Hello guys, I just received some explanation from the scxml w3c forum and what I was trying to do here does not follow Harel's Sematics.
So we have to stick on flow control statements (if, else, elseif) and so decide from there what part of the code gets executed and what part does not get executed. In my particular usecase I have a Bridge that gets calls from a Dispatcher class (which accepts GET/POST HTTP calls). I automatically call the init() method of the specific Bridge and if an exception occurs I just pass the control to the SCXML Engine throwing "app.error" event. Otherwise I forward the event corresponding to the service to the SCXML Engine which then go to an action state. >From that acton state some preconditions are evauated in the oninit handler and some of them can fire exceptions, however now it is different ... I cannot pass the control to the SCXML Engine hoping it will take care interrupting the flow. What I have to do is check return codes to decide if the onentry must continue using "if else" statements and later just test for those conditions in transaitions that lead the state machine to final targets. In those final targets or states the same process can be repeated. An important conclusion for my usecase is that within action states we never should handle events using transitions. They are mean to be used only from final states. I call those states final just to differentiate them from the action states. I guess a better term can be used for them but at this time I cannot come with a better clasification. Hope this help to anyone using SCXML to modelate Satate Oriented Business processes. Thanks!, Nestor --- Nestor Urquiza <[EMAIL PROTECTED]> wrote: > Sorry guys for the long thread but I continue making > tests and posting the information I get from them. > Now > I present an example in which a given error happens > in > the onentry event of the second (after initial) > state. > The error launch an event "app.error" that I handle > in > the new state and in the initial state just to > demonstrate that the FSM executes the transition > from > the initial state when it is in fact entering the > second state. Is this behavior correct? I would > expect > a different behavior here (going to aIDDLE2HD_ERROR > instead of going to IDDLE_ERROR) > test scxml file > ----------------- > <n:scxml version="1.0" initialstate="IDDLE" > xmlns:n="http://www.w3.org/2005/07/scxml" > xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" > xsi:schemaLocation="http://www.w3.org/2005/07/scxml > http://apis.buongiorno.us/white_labels/xml/scxml.xsd"> > <!-- first state uppon first inquire --> > <n:state id="IDDLE"> > <n:transition event="HandsetDetectionReq" > target="aIDDLE2HD"/> > <n:transition event="app.error" > target="IDDLE_ERROR"/> > </n:state> > <state id="aIDDLE2HD"> > <onentry> > <var name="dummy" > expr="HandsetDetectionReq.processError(12)"/> > </onentry> > <transition event="app.error" > target="aIDDLE2HD_ERROR"/> > </state> > <state id="IDDLE_ERROR"> > </state> > <state id="aIDDLE2HD_ERROR"> > </state> > </n:scxml> > > the logs: > -------- > Jul 11, 2006 6:06:10 AM > org.apache.commons.scxml.SCXMLExecutor logState > INFO: Current States: [IDDLE] > Jul 11, 2006 6:06:42 AM > org.apache.commons.scxml.SCXMLExecutor logState > INFO: Current States: [IDDLE_ERROR] > Jul 11, 2006 6:06:45 AM > org.apache.commons.scxml.SCXMLExecutor logState > INFO: Current States: [aIDDLE2HD] > > Thanks for any help, > Nestor > > --- Nestor Urquiza <[EMAIL PROTECTED]> wrote: > > > Going further thru the code and looking to some > tips > > from Rahul [1] I still am wondering if the only > > alternative must be to author custom actions > > extending > > the Commons SCXML Action abstract base class. > > > > In my specific case I cannot do so since I am > > relying > > in a lot of helper method I access from objects > that > > I > > have in the Root Context and for me it is > impossible > > to change all those objects. > > > > Consider for example: > > <n:var name="contentSupported" > > expr="userProfile.isContentSupported( askedContent > > )"/> > > > > userProfile is an instance of UserProfile class > > which > > cannot implement the Action abstract class. Still > I > > need any error in UserProfile#isContentSupported() > > to > > be able to call something like processError() that > > in > > turn can just suspend the execution of the state > > machine by means of calling exactly as suggested > in > > [1] : > > <transition event="err.foo" target="errorstate" /> > > > > I followed the code and it is not clear how I can > > accomplish this ... Maybe instead of > > SCXMLExecutor#triggerEvents we could add something > > like SCXMLExecutor#triggerException passing the > > name > > of the internal event to be passed (err.fo) and > > ensuring that that event can be the next step to > be > > done ignoring any other "standard" or "custom" > > action? > > > > I have been calling my var assignments with the > > result > > of my helper methods "custom actions" while I > > understand they are not really "custom actions" > from > > scxml definitions so I rather use thye term once > > suggested by Rahul I think "specific actions" > > > > Any feedback here?, Thanks > > > > [1] > > > http://marc.theaimsgroup.com/?l=jakarta-commons-user&m=114780510526564&w=2 > > > > --- Nestor Urquiza <[EMAIL PROTECTED]> wrote: > > > > > Hello guys, > > > I guess because of my own mistake at some point > I > > > used: > > > evts[0] = new TriggerEvent("app.error", > > > TriggerEvent.ERROR_EVENT, null); > > > to manage an error assumming that being the type > = > > > "ERROR" and having defined in my scxml: > > > <n:transition event="app.error" target="IDDLE"/> > > > the engine would go directly to IDDLE state and > no > > > other code would be executed from the scxml > file. > > > > > > Going thru the code I see that this is wrong > > > assumption since at least the last version of > > > SCXMLExecutor#triggerEvents() restores the rest > of > > > the > > > events data (restoreEventData(oldData);) and so > > the > > > scxml continues executing the uneeded code. > > > > > > My question is ... is this a feature > could/should > > be > > > implemented? This is, when firing an error > > > transition > > > just ignore the rest of the actions to be > executed > > > and > > > just execute the error event specified by > > > TriggerEvent() constructor. If this is not an > > > expected > > > behavior from the SCXML specification I guess I > > can > > > have the same behavior if I am able to call > > > SCXMLExecutor#setEventData() to just clear the > > > pending > > > events, however this method is private. > > > > > > Thanks as always for any help, > > > Nestor > > > > > > > __________________________________________________ > > > Do You Yahoo!? > > > Tired of spam? Yahoo! Mail has the best spam > > > protection around > > > http://mail.yahoo.com > > > > > > > > > --------------------------------------------------------------------- > > > To unsubscribe, e-mail: > > > [EMAIL PROTECTED] > > > For additional commands, e-mail: > > > [EMAIL PROTECTED] > > > > > > > > > > > > __________________________________________________ > > Do You Yahoo!? > > Tired of spam? Yahoo! Mail has the best spam > > protection around > > http://mail.yahoo.com > > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: > > [EMAIL PROTECTED] > > For additional commands, e-mail: > > [EMAIL PROTECTED] > > > > > > > __________________________________________________ > Do You Yahoo!? > Tired of spam? Yahoo! Mail has the best spam > protection around > http://mail.yahoo.com > > --------------------------------------------------------------------- > To unsubscribe, e-mail: > [EMAIL PROTECTED] > For additional commands, e-mail: > [EMAIL PROTECTED] > > __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]