Side effects of using keyword allowduplicates in the unstack function

Specific Domains Data
1

If I have this df

df1 = DataFrame(
           field = repeat(["A", "B", "C", "D", "E"], 12),
           data = rand(60),
           group = repeat(1:4, outer = 15))

if i try to do the following i get the suggestion to use allowduplicates

julia> unstack(df1,:group,:data)
ERROR: ArgumentError: Duplicate entries in unstack at row 21 for key ("A",) and variable 1. Pass allowduplicates=true to allow them.

following the suggestion I get the following result.

julia> unstack(df1,:field,:data,allowduplicates=true)
4×6 DataFrame
 Row │ group  A         B         C         D         E        
     │ Int64  Float64?  Float64?  Float64?  Float64?  Float64? 
─────┼─────────────────────────────────────────────────────────
   1 │     1  0.375523  0.341876  0.360489  0.925469  0.593533
   2 │     2  0.310565  0.21974   0.121332  0.730402  0.787124
   3 │     3  0.37437   0.164914  0.350874  0.857969  0.291186
   4 │     4  0.856412  0.905269  0.132294  0.168782  0.232174

which would be equivalent to the following code, but done explicitly.


cgl=combine(groupby(df1,[:group,:field]),:data=>last)

ucgl=unstack(cgl, :field,:data_last)

I am sure that the matter has been dealt with, but I have not found where and how.
I wonder if in these cases it is not useful to add a function that does aggregation, instead of kw = allowduplicates.

2

This could be discussed. The issue with arbitrary aggregation is that it might not have an eltype that matches eltype of the column.