It's a unityper type https://github.com/YingboMa/Unityper.jl

https://github.com/JuliaSymbolics/SymbolicUtils.jl/blob/d68e092a646ee746becc16fa5ec0e7e066a96f12/src/types.jl#L25-L62

So you need to check the EXPR type https://github.com/JuliaSymbolics/SymbolicUtils.jl/blob/d68e092a646ee746becc16fa5ec0e7e066a96f12/src/types.jl#L252

This breaks Symbolics.derivative(2x, 2x) == 1. IDK if we care about that.

I am fine with having more errors, more safety.

This also breaks differentiating with respect to an array of symbols (which we should presumably fix) and with respect to x(t) which occurs once in tests.

The crux of this issue is that we assume two expressions are constant with respect to eachother unless proven otherwise.

julia> @variables x, y

julia> Symbolics.derivative(x, y)
0

This PR does not fix that issue, and does introduce errors which previously worked e.g. all of these results will become errors:

julia> @variables x(t), y(t)

julia> Symbolics.derivative(x, x) # Good
1

julia> Symbolics.derivative(t, x) # Wrong
0

julia> Symbolics.derivative(x, t) # Good
Differential(t)(x(t))