error parameter for expect_snapshot()

[lionel-]

It might be useful to add an error = FALSE argument to expect_snapshot(). When TRUE, no error is a failure. When FALSE, error is a failure. NULL would correspond to the current behaviour (no failure either way). When non-null, presence or absence of error would be checked on CRAN even when cran = FALSE.

[lionel-]

Just a note that multiple errors can be flexibly and conveniently tested in this way:

cnd_cat <- function(x) {
  cat(paste0(conditionMessage(x), "\n"))
}

test_that("%|% fails with wrong types", {
  expect_snapshot({
    cnd_cat(expect_error(c(1L, NA) %|% 2))
    cnd_cat(expect_error(c(1, NA) %|% ""))
    cnd_cat(expect_error(c(1, NA) %|% call("fn")))
  })
})

This prints both the code and the error message in the snapshot file. expect_snapshot() fails if an expected error does not occur.

[lionel-]

Easier way suggested by @hadley:

test_that("%|% fails with wrong types", {
  expect_snapshot({
    (expect_error(c(1L, NA) %|% 2))
    (expect_error(c(1, NA) %|% ""))
    (expect_error(c(1, NA) %|% call("fn")))
  })
})

This prints the condition objects returned invisibly by expect_error(), including the error message.