On Sunday 11 January 2009 18:42:51 iker [..] ... wrote:
> Hello everyone,
>
> I have 3 Components:
>
> 1.- Component1 - Collect Data - Periodic
> 2.- Component2 - Process Data - NonPeriodic
> 3.- Component3 - Send Data - Periodic
>
> When Component1 collects all the data, it has to activate a Trigger, so
> that Component2 will activate.

So 2 needs to be a peer of 1, such that 1 can trigger it.

>
> Once all the data is processed, Data can be sent, but instead of send it
> all at once, it has to be sent in a periodic way using Component3.

So you'll need a buffer port between 2 and 3.

>
> After doing the first round, Component2 has to keep on waiting the Trigger
> (Component1 never stops collecting data), but Component3 has to continue
> sending the data (the previous cycle processed data). Component3's
> updateHook is always looking if the Processed Data is new data (every
> updateHook checks the ports).
>
> My question are:
>
> 1.- How can I configure the trigger, so that Component2 starts working
> after Component1 finished the data collection?

In RTT 1.6, you need to make 2 a peer of 1 and call (gasp):
peer->engine()->getActivity()->trigger() ; // code in 1's updatehook
Each time you want 2 to wake up.

In RTT 1.8, we have event driven data ports, which allow 2 to wake up each
time 1 wrote data.

>
> 2.- Component3 can not begin its execution until the first periodic is
> reached (after Component2 finishes the first data processing), but once the
> first period it is done, it has to be a periodic task. So it has to work
> like a periodic task with a delay.

This is a matter of taste. We don't allow switching between periodic and non-
periodic yet (you could write such an activity if you felt like it though). If
I were you, I would just poll and return from updateHoke() if there's no new
data. Otherwise, create an event which 2 will emit when data is ready, 3
subscribes to it and calls this->start() in the event handler. This would have
again been easier with 1.8-like event data ports.

Peter