df-cols are hidden before any other columns, which feels wrong

[DavisVaughan]

Consider the following df with a df-col up front

col <- tibble::tibble(foo = 1, bar = 2, baz = 3)
df <- tibble::tibble(col = col, x = 1, y = 2, z = 3, a = 4, b = 5, c = 6)

It's pretty confusing to me that, in a narrow console, baz is dropped first. I would have expected the final c column to be dropped first

It gets particularly hard to explain when you make it even more narrow, where c is dropped and some, but not all, of the df-col columns are hidden.

I have a feeling the current rule is something like:

Hide all but the first df-col columns first, then starting hiding non-df-col columns from last to first until it fits

But I think a simpler rule overall would just be:

Hide columns from last to first until it fits

So in this case you'd just hide c, then b, then a, etc, until it fits.

[krlmlr]

Thanks. The display shows as many columns as possible. More columns of a packed column are considered a "detail" that will be truncated first (similarly to an overly long character column), in favor of showing more columns. Does this explain the behavior?

In the proposed behavior, a deeply packed column would be expanded, hiding all other columns. Is this what we want?

I don't have a strong opinion; both options have advantages and could coexist in the same codebase, IIRC.

[DavisVaughan]

More columns of a packed column are considered a "detail" that will be truncated first (similarly to an overly long character column), in favor of showing more columns

Yea that does explain the behavior and is what I guessed was happening, I think I just find it unintuitive. Packed or not, I think I want to read the column names from left to right and only have columns at the far right be ones that get dropped off.

---

I guess I see truncation of the character column here as being fine

library(tibble)

tibble(
  a = 1,
  b = paste0(rep("x", 1000), collapse = ""),
  c = 1
)
#> # A tibble: 1 × 3
#>       a b                                                                      c
#>   <dbl> <chr>                                                              <dbl>
#> 1     1 xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx…     1

Because I still see all the column names a, b, c and only the "data" inside column b is truncated.

This just feels worse to me

col <- tibble::tibble(foo = 1, bar = 2)
df <- tibble::tibble(col = col, x = 1, y = 2, z = 3, a = 4, b = 5, c = 6)
df

I see how you could argue that col$bar is "data" inside column col as well, but I think the fact that we write out # ℹ 2 more variables: col$bar <dbl>, c <dbl> is proof that we are treating col$bar more like a column than as "data" within a column, so the normal column truncation rules of "drop the right columns first" should apply.