Use `tar_map()` to reduce duplicated code

[Robinlovelace]

From the targets manual:

Sometimes, a pipeline contains more targets than a user can comfortably type by hand. For projects with hundreds of targets, branching can make the _targets.R file more concise and easier to read and maintain.

Just saw this tip from @andrewheiss (thanks for sharing):

Finally figured out how to dynamically make {targets} targets through metaprogramming (https://books.ropensci.org/targets/static.html#metaprogramming) and it's SO NICE to have nice clean target names that are automatically generated #rstats

Source: https://fosstodon.org/@andrew@fediscience.org/111689215513772257

Currently our code has a lot of bits like this:

npt/_targets.R

Lines 322 to 328 in f1ae9eb

	tar_target(r_commute_fastest, {
	rs_commute_fastest[[1]]$routes
	}),
	tar_target(r_commute_quietest, {
	rs_commute_quietest[[1]]$routes
	}),

An iterator like tar_map() could compress those 3 targets into a single function call, reducing complexity and easing readability and maintenance.

[Robinlovelace]

I tested this on a modified version from the documentation:

  list(
    tarchetypes::tar_map(
      list(a = c(12, 34, 1), b = c(45, 78, 9)),
      targets::tar_target(x, a + b),
      targets::tar_target(y, x + a, pattern = map(x))
    )
  )

And the results show that the 'target factory' is accurate:

This can be part of a wider refactor discussed, will do some tests first but I think we can greatly simplify things as per #343 but taking it to the next level.

[Robinlovelace]

Taking a look at this now.

[Robinlovelace]

Heads-up @mem48 and @wangzhao0217 these are the regions I'm planning to use to geographically batch the build process, enabling all routes to be held in a single list:

[Robinlovelace]