New tib_custom() as a variant of tib_scalar() with different defaults?

[krlmlr]

For ingesting a datetime field, I use:

    tib_scalar("start", vctrs::new_datetime(), ptype_inner = character(), transform = ~ as.POSIXct(.x, tz = "UTC")),

Should we support

    tib_custom("start", ~ as.POSIXct(.x, tz = "UTC")),

with tib_custom() forwarding to tib_scalar() but with different defaults (in particular, ptype_inner = character() and transform required, perhaps ptype optional)?

as.POSIXct(character())
#> POSIXct of length 0

^{Created on 2023-05-26 by the reprex package (v2.0.1)}

[mgirlich]

I don't think we need tib_custom() for this. We could simply change tib_scalar() a little, e.g. adding

if (is_missing(ptype)) {
    if (is_missing(ptype_inner) || is_null(ptype_inner) || is_null(transform)) {
      cli_abort("Must provide {.arg ptype} or both of {.arg ptype_inner} and {.arg transform}.")
    }

    ptype <- transform(ptype_inner)
  }

But I also like the clarity of

tib_scalar(
  "start",
  vctrs::new_datetime(),
  ptype_inner = character(),
  transform = ~ as.POSIXct(.x, tz = "UTC")
)

It is a bit redundant but also a bit easier to understand what the outcome of this tib_scalar() is.

[krlmlr]

We don't need tib_custom(), it would be a handy helper with different defaults than tib_scalar() .

Do you really need is_missing() when you can always do is.null() ?