Tracking user-defined functions that do not create targets

[arnold-c]

Prework

• Read and agree to the code of conduct and contributing guidelines.
• If there is already a relevant issue, whether open or closed, comment on the existing thread instead of posting a new issue.
• For any problems you identify, post a minimal reproducible example so the maintainer can troubleshoot. A reproducible example is:
  • Runnable: post enough R code and data so any onlooker can create the error on their own computer.
  • Minimal: reduce runtime wherever possible and remove complicated details that are irrelevant to the issue at hand.
  • Readable: format your code according to the tidyverse style guide.

Question

I'm hoping to try and track changes to user-defined functions do not create targets, but are used in a notebook that is part of a pipeline. For example, I have a quick plotting function that I use repeatedly, but I do not want to have to create a target for every single plot in a notebook as this would clutter the _target.R file. However, if I made a change to that function, even if it is sourced into the notebook, the pipeline thinks it is up to date.

I believe this is different to issues #241 and #239 as these relate to packages, not standalone functions.

One thought I had was to write these functions in a separate R file that I can track using format = "file" like below, and this seems to work, but I just wanted to make sure this is a reasonable thing to do and wouldn't lead to unintended consequences.

In _targets.R:

list(
    ...
    tar_target(
        user_functions,
        here::here("funs", "user-defined-functions.R")
        ),
        format = "file",
    ...
    tar_render(analysis_notebook, "analysis-notebook.Rmd")
    )

In my user-defined-functions.R file:

library(tidyverse)

df <- tibble(thresholds = 0:10, sensitivity = seq(0, 1, 0.1), specificity = rev(seq(0, 1, 0.1)))

my_function <- function(data){
  data %>%
    pivot_longer(cols = sensitivity:specificity, names_to = "metric", values_to = "value") %>%
    ggplot(aes(x = thresholds, y = value, color = metric)) +
    geom_point()
}

threshold_metric_plot(data = df)

In my Rmd notebook:

tar_load(user_functions)

my_function(data = df)

Created on 2021-01-06 by the reprex package (v0.3.0)

[wlandau]

One thought I had was to write these functions in a separate R file that I can track using format = "file" like below, and this seems to work, but I just wanted to make sure this is a reasonable thing to do and wouldn't lead to unintended consequences.

Sounds like a great workaround to me.

Another approach is to define a target for all the function objects:

tar_target(functions, list(my_function, fun2, fun3, ...))

This technique is less trigger-happy than format = "file" because it does not invalidate the target if there are trivial changes to comments and whitespace. For a function in memory, targets cannot tell the difference between this:

my_function <- function(data){
  data %>%
    pivot_longer(cols = sensitivity:specificity, names_to = "metric", values_to = "value") %>%
    ggplot(aes(x = thresholds, y = value, color = metric)) +
    geom_point()
}

and this:

my_function <- function(data){
  data %>%

    pivot_longer(
      cols = sensitivity:specificity,
      names_to = "metric",
      values_to = "value"
    ) %>%

    ggplot(aes(x = thresholds, y = value, color = metric)) +

    geom_point()
}