symbolic_solve fails for time-dependent variables

[adhalanay]

As the title suggests symbolic_solve fails with an error if one tries to solve an algebraic equations involving time-dependent variables. An MWE is :
@variables t S(t) R(t)
expr=[S+R-2,2*S-3*R-6]
symbolic_solve(expr,[S,R])

which fails with

KeyError: key R not found

Stacktrace:
[1] getindex
@ ./iddict.jl:98 [inlined]
[2] linearity_1
@ ~/.julia/packages/Symbolics/rtkf9/src/linearity.jl:57 [inlined]
[3] mark_vars(expr::SymbolicUtils.BasicSymbolic{Real}, vars::OrderedCollections.OrderedSet{Vector{Num}})
@ Symbolics ~/.julia/packages/Symbolics/rtkf9/src/semipoly.jl:208
[4] (::Base.Fix2{typeof(Symbolics.mark_vars), OrderedCollections.OrderedSet{Vector{Num}}})(y::SymbolicUtils.BasicSymbolic{Real})
@ Base ./operators.jl:1144
[5] iterate
@ ./generator.jl:48 [inlined]
[6] collect_to!(dest::Vector{Symbolics.SemiMonomial}, itr::Base.Generator{Vector{Any}, Base.Fix2{typeof(Symbolics.mark_vars), OrderedCollections.OrderedSet{Vector{Num}}}}, offs::Int64, st::Int64)
@ Base ./array.jl:838
[7] collect_to_with_first!
@ ./array.jl:816 [inlined]
[8] _collect(c::Vector{Any}, itr::Base.Generator{Vector{Any}, Base.Fix2{typeof(Symbolics.mark_vars), OrderedCollections.OrderedSet{Vector{Num}}}}, ::Base.EltypeUnknown, isz::Base.HasShape{1})
@ Base ./array.jl:810
[9] collect_similar
@ ./array.jl:709 [inlined]
[10] map
@ ./abstractarray.jl:3371 [inlined]
[11] mark_vars(expr::SymbolicUtils.BasicSymbolic{Real}, vars::OrderedCollections.OrderedSet{Vector{Num}})
@ Symbolics ~/.julia/packages/Symbolics/rtkf9/src/semipoly.jl:204
[12] mark_and_exponentiate(expr::SymbolicUtils.BasicSymbolic{Real}, vars::OrderedCollections.OrderedSet{Vector{Num}})
@ Symbolics ~/.julia/packages/Symbolics/rtkf9/src/semipoly.jl:144
[13] semipolyform_terms(expr::SymbolicUtils.BasicSymbolic{Real}, vars::OrderedCollections.OrderedSet{Vector{Num}})
@ Symbolics ~/.julia/packages/Symbolics/rtkf9/src/semipoly.jl:160

[14] semipolynomial_form(expr::SymbolicUtils.BasicSymbolic{Real}, vars::Vector{Vector{Num}}, degree::Float64; consts::Bool)
@ Symbolics ~/.julia/packages/Symbolics/rtkf9/src/semipoly.jl:264
[15] semipolynomial_form
@ ~/.julia/packages/Symbolics/rtkf9/src/semipoly.jl:257 [inlined]
[16] polynomial_coeffs(expr::SymbolicUtils.BasicSymbolic{Real}, vars::Vector{Vector{Num}})
@ Symbolics ~/.julia/packages/Symbolics/rtkf9/src/semipoly.jl:301
[17] check_poly_inunivar(poly::Num, var::Vector{Num})
@ Symbolics ~/.julia/packages/Symbolics/rtkf9/src/solver/solve_helpers.jl:97
[18] symbolic_solve(expr::Vector{Num}, x::Vector{Num}; dropmultiplicity::Bool, warns::Bool)
@ Symbolics ~/.julia/packages/Symbolics/rtkf9/src/solver/main.jl:168
[19] symbolic_solve(expr::Vector{Num}, x::Vector{Num})
@ Symbolics ~/.julia/packages/Symbolics/rtkf9/src/solver/main.jl:145
[20] top-level scope
@ In[18]:3

[ChrisRackauckas]

@n0rbed can you take a look? This is a very important case to handle and the matching seems like it's taking too many liberties with the dependence

[n0rbed]

Hi @ChrisRackauckas , the error stems from lines 262-264 in ./src/solver/preprocess.jl
Here is the problem:

julia> some_expr = unwrap(1 + S)
1 + S(t)

julia> args = Symbolics.arguments(some_expr)
2-element Vector{Any}:
 1
  S(t)

julia> new_expr = Symbolics.term(+, args...)
1 + S(t)

julia> polynomial_coeffs(new_expr, S)
ERROR: MethodError: no method matching iterate(::SymbolicUtils.BasicSymbolic{Real})

julia> polynomial_coeffs(wrap(new_expr), S)
ERROR: MethodError: no method matching iterate(::SymbolicUtils.BasicSymbolic{Real})

There was one small mistake causing the "key R not Found" but thats easily fixed, this is the root problem. Let me know if you have any clues, not sure why this "reconstruction" of the poly is breaking stuff.

[n0rbed]

This reconstruction above can be removed actually. And i have a branch that does so, last problem is now #1323 ; I'll make a PR after that's fixed