combine() error with std() on GroupedDataFrame with Unitful columns

[tomyun]

using DataFrames, Statistics, Unitful
df = DataFrame(a=[:a, :a, :b, :b], b=[0, 1, 2, 3]u"m")

julia> combine(df, :b => std)
1×1 DataFrame
 Row │ b_std
     │ Quantity…
─────┼───────────
   1 │ 1.29099 m

julia> combine(groupby(df, :a), :b => std)
ERROR: DimensionError: 0.0 m and 0.25 m² are not dimensionally compatible.
Stacktrace:
 [1] +(::Quantity{Float64,𝐋,Unitful.FreeUnits{(m,),𝐋,nothing}}, ::Quantity{Float64,𝐋²,Unitful.FreeUnits{(m²,),𝐋²,nothing}}) at /Users/tomyun/.julia/packages/Unitful/1t88N/src/quantities.jl:137
 [2] groupreduce!(::Array{Quantity{Float64,𝐋,Unitful.FreeUnits{(m,),𝐋,nothing}},1}, ::DataFrames.var"#535#538"{Array{Quantity{Float64,𝐋,Unitful.FreeUnits{(m,),𝐋,nothing}},1}}, ::typeof(+), ::Nothing, ::Function, ::Bool, ::Array{Quantity{Int64,𝐋,Unitful.FreeUnits{(m,),𝐋,nothing}},1}, ::GroupedDataFrame{DataFrame}) at /Users/tomyun/.julia/packages/DataFrames/yqToF/src/groupeddataframe/fastaggregates.jl:174
 [3] (::DataFrames.Aggregate{typeof(var),Nothing})(::Array{Quantity{Int64,𝐋,Unitful.FreeUnits{(m,),𝐋,nothing}},1}, ::GroupedDataFrame{DataFrame}) at /Users/tomyun/.julia/packages/DataFrames/yqToF/src/groupeddataframe/fastaggregates.jl:248
 [4] (::DataFrames.Aggregate{typeof(std),Nothing})(::Array{Quantity{Int64,𝐋,Unitful.FreeUnits{(m,),𝐋,nothing}},1}, ::GroupedDataFrame{DataFrame}) at /Users/tomyun/.julia/packages/DataFrames/yqToF/src/groupeddataframe/fastaggregates.jl:254
 [5] _combine_process_agg(::Pair{Int64,var"#s151"} where var"#s151"<:(Pair{var"#s150",Symbol} where var"#s150"<:Function), ::Bool, ::DataFrame, ::GroupedDataFrame{DataFrame}, ::Dict{Symbol,Tuple{Bool,Int64}}, ::Array{DataFrames.TransformationResult,1}, ::Array{Int64,1}) at /Users/tomyun/.julia/packages/DataFrames/yqToF/src/groupeddataframe/splitapplycombine.jl:203
 [6] _combine(::GroupedDataFrame{DataFrame}, ::Array{Any,1}, ::Array{Bool,1}, ::Bool, ::Bool, ::Bool) at /Users/tomyun/.julia/packages/DataFrames/yqToF/src/groupeddataframe/splitapplycombine.jl:532
 [7] _combine_prepare(::GroupedDataFrame{DataFrame}, ::Union{Regex, AbstractString, Function, Signed, Symbol, Unsigned, Pair, AbstractArray{T,1} where T, Type, All, Between, Cols, InvertedIndex}; keepkeys::Bool, ungroup::Bool, copycols::Bool, keeprows::Bool, renamecols::Bool) at /Users/tomyun/.julia/packages/DataFrames/yqToF/src/groupeddataframe/splitapplycombine.jl:66
 [8] #combine#509 at /Users/tomyun/.julia/packages/DataFrames/yqToF/src/groupeddataframe/splitapplycombine.jl:638 [inlined]
 [9] combine(::GroupedDataFrame{DataFrame}, ::Pair{Symbol,typeof(std)}) at /Users/tomyun/.julia/packages/DataFrames/yqToF/src/groupeddataframe/splitapplycombine.jl:638
 [10] top-level scope at REPL[16]:1

Tested on Julia 1.5.3 with DataFrames 0.22.2 and Unitful 1.5.0.

[bkamins]

@nalimilan - could you please have a look at it. It seems to be the problem with the fast path that you have implemented (if you do not have time I can check it - please let me know then).

[nalimilan]

Good catch. This kind of thing is hard to get right until somebody actually tries it. I'll push a fix tomorrow.

[nalimilan]

If you need a workaround in the meantime, note that combine(..., :b => (x -> std(x))) avoids the optimized method that fails.

[nalimilan]

See #2601.