[WIP] work towards updating to DataFrames v0.11+ framework #1088
Conversation
i haven't been able to fix everything due to slight inconsistencies between PooledDataArrays and CategoricalArrays behavior. See JuliaData/CategoricalArrays.jl#117 for details. Working towards a fix for GiovineItalia#1065
|
thanks for tackling this. i don't use DataFrames much, so am not familiar with the API changes. please though make use of the regression testing infrastructure in Gadfly before you merge. |
|
I definitely don't plan on merging this until it's thoroughly vetted since it's a lot of major changes. |
| @@ -1,13 +1,13 @@ | |||
| #Is this usable data? | |||
| isconcrete{T<:Number}(x::T) = !isna(x) && isfinite(x) | |||
| isconcrete{T<:Number}(x::T) = !ismissing(x) && isfinite(x) | |||
There was a problem hiding this comment.
I guess it should be something like:
isconcrete(::Missing) = false
isconcrete(x::Number) = isfinite(x)There was a problem hiding this comment.
Good catch. This should fix #1083 too, I think
|
Cc: @andreasnoack |
|
Doh. I've been through almost all the same changes earlier this week. A good reminder to open a I had a lot of issues with the wrapping behavior of |
It would be great if you could push your changes to this PR or make a new PR, which ever is easier.
Agreed. And this was where I got hung up as well. |
|
Out of interest, JuliaArrays/IndirectArrays.jl#12 |
| PooledDataArray(newdata, newpool, [false for _ in newdata]) | ||
| end | ||
| cat_aes_var!{T}(xs::CategoricalArray{T}, ys::CategoricalArray{T}) = vcat(xs, ys) | ||
| cat_aes_var!{T, U}(xs::CategoricalArray{T}, ys::CategoricalArray{U}) = vcat(xs, ys) |
There was a problem hiding this comment.
BTW, cat_aes_var{T}(...) is pre-0.6 syntax. Since DataFrames 0.11 requires 0.6, it's better write it as cat_aes_var(...) where {T}, although in this particular case it could be as simple as (replaces both definitions)
cat_aes_var!(xs::CategoricalArray, ys::CategoricalArray) = vcat(xs, ys)
| if typeof(vals) <: PooledDataArray | ||
| setfield!(aes_grid[i, j], var, | ||
| make_pooled_data_array(typeof(vals), staging[i, j])) | ||
| if typeof(vals) <: CategoricalArray |
There was a problem hiding this comment.
vals isa CategoricalArray or isa(vals, CategoricalArray) is shorter
|
Closed in favor of #1090 |
i haven't been able to fix everything due to slight inconsistencies between PooledDataArrays and CategoricalArrays behavior. See JuliaData/CategoricalArrays.jl#117 for
details. Working towards a fix for #1065.
It would nice to have this soon as we're holding back the whole DataFrames ecosystem until we do. Maybe @alyst and @nalimilan could also provide us some feedback to get this working with v0.11