fix: Throw better warning when a guess is missing

[vyudu]

Fix #3449

[ChrisRackauckas]

only one?

[vyudu]

only one what?

[SebastianM-C]

I think the message is a bit confusing since one could interpret it as asking for more than one guess for a single variable.

[vyudu]

Ah okay, will change

[ChrisRackauckas]

Won't this have a list of variables that need guesses? Not just one?

[vyudu]

Assuming test pass I think this is good

[ChrisRackauckas]

If it's just missing, what's the val?

[vyudu]

The symbolic value that it tries to assign to the variable

[ChrisRackauckas]

That's if it's cyclic. But this also is for if a guess is just missing, right? In which case, what does the error look like?

[vyudu]

If a guess it just missing it throws a different error (IncompleteInitializationError)

[vyudu]

ModelingToolkit.jl/src/systems/diffeqs/abstractodesystem.jl

Lines 1345 to 1353 in 0ad773d

	const INCOMPLETE_INITIALIZATION_MESSAGE = """
	Initialization incomplete. Not all of the state variables of the
	DAE system can be determined by the initialization. Missing
	variables:
	"""
	struct IncompleteInitializationError <: Exception
	uninit::Any
	end

[ChrisRackauckas]

That error doesn't make much sense to the user though? Why wouldn't the case I mentioned be a missing guess error?

[vyudu]

It probably should be, this was just already in the code. I can make it throw a MissingGuessError instead of this one.

[vyudu]

there is also this error but I need to dig a bit to figure out when it throws IncompleteInitializationError vs this one, maybe @AayushSabharwal knows better

ModelingToolkit.jl/src/variables.jl

Lines 210 to 214 in 0ad773d

	const MISSING_VARIABLES_MESSAGE = """
	Initial condition underdefined. Some are missing from the variable map.
	Please provide a default (`u0`), initialization equation, or guess
	for the following variables:
	"""

[ChrisRackauckas]

    @parameters g
    @variables x(t) y(t) [state_priority = 10] λ(t)
    eqs = [D(D(x)) ~ λ * x
           D(D(y)) ~ λ * y - g
           x^2 + y^2 ~ 1]
    @mtkbuild pend = ODESystem(eqs, t)

    @test_throws ModelingToolkit.MissingGuessError ODEProblem(pend, [x => 1], (0, 1), [g => 1], guesses = [y => λ])

where it's missing a guess for λ, what does this print?

[vyudu]

ERROR: Unable to resolve numeric guesses for all of the variables in the system. This may be because your guesses are cyclic. In order for the problem to be initialized, all of the variables must have a numeric value to serve as a starting point for the nonlinear solve.

Symbolic values were found for the following variables/parameters in the map; please provide additional numeric guesses so they can resolve to numbers:

y(t) => 50 + λ(t)
λ(t) => 2551 + y(t)

[ChrisRackauckas]

y(t) => 50 + λ(t)
λ(t) => 2551 + y(t)

how does it get that? That doesn't look right.

What does it print for:

    @parameters g
    @variables x(t) y(t) [state_priority = 10] λ(t)
    eqs = [D(D(x)) ~ λ * x
           D(D(y)) ~ λ * y - g
           x^2 + y^2 ~ 1]
    @mtkbuild pend = ODESystem(eqs, t)

    ODEProblem(pend, [x => 1], (0, 1), [g => 1])

?

[vyudu]

ERROR: Initial condition underdefined. Some are missing from the variable map.
Please provide a default (u0), initialization equation, or guess
for the following variables:

λ(t), y(t)

[vyudu]

Sorry, the first case you posted actually prints this, my output was from the test I wrote

  Initial condition underdefined. Some are missing from the variable map.
  Please provide a default (`u0`), initialization equation, or guess
  for the following variables:

  λ(t)

[ChrisRackauckas]

Okay so if any are missing you get that message, and if it's cyclic you get this other method. Given that, can we refine the error message? We know it's not a missing guess and instead a cyclic case, so we should make this more explicit.