use of tibble::num() breaks display by other packages

[twest820]

Using num() to get better tibble display formatting has the ironic side effect of causing formatting paths in other packages to fail. Here's a trivial repex.

foo = tibble(a = tibble::num(seq(0, 10, by = 3.2), digits = 1))
table(foo$a)
Error in `as.character()`:
! Can't convert `x` <pillar_num:.1!> to <character>.
# workaround: table(as.numeric(foo$a))

It's difficult to inventory all the code paths which can call as.character() on a column using the num() constructor but, from the few places I've used num() with tibble 3.2.1 so far, impact seems likely widespread across at least base and ggplot.

This was raised in ggplot nearly two years ago (tidyverse/ggplot2#5284) but doesn't appear to have gotten forwarded to tibble or pillar as suggested there.

[krlmlr]

Similar problem:

x <- tibble::num(c(0.001, 0.005))

# Expected output
vctrs::vec_match(x, x)
#> [1] 1 2

# This bug
match(x, x)
#> Error in `as.character()`:
#> ! Can't convert `x` <pillar_num> to <character>.

^{Created on 2025-03-18 with reprex v2.1.1}

[moodymudskipper]

pillar::num() creates numerics with a class and attributes, the class inherits from "vctrs_vctr", which makes as.character() fail on such objects.

However as.character() is called by some important functions such as table().

We want to define, in the pillar package, a method as.character.pillar_num() that calls as.character.default() so we don't fall back to as.character.vctrs_vctr()