search

MyRobotLab / InMoov

inmoov repo
http://myrobotlab.org
87 stars 64 forks source link

Possibility to stop a running gesture #165

Closed moz4r closed 5 years ago

moz4r commented 6 years ago

gestures pushed inside a thread ? any other "easy" idea ?

supertick commented 6 years ago

Its a multi-level rabbit hole ... First level is Python/Jython - interrupting python Second level (potentially) is Arduino - interrupting movement

So lets go with first level first .. Please first tell how/where in the Python or InMoov script is the gestures started ?

You can't know ending without the beginning ...

kwatters commented 6 years ago

Agreed, this is a VERY deep rabbit hole. We need a map before we go into it. Problems is that we require the python service to drive the gestures. If we have defined callbacks for when a gesture starts and is complete.. this would be a good start. (we will have to account for long running gestures...) Then when we have a status for the robot , if it's performing a gesture or not.. we can accept, queue, or reject a new gesture request coming in.

As GroG points out, this is a big task... Let's tackle it together.

supertick commented 6 years ago

Historically I think I tried to re-initialize the interpreter - that was bad idea. Looking and testing I can see we create a new Java/Python Thread PIThread for "most" new scripts that are executed ..

The executing Java/Python thread will block when executing a script

There is one "global" reference to the last thread created, for the last script run (all other references are lost).

"If" you have a reference to the PIThread calling interrupt is effectively doing a Ctrl-C signal to the script.

My proposal would be to maintain PIThread references in a map and when stop() is called - iterate through all alive threads and interrupt them - then clear the map. Currently there is a stop() function but it (partially) destroys the interpreter

This solution would just make stop() work correctly

it offers no solution for scripts that are setup to block (but this seems to be an edge case)

The map will fill during normal operation of dead thread references - but the overhead seems nominal to the desired effect.

supertick commented 6 years ago

please test - if its to satisfaction lets close it

moz4r commented 6 years ago

You are just magic ! Welcome to the new stopGesture()

image

Possibility to catch the error ? by exemple a log.info(Python process : KeyboardInterrupt)

KeyboardInterrupt: interrupted sleep

at org.python.modules.time.Time.sleep(Time.java:469)
at sun.reflect.GeneratedMethodAccessor45.invoke(Unknown Source)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(Unknown Source)
at java.lang.reflect.Method.invoke(Unknown Source)
at org.python.core.PyReflectedFunction.__call__(PyReflectedFunction.java:188)
at org.python.core.PyReflectedFunction.__call__(PyReflectedFunction.java:206)
at org.python.core.PyObject.__call__(PyObject.java:480)
at org.python.core.PyObject.__call__(PyObject.java:484)
at org.python.pycode._pyx116.loop$1(<string>:6)
at org.python.pycode._pyx116.call_function(<string>)
at org.python.core.PyTableCode.call(PyTableCode.java:171)
at org.python.core.PyBaseCode.call(PyBaseCode.java:125)
at org.python.core.PyFunction.__call__(PyFunction.java:403)
at org.python.pycode._pyx193.f$0(<string>:1)
at org.python.pycode._pyx193.call_function(<string>)
at org.python.core.PyTableCode.call(PyTableCode.java:171)
at org.python.core.PyCode.call(PyCode.java:18)
at org.python.core.Py.runCode(Py.java:1614)
at org.python.core.Py.exec(Py.java:1658)
at org.python.util.PythonInterpreter.exec(PythonInterpreter.java:267)
at org.myrobotlab.service.Python$PIThread.run(Python.java:159)