Add more informative feedback in expect_named()

[EmilHvitfeldt]

Hope to close #2091 and close #1558

This PR aims to add more informative errors.

library(testthat)

x <- setNames(nm = 1:100)

expect_named(x, as.character(c(1:56, 58:100)), ignore.order = TRUE)
#> Error: Names of `x` (`actual`) and as.character(c(1:56, 58:100)) (`expected`) don't have the same values.
#> * Only in `actual`: "57"

^{Created on 2025-07-28 with reprex v2.1.1}

[EmilHvitfeldt]

This expect() block is copied and modified slightly from expect_setequal() to include "Names of ".

[hadley]

Could you explore extracting out a single helper function that we could use here, in expect_setequal() and expect_mapequal()? A expectation helper should end in _ and have an argument trace_env = caller_env() that is passed to every use of fail().

[EmilHvitfeldt]

@hadley for the ignore.cases = FALSE (default case) I was thinking that the waldo::compare() that powers expect_identical() would provide a high value reporting of the issues here. If you agree would I need to include if (edition_get() >= 3) {?

[hadley]

Can you hold off a day or two until I've finished refactoring how we write expectations? I'll have a better set of recommendations then.

[hadley]

Ok, can you please take another stab at it, following the updated custom-expectations vignette? That will also help me test if I've documented it adequately 😄

[EmilHvitfeldt]

The addition of error_prefix here and in expect_setequal_() are done to avoid duplication of these function by allowing me to add "Names of ", otherwise I would need some gnarly $lab modifications