Transformation expressions are part of
the BPMN2 specification. For example Person.name assignment would
have a transformation expression defined which evaluates this
expression and its result is copied to the target reference.
Can't think right now why it would be useful to have data
association abilities as stand-alone outside of the process
modelling environment but maybe I don't understand what you mean.
Thanks.
On 8/16/12 7:12 AM, Mauricio Salatino wrote:
And we should think about having a component that can
be used outside the process designer as well.
When you say transformations what do you mean?
On Thu, Aug 16, 2012 at 12:11 PM,
Tihomir Surdilovic <tsurdilo@redhat.com> wrote:
Agreed, Designer already supports transformations to
do what you are asking for, but the data editors should
allow users to use this feature better. Will look at it.
Thanks
On 8/16/12 7:06 AM, Mauricio Salatino wrote:
Just a quick comment about
mapping data,
I think that we need a specialized component to
do that, because is a cross cutting concern.
No matter if we are mapping task inputs and
task outputs or a form field, we need a graphical
mapping component to take care of that.
We need a graphical way to do something like
Input: Person -----> Person.name ----
<mapped to> ----> name
and the other way around
Output Person.name <----<mapped to>---
name
On Thu, Aug 16, 2012 at
11:20 AM, Tihomir Surdilovic <tsurdilo@redhat.com>
wrote:
There are a number of ideas floating
around on this - and yes data mapping
within a BPMN2 process is probably now one
of the hardest things to set up and
maintain.
a) Have ability to map data outputs of a
task to data inputs of the connected
tasks(s) with a single button click
b) Ability to visualize the data flow of
the process
c) Notifications of dependency errors when
users decide to remove a process var,
globals, imports, etc
d) Ability to "copy" existing data
elements such as
variables/globals/imports/etc from an
already existing process or a process
template..this has been asked in the
community already and we are looking at
it.
jBPM follows the BPMN2 specification for
data setup and mapping closely which is a
good thing as it does not introduce any
wrappers around it (you might see other
tools use extensions elements for this
heavily). I don't think writing wrappers
is a good idea, however making working
with data as simple and intuitive as
possible for users is very important so
any ideas/thoughts/contributions are very
welcome :)
Jean-Jacques, If you want to jump on IRC
and I'd be happy to talk to you in detail
about all of this as well.
Thanks.
Tihomir
On 8/16/12 3:38 AM, Esteban Aliverti
wrote:
The problem is that
BPMN2 specification requires the variables
to be defined.
What you will need to do is to modify
Designer to automatically create the
required variables according to the IO
specification of the tasks you have.
Unfortunately this is not as easy as it
sounds ;).
Designer keeps a representation of
the process in JSON format. That JSON is
then converted into BPMN2 when it is
saved. When you open a process Designer
first converts it from BPMN2 to JSNO
back again.
You will need to tweak the
Marshalling/Unmarshalling process..
The classes in charge of this are:
org/jbpm/designer/bpmn2/impl/Bpmn2JsonMarshaller.java
and
java/org/jbpm/designer/bpmn2/impl/Bpmn2JsonUnmarshaller.java