should `.by` be optional when spreading ?

[moodymudskipper]

We could display the explicit call, a bit like in dplyr joins when by is not explicit.

Desired behavior:

df <- data.frame(x = c("a", "b"), y = c(3, 4), z = c(5, 6))
df
#>   x y z
#> 1 a 3 5
#> 2 b 4 6
df %tb>% .[`{x}` = y]
#> In `.['{x}' = y]` using remaining column(s) as group(s), you might want to be explicit and use `.['{x}' = y, .by = "z"]`
#>   z  a  b
#> 1 5  3 NA
#> 2 6 NA  4

^{Created on 2020-01-27 by the reprex package (v0.3.0)}

[moodymudskipper]

I'm not sure we should support this, it looks a lot like a mutate call so it's confusing, maybe by = NA could mean by is whatever columns were unmentioned ? we could leverage this for summarizing too.

[moodymudskipper]

implemented but not very robust at the moment as it doesn't handle the dot notation (if dot used on the right, we should find the names on the lhs).

[moodymudskipper]

Now the spreaded column is taken out thanks to https://github.com/moodymudskipper/tb/issues/36

This is not quite enough though, we should handle the "summarize_at" notation too.

[moodymudskipper]

now implemented, though not very economic as we recompute and reparse things that we went through before.