Should expect_mapequal() be recursive?

[nealrichardson]

It seems only to make unordered comparisons at the top level. I would expect nested lists to get the same mapequal treatment.

testthat::expect_mapequal(
  list(
    outer_1 = 1,
    outer_2 = list(
      inner_1 = 1,
      inner_2 = 2
    )
  ),
  list(
    outer_2 = list(
      inner_2 = 2,
      inner_1 = 1
    ),
    outer_1 = 1
  )
)
#> Error: act$val[exp_nms] not equal to exp$val.
#> Component "outer_2": Names: 2 string mismatches
#> Component "outer_2": Component 1: Mean relative difference: 1
#> Component "outer_2": Component 2: Mean relative difference: 0.5

^{Created on 2021-12-31 by the reprex package (v2.0.1)}

[hadley]

Hmmm, I'll need to think about whether it's worth changing expect_mapequal() or just advertising list_as_map = TRUE more.

library(testthat)

test_that("can recursively ignore differences", {
  local_edition(3)
  expect_equal(
    list(
      outer_1 = 1,
      outer_2 = list(
        inner_1 = 1,
        inner_2 = 2
      )
    ),
    list(
      outer_2 = list(
        inner_2 = 2,
        inner_1 = 1
      ),
      outer_1 = 1
    ),
    list_as_map = TRUE
  )

})
#> Test passed 🌈

^{Created on 2022-01-02 by the reprex package (v2.0.1)}

[nealrichardson]

Oh cool, I didn't know about the list_as_map argument, that's nice. Though I'm surprised that expect_mapequal() behaves differently and isn't just expect_equal(list_as_map = TRUE).

[hadley]

Yeah. The main thought that’s required if it can just change that behaviour for 3e. It probably isn’t used very often so it’s likely to be possible but it’ll need a little research and to wait for a minor release.

[hadley]

Usage looks pretty minimal, so the change is likely to be fine: https://cs.github.com/?scopeName=All+repos&scope=&q=language%3Ar%20expect_mapequal%20-repo%3Ar-lib%2Ftestthat

[hadley]

FYI this broke a bunch of revdeps so I'm going to revert. It will no longer be recursive.