[help] Define many targets from one source, without triggering re-run of all targets when source changes

[pdavidboll]

Help

• I understand and agree to https://books.ropensci.org/targets/help.html.

Description

Hi,

I am using targets to run a large simulation study, in which I need to first simulate a number of DGPs and then analyse the simulated data for each DGP. So far, I am defining the DGPs as a list in a separate .R file as follows:

dgp_list <- list(
    dgp_1A = function(inputargs, COEF) {
        x <- somefunc(inputargs)
        eps <- someotherfunc(inputargs)
        y <- x + eps
        data.frame(y = y, x = x, eps = eps)
      },
    dgp_1B = function(inputargs, COEF) {
        x <- somenewfunc(inputargs)
        eps <- someothernewfunc(inputargs)
        y <- x + eps
        data.frame(y = y, x = x, eps = eps)
      }
)

and I source this .R file at the top of the _targets.R file. Then, I use this list to build static targets via tar_map.

The problem is that whenever I change anything in one of the DGP functions, targets re-runs the whole pipeline. I understand why, but I would like to figure out a way for me define the DGPs in such a way that I can change one of the DGPs or add a new DGP to the list without triggering a re-run of all simulations for the unaffected DGPs.

Any help would be much appreciated.

[joelnitta]

Why are you putting the functions in a list? If you define each one separately, changes to one won't affect the other.

I suppose it might have something to do with static branching, but you could also consider dynamic branching.

[wlandau]

I agree with @joelnitta, I suggest defining the functions separately and not in a list or class.

For simulation studies, you may be interested in tarchetypes::tar_map_rep(), which my group and I use for simulation studies routinely.