New failure in 'adaptalint'

[MichaelChirico]

Haven't had time to explore yet, but there's a new failure in the 'adaptalint' test suite [v3.0.2 vs current HEAD]:

── Error (test-basics.R:29:3): extract works in simple case ────────────────────
<tibble_error_column_names_cannot_be_empty/tibble_error/rlang_error/error/condition>
Error in `as_tibble(.)`: Columns 1, 2, 3, 4, 5, and 18 more must be named.
Use `.name_repair` to specify repair.
Caused by error in `repaired_names()`:
! Names can't be empty.
✖ Empty names found at locations 1, 2, 3, 4, 5, etc.

It's a simple test of adaptalint::extract_style(); that function looks like:

filename %>%
  lint() %>%
  as_tibble() %>%
  mutate(total_lints = n()) %>%
  group_by(.data$linter) %>%
  summarise(count = n(),
            total_lints = mean(.data$total_lints)
  ) %>%
  mutate(adjusted = .data$count/.data$total_lints)

The first thing that comes to mind is the new S3 dispatch... we might need to add an as_tibble.lints method, perhaps?

[MichaelChirico]

Yep... here's the MRE:

tibble::as_tibble(lintr::lint(text="a=1", linters=lintr::assignment_linter()))

I think we get to use an Enhances dependency, so that's fun :) Scratch that, we already Suggest tibble :(

[AshesITR]

Why doesn't tibble:::as_tibble.default get dispatched and return a good tibble using lintr:::as.data.frame.lints?

> tibble::as_tibble(lintr::lint(text="a=1", linters=lintr::assignment_linter()))
# A tibble: 1 × 7
  filename line_number column_number type  message                        line  linter           
  <chr>          <dbl>         <dbl> <chr> <chr>                          <chr> <chr>            
1 <text>             1             2 style Use <-, not =, for assignment. a=1   assignment_linter
> getRversion()
[1] ‘4.3.1’

[MichaelChirico]

that's what used to happen, now as_tibble.list intercedes

[MichaelChirico]

Maybe there's a tibble version consideration here too?

[AshesITR]

Ah that makes sense. Can we export an S3 method without importing tibble?