mtk_to_si no longer defined/exported with new functionality.

[TorkelE]

Simple MWE:

using StructuralIdentifiability
mtk_to_si # ERROR: UndefVarError: `mtk_to_si` not defined
using ModelingToolkit
mtk_to_si # ERROR: UndefVarError: `mtk_to_si` not defined

I think this was caused by the latest update.

Currently, this is breaking tests in Catalyst, hence I discovered it: https://github.com/SciML/Catalyst.jl/actions/runs/7743939220/job/21116510775?pr=776

I checked the source code. I am not really sure, but when I make extensions I do not put export statements in the extension file. Instead, I create a mtk_to_si function (without methods) in the main file, and then add methods to that in the extension. E.g. in Catalyst's HomotopyContinuation extension I have this in Catalyst.jl

# HomotopyContinuation
function hc_steady_states end
export hc_steady_states

and this in homotopy_continuation_extension.jl

function Catalyst.hc_steady_states(rs::ReactionSystem, ps; filter_negative=true, neg_thres=-1e-20, u0=[], kwargs...)
    ss_poly = steady_state_polynomial(rs, ps, u0)
    sols = HC.real_solutions(HC.solve(ss_poly; kwargs...))
    reorder_sols!(sols, ss_poly, rs)
    return (filter_negative ? filter_negative_f(sols; neg_thres) : sols)
end

This might be worth a try.

[pogudingleb]

@TorkelE
Thanks for reporting the issue and suggesting an elegant solution!
It did not occur to me that I can still have the function definition, I will do this right away.

[TorkelE]

I actually just finished a PR for it, see: #280
(but feel welcome to just do your own solution)

I encountered another problem there which might need checking as well.

Yes, I struggled quite a bit when I wrote the extensions. They work, but it is not super intuitive, and there are no good descriptions of how to do it.

[TorkelE]

If you haven't already, could you tag a new release so that Catalyst could use it for the tests?

[pogudingleb]

Sure, this is not version 0.5.3

[TorkelE]

Thanks :)