.names parameter for pack/unpack

[vorpalvorpal]

Currently pack/unpack have a .name_sep argument that packs/unpacks names as <Outer><sep><Inner>. I am working with a data set where the names are <Inner><sep><Outer>. There is no easy way of dealing with this set up. The .names parameter used by across deals with this easily by having a glue specification with {.col} being used as a mask. It seems like something similar would be simple to achieve for pack/unpack where .names would have masks for {.inner} and {outer}, perhaps with a default of "{.outer}.{.inner}".

If this is something you would consider, I can try to put together a PR.

[DavisVaughan]

Can you provide an actual reprex of your issue? This doesn't make much sense to me with how pack/unpack work.

[vorpalvorpal]

I have a water quality dataset I am working with that is very wide (> 100 cols). A simplified version that hopefully shows what I am getting at is below:

df <- tribble(~value, ~name.analyte, ~units.analyte, ~valence.analyte, 
              ~name.lab, ~method.lab, ~name.project, ~start_date.project) 
# plus a lot move columns of form `x.y`
# For certain purposes it is easier to pack related columns up so that we have fewer columns
df1 <-
  df |>
  pack(analyte = ends_with(".analyte"),
       lab = ends_with(".lab"),
       project = ends_with(".project"),
       .names = "{.inner}.{.outer}")
names(df1)
#> [1] "value"              "analyte"          "lab"          "project"  
names(df1$analyte)
#> [1] "name"       "units"      "valence"    
# And then unpack again
df1 |>
  unpack(cols = c(analyte, lab, project), names = "{.inner}.{outer}") |>
  names()
#> [1] "value"              "name.analyte"       "units.analyte"      "valence.analyte"    
#> [5] "name.lab"           "method.lab"         "name.project"       "start_date.project"