Improve expect_setequal() feedback

[hadley]

Fixes #1657

[DavisVaughan]

How would you feel about framing this as Only in expected: instead of Missing from expected:?

I have trouble reading this example:

x <- c("a", "b")
y <- c("b", "c")

expect_snapshot_failure(expect_setequal(x, y))
#> `x` and `y` don't have the same values
#> * Missing from expected: "a"
#> * Missing from actual: "c"

and I think it is because my brain is juggling too many things. I have to realize that y is "expected" but then look over at x to see that it did indeed have "a" which isn't in y.

This makes more sense to me

x <- c("a", "b")
y <- c("b", "c")

expect_snapshot_failure(expect_setequal(x, y))
#> `x` (`actual`) and `y` (`expected`) don't have the same values
#> * Only in `actual`: "a"
#> * Only in `expected`: "c"

I also like seeing `x` (`actual`) because I can't ever remember which is which. It is also what expect_equal() and expect_identical() seem to do

> testthat::expect_identical(1, 1.5)
Error: 1 (`actual`) not identical to 1.5 (`expected`).

  `actual`: 1.0
`expected`: 1.5
> testthat::expect_equal(1, 1.5)
Error: 1 (`actual`) not equal to 1.5 (`expected`).

  `actual`: 1.0
`expected`: 1.5

[hadley]

I like it.

[DavisVaughan]

Can't you use expect_setequal() on more than strings though? How does this look on numeric values?

[hadley]

Oops, yes.