[feature:] move the SS3 converter function to r4ss?

[iantaylor-NOAA]

Describe the situation that led to the request and a solution

@kellijohnson-NOAA, what do you think about moving get_ss3_data out of this repo into r4ss? That would make it easier for SS3 users to find and streamline the process of getting more people to try out FIMS.

Alternative solutions

Add converter functions for SS3 and other models to FIMS or some new repo within NOAA-FIMS which could be structured as an R package, unlike case-studies.

Statistical validity

No response

Additional context

No response

[kellijohnson-NOAA]

I ❤️ the idea of moving it to {r4ss}!

[iantaylor-NOAA]

Great. I see that r4ss is already installed when rendering the case studies via https://github.com/NOAA-FIMS/case-studies/blob/main/content/setup.qmd#L36-L39, so it's just a matter of moving the file and adding r4ss:: in front of the function calls.

@kellijohnson-NOAA, do you think it's worth preserving the git history? The steps discussed in https://kokkonisd.github.io/2021/06/23/import-git-history seem complex and you, me, and @MOshima-PIFSC would continue to be listed as authors even if we lost the history of our individual commits.

[kellijohnson-NOAA]

I see VERY limited value in preserving the history.

[iantaylor-NOAA]

I have started work on this issue in the convert-data-to-fims branch of r4ss and the petrale-with-new-r4ss-converter branch of case-studies.

I couldn't help but make some changes along the way, including extracting the age-to-length-conversion from the SS3 output in the revised converter function in r4ss, and no longer removing one of the catch fleets from the petrale model.

Presumably because of some of those changes, I'm now getting some errors (pasted below for my future reference).

I'm on leave for the next 2 weeks but will happily continue work on this when I'm back on March 2.

r$> # Run the model without optimization to help ensure a viable model
    test_fit <- parameters |>
        FIMS::initialize_fims(data = data_4_model) |>
        FIMS::fit_fims(optimize = FALSE)
Error: lexical error: invalid string in json text.
           "objective_function_value": nan,     "max_gradient_componen
                     (right here) ------^                                                                                                      
                                                                                                          
r$> test_fit <- parameters |>
        FIMS::initialize_fims(data = data_4_model) |>
        FIMS::fit_fims(optimize = TRUE)
v Starting optimization ...
Error in nlminb(start = par, objective = fn, gradient = gr, control = control) : 
  NA/NaN gradient evaluation
In addition: Warning message:
In nlminb(start = par, objective = fn, gradient = gr, control = control) :
  NA/NaN function evaluation

[kellijohnson-NOAA]

Thank you Ian for putting this in r4ss. I think that move is great. I will use the branch of r4ss in renv until you let me know that the changes are in main.

[kellijohnson-NOAA]

@iantaylor-NOAA I am wondering about the timing of the converter getting into main. No big deal on the answer, I am just thinking about changing the setup.qmd script in main to reference the branch of r4ss instead of main.

[iantaylor-NOAA]

@kellijohnson-NOAA, I was hoping to have at least one functioning model that used the converter before merging, but still struggling with petrale. I can try updating the Opakapaka model or apply it to hake and see if that works.
But also if it's more convenient, I can just bring it to main now since nobody is using the function and no harm will come.

[kellijohnson-NOAA]

Nope, it is not a big deal at all for it to be in a branch. I can try it out on Pacific Hake in a bit as well and help with some development if needed.