

> What'd be really cool is you tried to leverage the VE to do this for
> Penumbra (cool name by the way),

Thanks -- it's nice when you stumble across a good name while chatting
with folks at work about a (real) lunar eclipse ;)

> and on the dev mailing list get a
> discussion going about how to build an editor similar to what you have in
> mind.

I'm definitely considering this, and at some point soon I may dive into
the code to feel it out.

> Forget the VE as an end-user out of the box GUI builder with fixed
> regeimes for how it generates and persists its model, but view it is instead
> as a series of layers and which would you keep and which do you need to swop
> out for your own ones.

I'm glad you're viewing it this way; I think it's the key to any next-
generation builder's success.

Later,
-- Scott

Report message to a moderator

Re: Connections [message #82950 is a reply to message #17151]

Wed, 09 March 2005 14:59

Luis Montes

Messages: 11
Registered: July 2009

Junior Member

Hey Scott,

Been a little more than a year, so I'm not sure if you're still hanging
around here.
I for one am still interested in some type of connector technology for
Eclipse.

Any progress?

Thanks,

Luis

On Tue, 27 Jan 2004 06:42:31 -0500, Scott Stanchfield wrote:

> In article <401590BE.CA787B6E@uk.ibm.com>, winchest@uk.ibm.com says...
>> Hi Scott,
>
> Hey Joe!
>
>> > I've been keeping Penumbra on hold to see where VE seems to be going.
>> > [...]
>>
>> The VE is designed to be a framework as well as a finished GUI builder. I
>> don't want to argue the case on round tripping versus top down code
>> generation, basically because both points of view are correct. It's horses
>> for courses.
>
> I'll agree it's not worth debating right now. It's really a matter of
> whether one views "visual programming" as a higher-level representation
> (3GL++) or simply another view on an existing language (3GL). (Note I
> don't want to say 4GL or folks will start thinking of stuff like CSP
> again. Not that a next-generation COBOL is a bad thing... Oh wait... ;)
>
>>
>> > I'm going to start working on Penumbra again. Hopefully I'll have some
>> > time soon. I may decide to open source it and work with a small team of
>> > folks who have more time than I to code -- not sure yet.
>>
>> Don't give up on the VE just yet. It is designed into a series of plugins
>> that introduce functionality in well defined layers [...]
>
> Understood, but I think that brings along more baggage than I'd like
> right now. I'll be keeping a close eye on it, and if I feel it meets my
> goals, I'll likely abandon Penumbra. I just don't think it does...
>
> Ultimately the two tools may be able to leverage some of each other's
> code, but of course that depends on licensing and how much tweaking is
> necessary...
>
>> However ... one of the big reasons we wanted to become open-source was not
>> just to give away a GUI builder in Eclipse, but to try and foster the whole
>> idea of having an extensible framework for GUI builders. To do the kind of
>> high level VCE artifacts such as property connections or even higher than
>> that probably does require an XML model to capture this as metadata and top
>> down generation (just cause it's a lot harder to parse from source) but it
>> also requires a lot of GUI stuff that the VE does do in its lower layers.
>
> Agreed. There's a lot of basic plumbing that will be pretty common. Much
> of this I see as GEF's job, but there are more advanced layers that can
> be built on GEF. The trouble is, the more specialized those layers get,
> the more a specific design philosphy gets embedded.
>
> I haven't dug through the VE code yet; I've wanted to stay fairly clean-
> room for now and listen to what's going on, what plans are being made,
> etc. I like some of the concepts in VE, but I have a design plan that I
> want to follow through on and see where it goes...
>
>> What'd be really cool is you tried to leverage the VE to do this for
>> Penumbra (cool name by the way),
>
> Thanks -- it's nice when you stumble across a good name while chatting
> with folks at work about a (real) lunar eclipse ;)
>
>> and on the dev mailing list get a
>> discussion going about how to build an editor similar to what you have in
>> mind.
>
> I'm definitely considering this, and at some point soon I may dive into
> the code to feel it out.
>
>> Forget the VE as an end-user out of the box GUI builder with fixed
>> regeimes for how it generates and persists its model, but view it is instead
>> as a series of layers and which would you keep and which do you need to swop
>> out for your own ones.
>
> I'm glad you're viewing it this way; I think it's the key to any next-
> generation builder's success.
>
> Later,
> -- Scott

Report message to a moderator