You're suspicion is spot on, passing the app object to the worker will not work (as doing so sends a copy, or in this case, an invalid copy). Instead, you need to make your testExecutionWorker static and pass in arguments directly instead of the app:
classdef ExampleWorkflowAbortAsyn < matlab.apps.AppBase
% Static is important here as you want this to be callable without the app object.
methods (Access = public, Static)
function testExecutionWorker(dataQueue, n, myIn1, myIn2, ..)
for i = 1:n
% Do stuff then
send(dataQueue, someResult);
end
end
end
% In the start button callback:
methods (Access = private)
function EnterTestState(app, event)
switch app.StartButton.Text
case 'Start'
..
myIn1 = app.SomeProperty;
myIn2 = app.SomeOtherProperty;
n = app.N
dataQueue = parallel.pool.PollableDataQueue;
% Have to use the classname here instead of app as this
% function is static.
fut = parfeval(@ExampleWorkflowAbortAsyn.testExecutionWorker, 0, dataQueue, n, myIn1, myIn2, ..);
% If for any reason testExecutionWorker fails, we only
% find out by checking fut. One way to deal with this
% is to check fut periodically with ~isempty(fut.Error)
% Another way is to do the following, which will add a
% callback once fut finishes:
afterAll(fut, @app.checkIfFutureFailed, 0, PassFuture=true);
..
% Rest of code is same
end
end
function checkIfFutureFailed(app, fut)
if ~isempty(fut.Error)
errordlg("Something went wrong!", fut.Error.getReport());
end
end
end
end
Several other tips along the way:
- For your use-case here, if you're using a recent version of MATLAB (I think R2021b or later), you can use backgroundPool instead of parpool("Threads", 1). It's functionally the same, except it is available in base MATLAB and doesn't require Parallel Computing Toolbox (just do "parfeval(backgroundPool,..)" in your code).
- You might want to look into afterEach / afterAll and parallel.pool.DataQueue. The current code (waiting on DataQueue/poll(inf)) will have periods of time where the app does not respond to button presses. By adding callbacks to when data messages are received (instead of waiting for those data messages directly), you allow other callbacks like button presses to still work. A full example of this is Create Responsive Apps by Running Calculations in the Background.
