possible across update to support functions w > 1 arg

[drizk1]

Did some experimenting to add support for multiple function args in an across function and with relatively limited changes to parse_across managed to get it to work. suffix corresponds to function order in across. However, I fully expect there is a different way to do this under the hood

julia> using TidierData, TidierFiles

julia> mtcars = read_file( "https://gist.githubusercontent.com/seankross/a412dfbd88b3db70b74b/raw/5f23f993cd87c283ce766e7ac6b329ee7cc2e1d1/mtcars.csv");

julia> wt_mean(x,y) = x' * y / sum(y);

julia> wt_mean2(x,y,n) = x' * y / sum(y) + n;

julia> @chain mtcars begin
         @group_by(cyl)
         @summarize(across((!(wt, cyl, model)), (x -> wt_mean(x, wt))))
       end
3×10 DataFrame
 Row │ cyl    mpg_1    disp_1   hp_1      drat_1   qsec_1   vs_1      am_1      gear_1   carb_1  
     │ Int64  Float64  Float64  Float64   Float64  Float64  Float64   Float64   Float64  Float64 
─────┼───────────────────────────────────────────────────────────────────────────────────────────
   1 │     6  19.6458  185.239  121.558   3.56917  18.122   0.621219  0.378781  3.82104  3.33616
   2 │     4  25.935   110.35    83.3945  4.03126  19.3804  0.914887  0.649803  4.04725  1.57225
   3 │     8  14.8064  362.125  209.373   3.20566  16.8853  0.0       0.120381  3.24076  3.53955

julia> @chain mtcars begin
         @group_by(cyl)
         @summarize(across((!(wt, cyl, model)), (x -> wt_mean(x, wt), x -> wt_mean2(x, wt, 5), minimum)))
       end
3×28 DataFrame
 Row │ cyl    mpg_1    mpg_2    mpg_3    disp_1   disp_2   disp_3   hp_1      hp_2      hp_3   drat_1   drat_2   d ⋯
     │ Int64  Float64  Float64  Float64  Float64  Float64  Float64  Float64   Float64   Int64  Float64  Float64  F ⋯
─────┼──────────────────────────────────────────────────────────────────────────────────────────────────────────────
   1 │     6  19.6458  24.6458     17.8  185.239  190.239    145.0  121.558   126.558     105  3.56917  8.56917    ⋯
   2 │     4  25.935   30.935      21.4  110.35   115.35      71.1   83.3945   88.3945     52  4.03126  9.03126
   3 │     8  14.8064  19.8064     10.4  362.125  367.125    275.8  209.373   214.373     150  3.20566  8.20566
                                                                                                  16 columns omitted

> mtcars %>%
+   group_by(cyl) %>% 
+   summarize(across(-wt, ~weighted.mean(., wt)))
# A tibble: 3 × 11
    cyl   mpg  disp    hp  drat  qsec    vs    am  gear  carb weight
  <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>  <dbl>
1     4  25.9  110.  83.4  4.03  19.4 0.915 0.650  4.05  1.57  0.483
2     6  19.6  185. 122.   3.57  18.1 0.621 0.379  3.82  3.34  0.429
3     8  14.8  362. 209.   3.21  16.9 0     0.120  3.24  3.54  0.512

edit: that failing nightly test has to do w column order getting changed in an unnest. not sure if this is something that needs to be treated as a bug, or resolved with a @relocate. in a dif PR i just resovled w a @relocate

[kdpsingh]

Awesome @drizk1, I will take a look at this and the other PRs.