|
|
@@ -157,3 +157,94 @@ Cell
|
|
|
----
|
|
|
|
|
|
The cell tracer is discussed separately, as it has very specific behaviour and is only applicable to a select number of models.
|
|
|
+
|
|
|
+Custom
|
|
|
+------
|
|
|
+
|
|
|
+Additionally, it might necessary to define your own custom tracer.
|
|
|
+This can be done by defining a class similar to the following template.
|
|
|
+For each trace method, an *aDEVS* parameter is passed, being a reference to the atomic DEVS model doing the transition.
|
|
|
+On this *aDEVS* object, the following functions and attributes can be accessed:
|
|
|
+
|
|
|
+ - *aDEVS.getModelFullName()*: full hierarchical name of the model
|
|
|
+ - *aDEVS.IPorts*: reference to all input ports
|
|
|
+ - *aDEVS.OPorts*: reference to all output ports
|
|
|
+ - *aDEVS.state*: state of model
|
|
|
+ - *aDEVS.time_last[0]*: Time of next transition
|
|
|
+ - *aDEVS.time_next[0]*: Time of next transition
|
|
|
+ - *aDEVS.my_output*: dictionary of output events
|
|
|
+ - *aDEVS.my_input*: dictionary of input events
|
|
|
+ - *aDEVS.elapsed*: elapsed time before transition
|
|
|
+
|
|
|
+A custom tracer can be defined as follows::
|
|
|
+
|
|
|
+ class TracerCustom(object):
|
|
|
+ def __init__(self, uid, server, filename):
|
|
|
+ """
|
|
|
+ Both uid and server can be ignored, as these are only required for distributed simulation
|
|
|
+ filename contains the name of the file in which we should write the trace
|
|
|
+ """
|
|
|
+ pass
|
|
|
+
|
|
|
+ def startTracer(self, recover):
|
|
|
+ """
|
|
|
+ Recover is a boolean representing whether or not this is a recovered call (e.g., should the file be overwritten or appended to?)
|
|
|
+ """
|
|
|
+ pass
|
|
|
+
|
|
|
+ def stopTracer(self):
|
|
|
+ """
|
|
|
+ Stops the tracer (e.g., flush the file)
|
|
|
+ """
|
|
|
+ pass
|
|
|
+
|
|
|
+ def traceInternal(self, aDEVS):
|
|
|
+ """
|
|
|
+ Called for each atomic DEVS model that does an internal transition.
|
|
|
+ """
|
|
|
+ pass
|
|
|
+
|
|
|
+ def traceExternal(self, aDEVS):
|
|
|
+ """
|
|
|
+ Called for each atomic DEVS model that does an external transition.
|
|
|
+ """
|
|
|
+ pass
|
|
|
+
|
|
|
+ def traceConfluent(self, aDEVS):
|
|
|
+ """
|
|
|
+ Called for each atomic DEVS model that does a confluent transition.
|
|
|
+ """
|
|
|
+ pass
|
|
|
+
|
|
|
+ def traceInit(self, aDEVS, t):
|
|
|
+ """
|
|
|
+ Called upon initialization of a model.
|
|
|
+ The parameter *t* contains the time at which the model commences (likely 0).
|
|
|
+ """
|
|
|
+ pass
|
|
|
+
|
|
|
+ def traceUser(self, time, aDEVS, variable, value):
|
|
|
+ """
|
|
|
+ Called upon so called *god events* during debuggin, where a user manually alters the state of an atomic DEVS instance.
|
|
|
+ """
|
|
|
+ pass
|
|
|
+
|
|
|
+For some "example" tracers, have a look at the built-in tracers of PythonPDEVS, which can be found in *src/tracers*.
|
|
|
+
|
|
|
+Note that in optimistic synchronization the destructive parts of this operation should be separated.
|
|
|
+This can be done using the *runTraceAtController* function::
|
|
|
+
|
|
|
+ runTraceAtController(server, uid, aDEVS, [time, trace_text])
|
|
|
+
|
|
|
+Both the *server* and *uid* are those passed to the constructor of the tracer.
|
|
|
+
|
|
|
+Finally, after the tracer is defined, it needs to be registered for the simulator to use it.
|
|
|
+This is done using the following call on the instantiated simulator::
|
|
|
+
|
|
|
+ sim.setCustomTracer(self, tracerfile, tracerclass, args):
|
|
|
+
|
|
|
+Where:
|
|
|
+
|
|
|
+ - *tracerfile*: the Python class containing the implementation of the tracer, which is dynamically imported.
|
|
|
+ - *tracerclass*: the name of the class implementing the tracing functionality.
|
|
|
+ - *args*: the list of arguments that must additionally be passed to the tracer (e.g., filename)
|
