Name repair of `as_tibble.matrix()` with duplicate `V*` columns

[DavisVaughan]

This is a pathological case, but actually does come up in the wild, see tidymodels/workflows#48

When as_tibble() is given a matrix with duplicate names, I see two issues:

1. The warning message is about not having any column names at all, which doesn't seem right

2. It is possible for the result to still have duplicate names after the repair if the original column names were already V* where * is a numeric value.

library(tibble)

x <- matrix(1:2, ncol = 2, dimnames = list(NULL, c("V2", "V2")))

# lifecycle warning, but its an odd warning
as_tibble(x)
#> Warning: The `x` argument of `as_tibble.matrix()` must have column names if `.name_repair` is omitted as of tibble 2.0.0.
#> Using compatibility `.name_repair`.
#> This warning is displayed once every 8 hours.
#> Call `lifecycle::last_warnings()` to see where this warning was generated.
#> # A tibble: 1 x 2
#>      V2    V2
#>   <int> <int>
#> 1     1     2

# warning is silent now, and still no fix
as_tibble(x)
#> # A tibble: 1 x 2
#>      V2    V2
#>   <int> <int>
#> 1     1     2

I wonder if in the duplicated names case we should be using vec_as_names(names, repair = "unique") or something similar to that, rather than using this V* repair. The V* repair seems reasonable when there are no names at all, but when some exist I think they should be used and just repaired in the standard way with ...i

[krlmlr]

We still need to specify explicit name repair here, the default is NULL for as_tibble.matrix(), which leads to compatibility mode.

library(tibble)

x <- matrix(1:2, ncol = 2, dimnames = list(NULL, c("V2", "V2")))

as_tibble(x, .name_repair = "unique")
#> New names:
#> * V2 -> V2...1
#> * V2 -> V2...2
#> # A tibble: 1 x 2
#>   V2...1 V2...2
#>    <int>  <int>
#> 1      1      2

^{Created on 2020-07-06 by the reprex package (v0.3.0)}

The compatibility check for duplicate names was introduced in c74da6b, pretty sure it was due to some package expecting different behavior.

We should change this with the next major release.

[krlmlr]

Now we at least mention "unique" column names:

---

author: kirill
date: 2021-07-17
output: "reprex::reprex\_document"
title: dopey-nyala_reprex.R

library(tibble)

x <- matrix(1:2, ncol = 2, dimnames = list(NULL, c("V2", "V2")))

# lifecycle warning, but its an odd warning
as_tibble(x)
#> Warning: The `x` argument of `as_tibble.matrix()` must have unique column names if `.name_repair` is omitted as of tibble 2.0.0.
#> Using compatibility `.name_repair`.
#> # A tibble: 1 x 2
#>      V2    V2
#>   <int> <int>
#> 1     1     2

# warning is silent now, and still no fix
as_tibble(x)
#> # A tibble: 1 x 2
#>      V2    V2
#>   <int> <int>
#> 1     1     2

^{Created on 2021-07-17 by the reprex package (v2.0.0)}

[krlmlr]

To rephrase: The problem here is the second V2 column -- it's a duplicate and should be repaired, but uses the name V2 from the compatibility name repair. Postponing for tibble 4.0.0 .