State events generated inside algorithm sections

[casella]

Consider the following MWE, courtesy of @AnHeuermann (OpenModelica/OpenModelica#7484):

model algorithmTest
  Real x;
algorithm 
  x := sin(time);
  if x < 0 then  // <-- relation 1
    x := 0;
  end if;
  x := cos(time);
  if x < 0 then  // <-- relation 2
    x := 0;
  end if;
  x := pi;
end algorithmTest;

Is this supposed to generate events every when relation 1 changes, when relation 2 changes, when any of the two change, or never?

I read Section 8.5 which doesn't explicitly mention event generation inside algorithms, though the last example features a (time) event generating expression inside an algorithm section. Section 11.1.2 says

An algorithm section is treated as an atomic vector-equation, which is sorted together with all other
equations. For the sorting process (BLT), every algorithm section with N different left-hand side variables,
is treated as an atomic N-dimensional vector-equation containing all variables appearing in the algorithm
section. This guarantees that all N equations end up in an algebraic loop and the statements of the
algorithm section remain together.

but it doesn't explain how events should be handled.

Is there any consensus on when events should be generated in the reported MWE?

[gkurzbach]

At least SimulationX generates events for both relations.

[henrikt-ma]

System Modeler also keeps track of both relations. To me it seems obvious that it needs to work this way even if the specification fails to say it clearly.

Unlike some of the other related problems we have been discussing lately, I don't see a reason to introduce general restrictions in the specification for not allowing event-generating expressions in algorithms (the use of a relation in the MWE is so basic that I would expect that only a general ban on event generation could make it illegal).

[HansOlsson]

Also two separate events for Dymola. I added an integer to see it.

model algorithmTest
  Real x;
  Integer i;
algorithm 
  i:=0;
  x := sin(time);
  if x < 0 then  // <-- relation 1
    x := 0;
    i:=1;
  end if;
  x := cos(time);
  if x < 0 then  // <-- relation 2
    x := 0;
    i:=i+2;
  end if;
  x := 3.14159;
  annotation (experiment(StopTime=10));
end algorithmTest;

[casella]

OK. Could it be worth mentioning explicitly in a comment in the language spec?

[HansOlsson]

@casella Possibly, and to be clearer we could in https://specification.modelica.org/master/equations.html#events-and-synchronization after

"The integration is halted and an event occurs whenever an event generation expression, e.g., x > 2 or floor(x), changes its value. An event generating expression has an internal buffer, and the value of the expression can only be changed at event instants. "

add something like:

"In a non-function algorithm each expression is compared to the previous value when executing the corresponding statement, where each iteration of statements inside for-loops are seen as different statements."

The main issues are algorithms with for-loops (needs restrictions in some way), or while-loops (I don't see how that would work), cp. the restrictions in https://specification.modelica.org/master/operators-and-expressions.html#discrete-time-expressions

[HansOlsson]

Language group: Add example.