Inline variable initialization

[Madars Vitolins <m...@silodev.eu>]

Hello!
	What is current Ode status with inline variable initialization? Does
someone works on it or maybe it is finished? From Jira seems it is open
and unassigned ( http://issues.apache.org/jira/browse/ODE-236 )
	Can I take over it? And guys can you point out me what directions I
should do research? Would it affect BPEL compiler?
	I see possible way that compiler would generate extra object model for
variable initialization in places where execution is performed - in
processes or scopes. For <sequence> activities that would be extra code
in the start of element. For other activities it would require to
generate extra "<sequence>" to moderate variable initialization and
other activities. What do you think? Is this right way?

Thanks a lot in advance,
Madars

Re: Inline variable initialization

[Matthieu Riou <ma...@gmail.com>]

On Sun, Feb 8, 2009 at 12:22 PM, Madars Vitolins <m...@silodev.eu> wrote:

> Hello!
>        What is current Ode status with inline variable initialization? Does
> someone works on it or maybe it is finished? From Jira seems it is open
> and unassigned ( http://issues.apache.org/jira/browse/ODE-236 )


It's filed in Jira and in our compliance page but nobody so far has
volunteered to implement it.


>
>        Can I take over it?


Sure, that'd be pretty cool :)


> And guys can you point out me what directions I
> should do research? Would it affect BPEL compiler?


It would definitely affect the compiler. You'll probably have to modify some
runtime classes too but I don't foresee anything that would break backward
compatibility. Although we'll have to keep an eye on it.


>
>        I see possible way that compiler would generate extra object model
> for
> variable initialization in places where execution is performed - in
> processes or scopes. For <sequence> activities that would be extra code
> in the start of element. For other activities it would require to
> generate extra "<sequence>" to moderate variable initialization and
> other activities. What do you think? Is this right way?
>

Depends what you mean :) So I don't think we can implement this strictly as
syntax sugar (only in the compiler, without modifying any OModel (the O*
classes) or runtime classes). When a process starts, it's whole container
chain starts but there's no way to stick activities before that. So the
initialization will have to be done in the scope the variable is in and we
can't really try to have the compiler generate an assignment before.

My expectations would be that you'll have to modify OScope.Variable to add
the from-spec (reusing the ones in OAssign probably) and the compiler to set
that. Then at runtime, you'll have to alter SCOPE to set the variables when
it gets instantiated.

Hope that helps.

Cheers,
Matthieu


>
> Thanks a lot in advance,
> Madars
>
>