Support automatic test case naming about patrick HOT 3 CLOSED

deparse() comes with its own headaches (though I do think it's probably the right way to go) -- a minimal-overhead approach is to just use seq_along to number the tests. At least that way the test failures will have unique names.

from patrick.

I think a good use case to have in mind here is replacing a simple for loop with a patrick test.

With for loop:

for (name in names) {
  test_that(sprintf("works: %s", name), {
    # expectations depending on 'name'
  })
}

replaced with

with_parameters_test_that(
  "works:",
  {
    # expectations depending on 'name'
  },
  name = names
)

A big appeal of the for loop is that names can be a literal & we don't need to store any variables in the test env, nor to repeat ourselves by spelling out names >once.

Whereas the appeal of doing this through patrick is (1) API consistency within a test suite (2) makes it clearer the purpose is to do a templated series of tests (slightly less clear in for loop approach, especially with a bigger {...} body) and (3) obviates the distracting sprintf() test name construction part.

from patrick.

I think this is the same as a new issue I was about to post:

The more I use patrick() the more I find myself writing code like:

with_parameters_test_that(
  "a test name",
  { some_test_exprs(single_param) },
  .test_name = FOO,
  single_param = FOO
)

This feels pretty redundant. It seems like using .test_name = params[[1]] when length(params) == 1 is a reasonable default.

Alternatively, accepting syntax like

single_param = FOO, .test_name = single_param (or even "single_param" to avoid NSE headaches)

would generalize to higher-cardinality tests where .test_name is still drawn directly from one of the test parameters

from patrick.