[ve-dev] Re: Using JEM to draw cartoon images of widgets (Rich Kulp)

["Namrata" <namrata@xxxxxxxxxxxxxx>]

Hi Rich,

I have a few more follow up questions. I'm trying to understand what
ComponentManager does with the expression, and what the Adapter is expected
to do to the expression before it passes it to ComponentManager.

ComponentManager has two bean proxies - fComponentManagersProxy and
fComponentBeanProxy.

Do they correspond to the container and component (ex: panel and button)?

What does the method ContainerProxyAdapter::addComponentWithConstraint do? I
was looking at it while debugging the VE flow on adding a control. It calls
instantiateSettingBean(). Could not figure out what it instantiates. My
understanding is that it instantiates the constraint object. Like
Java.awt.BorderLayout.EAST. Is this correct?

Thanks
Namrata

> Hi,
>
> This was an oversight on our part. We just did a major conversion to using
> IExpressions for just about everything for performance reasons. We used to
> do everything through IMethodProxy's, but on a remote vm with a
> complicated java class the performance was not good. This was because each
> call to do an invoke on an IMethodProxy was round-trip. Talking through
> sockets to a remote vm using many small transactions has a lot of latency
> delay which adds up. So we changed to use IExpressions. These just keeping
> pushing the commands to the remote vm without stopping and waiting for a
> response until the entire transaction of all components and setting them
> up has been completed. At that time we wait for the expressions to be
> evaluated and completed. This helps tremendously when using remote vm's.
> However, IExpressions when using IDERegistry, has a significant
> performance overhead. This is because since everything is local, there is
> relatively no latency for the IMethodProxy invokes compared to the setup
> and processing costs of the IExpression itself.
>
> Now we had put this in recently and forgot that ComponentManager could be
> used outside of remote vm's, such as you are doing. So for now you need to
> wrapper the calls within a IExpression. We do intend to change
> ComponentManager to also allow direct non-expression access but I doubt we
> will get to it for VE1.1 because 1.1 is going into final test stage in
> about a week.
>
> Sorry for the inconvenience on this. You should open a defect on Bugzilla
> against VE subcomponent JFC/Swing and make me (richkulp@xxxxxxxxxx) the
> person to be assigned to asking that the ComponentManager be changed to
> work without IExpressions too. That way you will be notified when we do
> actually fix it.
>
> Thanks,
> Rich