System.AccessViolationException with Python Eval

[rsatapat]

Hi, I am new to Bonsai. I wanted to replicate one of my python analysis pipelines in Bonsai. The final step in this pipeline involves the scipy.signal.find_peaks() function. But, the workflow always crashes with this error trace

Unhandled Exception: System.AccessViolationException: Attempted to read or write protected memory. This is often an indication that other
memory is corrupt.
at Python.Runtime.Runtime.PyRun_String(String code, RunFlagType st, BorrowedReference globals, BorrowedReference locals)
at Python.Runtime.PyModule.Eval(String code, PyDict locals)
at System.Reactive.Linq.ObservableImpl.Select2._.OnNext(TSource value) at System.Reactive.AutoDetachObserver1.OnNextCore(T value)
at System.Reactive.ObserverBase1.OnNext(T value) at System.Reactive.Linq.ObservableImpl.Do1..OnNext(TSource value)
at System.Reactive.Linq.ObservableImpl.Select2._.OnNext(TSource value) at System.Reactive.Linq.ObservableImpl.Buffer1..OnNext(TSource value)
at System.Reactive.AutoDetachObserver1.OnNextCore(T value) at System.Reactive.ObserverBase1.OnNext(T value)
at System.Reactive.Linq.ObservableImpl.Do1._.OnNext(TSource value) at System.Reactive.Linq.ObservableImpl.Select2..OnNext(TSource value)
at System.Reactive.AutoDetachObserver1.OnNextCore(T value) at System.Reactive.ObserverBase1.OnNext(T value)
at System.Reactive.Linq.ObservableImpl.Do1._.OnNext(TSource value) at System.Reactive.Linq.ObservableImpl.Select2..OnNext(TSource value)
at System.Reactive.AutoDetachObserver1.OnNextCore(T value) at System.Reactive.ObserverBase1.OnNext(T value)
at System.Reactive.Linq.ObservableImpl.Do1._.OnNext(TSource value) at System.Reactive.Linq.ObservableImpl.Select2..OnNext(TSource value)
at System.Reactive.AutoDetachObserver1.OnNextCore(T value) at System.Reactive.ObserverBase1.OnNext(T value)
at System.Reactive.Linq.ObservableImpl.Do1._.OnNext(TSource value) at System.Reactive.Linq.ObservableImpl.Do1..OnNext(TSource value)
at System.Reactive.AutoDetachObserver1.OnNextCore(T value) at System.Reactive.ObserverBase1.OnNext(T value)
at System.Reactive.Linq.ObservableImpl.Do1._.OnNext(TSource value) at System.Reactive.Linq.ObservableImpl.Select2..OnNext(TSource value)
at System.Reactive.AutoDetachObserver1.OnNextCore(T value) at System.Reactive.ObserverBase1.OnNext(T value)
at System.Reactive.Linq.ObservableImpl.Do1._.OnNext(TSource value) at System.Reactive.Linq.ObservableImpl.Select2..OnNext(TSource value)
at System.Reactive.AutoDetachObserver1.OnNextCore(T value) at System.Reactive.ObserverBase1.OnNext(T value)
at System.Reactive.Linq.ObservableImpl.Do1._.OnNext(TSource value) at System.Reactive.AutoDetachObserver1.OnNextCore(T value)
at System.Reactive.ObserverBase`1.OnNext(T value)
at Bonsai.Vision.FileCapture.<>c__DisplayClass32_0.b__1()
at System.Threading.Tasks.Task.Execute()
at System.Threading.ExecutionContext.RunInternal(ExecutionContext executionContext, ContextCallback callback, Object state, Boolean preserveSyncCtx)
at System.Threading.ExecutionContext.Run(ExecutionContext executionContext, ContextCallback callback, Object state, Boolean preserveSyncCtx)
at System.Threading.Tasks.Task.ExecuteWithThreadLocal(Task& currentTaskSlot)
at System.Threading.Tasks.Task.ExecuteEntry(Boolean bPreventDoubleExecution)
at System.Threading.ExecutionContext.RunInternal(ExecutionContext executionContext, ContextCallback callback, Object state, Boolean preserveSyncCtx)
at System.Threading.ExecutionContext.Run(ExecutionContext executionContext, ContextCallback callback, Object state, Boolean preserveSyncCtx)
at System.Threading.ExecutionContext.Run(ExecutionContext executionContext, ContextCallback callback, Object state)
at System.Threading.ThreadHelper.ThreadStart(Object obj)

I am also attaching the workflow and .py files.
trial.zip
python_trial.py

[glopesdev]

@rsatapat it looks like you might have forgotten to capture the GIL with the ObserveOnGIL operator.

Whenever you have an observable sequence running asynchronously to Python you will need to use it (just once at the point you want to switch contexts, just before the Python call).

Be aware acquiring the GIL is a mutually exclusive region: it will likely affect performance if your camera is running too fast, and will also bottleneck if you are running multiple cameras at the same time.

In upcoming versions we will likely be able to configure Python contexts to run without the GIL, but this is not working yet.

[rsatapat]

Thank you so much! That solved the issue.
I have another related questions. I have defined a function inside script that use for CreateRuntime. I am trying to call this function but it raises errors. I have tried func(), func(it) etc. What is the issue here? Also, do I have the right model in my mind, is this how the Eval class is meant to be used?

[banchan86]

The follow-up question was posted to https://github.com/orgs/bonsai-rx/discussions/2533 and answered there; just adding a breadcrumb in case anyone stumbles on this thread.