@combine and unique

[juliohm]

The following MWE is producing a different result than what I expected:

df = DataFrame(C = [:a,:a,:a,:b,:b,:b,:b], V = [1,2,3,4,5,6,7])

@chain df begin
  @groupby(:C)
  @combine(unique(:V))
end

7×2 DataFrame
 Row │ C       V_unique 
     │ Symbol  Int64    
─────┼──────────────────
   1 │ a              1
   2 │ a              2
   3 │ a              3
   4 │ b              4
   5 │ b              5
   6 │ b              6
   7 │ b              7

I expected a dataframe with two rows:

DataFrame(C = [:a,:b], V_unique = [[1,2,3],[4,5,6,7]])

2×2 DataFrame
 Row │ C       V_unique     
     │ Symbol  Array…       
─────┼──────────────────────
   1 │ a       [1, 2, 3]
   2 │ b       [4, 5, 6, 7]

[jkrumbiegel]

That is the normal behavior for combine, if you return an array per group they get concatenated into rows. You can use Ref around the expression to avoid that.

[juliohm]

But that seems inconsistent with other query frameworks. Query.jl will return the result with two rows as expected. I'll try with other packages to see what happens.

On Mon, Jun 28, 2021, 03:25 jkrumbiegel @.***> wrote:

That is the normal behavior for combine, if you return an array per group they get concatenated into rows. You can use Ref around the expression to avoid that.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/jkrumbiegel/DataFrameMacros.jl/issues/6#issuecomment-869396240, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZQW3JEGCAVRFO2KSFVGYTTVAIUTANCNFSM47M2LZNQ .

[jkrumbiegel]

Might be, but that's DataFrames.jl behavior. I'll close this

[juliohm]

I think it is not about being a DataFrames.jl behavior. This is more like a general split-apply-combine behavior? I thought that any combine should guarantee a number of rows equal to the number of classes in the groupby. In user code, if one cannot know the number of rows beforehand that becomes an issue. Makes sense?

[juliohm]

What I am trying to say is that Ref behavior is a more natural default than broadcasting the groupby column.

[jkrumbiegel]

https://dataframes.juliadata.org/stable/man/split_apply_combine/

combine: does not put restrictions on number of rows returned, the order of rows is specified by the order of groups in GroupedDataFrame; it is typically used to compute summary statistics by group;

[juliohm]

I noticed that R produces the same results so I guess Query.jl is the exception to the rule.

Em seg., 28 de jun. de 2021 às 07:44, jkrumbiegel @.***> escreveu:

https://dataframes.juliadata.org/stable/man/split_apply_combine/

combine: does not put restrictions on number of rows returned, the order of rows is specified by the order of groups in GroupedDataFrame; it is typically used to compute summary statistics by group;

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/jkrumbiegel/DataFrameMacros.jl/issues/6#issuecomment-869578150, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZQW3MZ63MXCAYDRECWWLTTVBHBNANCNFSM47M2LZNQ .