Loosen constructor for a DataFrame #1108
Conversation
It seems to be a little more difficult than it should be to pass a container of columns to the `DataFrame`'s constructor. This change lets you, e.g.
```julia
DataFrame([[1,2,3]], [:a])
```
Before this would complain because `[[1,2,3]]` is of type `Vector{Vector{Int}}` which is not a subtype of (concrete) type `Vector{Any}` (type parameters aren't convariant). This will automatically loosen the type parameter for the column container, if necessary.
| @@ -102,9 +102,9 @@ function DataFrame(; kwargs...) | |||
| return result | |||
| end | |||
|
|
|||
| function DataFrame(columns::Vector{Any}, | |||
| function DataFrame{T}(columns::Vector{T}, | |||
There was a problem hiding this comment.
You don't need the T AFAICT. Also, while you're at it, change this to AbstractVector.
|
Why not, though this triggers an ambiguity error that should be fixed. Could you add a test too? |
|
Thanks. Still needs a test though. |
|
OK. I'm done now. Sorry, I've only had a few seconds here and there. |
| @@ -185,6 +185,7 @@ module TestDataFrame | |||
| @test typeof(df[:,:a]) == NullableVector{Int} | |||
| @test typeof(df[:,:b]) == NullableVector{Char} | |||
|
|
|||
| @test DataFrame([[1,2,3], [2.5,4.5,6.5]], [:A, :B]) = DataFrame(A = [1,2,3], B = [2.5,4.5,6.5]) | |||
There was a problem hiding this comment.
This is causing syntax errors:
FAILED: dataframe.jl
LoadError: syntax: "[[1,2,3],[2.5,4.5,6.5]]" is not a valid function argument name
in macro expansion; at C:\Users\appveyor\.julia\v0.5\DataFrames\test\runtests.jl:43 [inlined]
in anonymous at .\<missing>:?
in include_from_node1 at .\loading.jl:488
in process_options at .\client.jl:262
in _start at .\client.jl:318
while loading C:\Users\appveyor\.julia\v0.5\DataFrames\test\dataframe.jl, in expression starting on line 188
There was a problem hiding this comment.
Wow, that is a really strange error message. The problem was I used = instead of ==
Sorry, like I said I only have a few moments here and there so I used the github web interface and relying on CI for testing.
|
Hmm now EDIT: the culprit is |
`==` between `Nullable{T}` and `T` is always false, for some reason.
|
Thanks! |
It seems to be a little more difficult than it should be to pass a container of columns to the
DataFrame's constructor. This change lets you, e.g.Before this would complain because
[[1,2,3]]is of typeVector{Vector{Int}}which is not a subtype of (concrete) typeVector{Any}(type parameters aren't convariant). This will automatically loosen the type parameter for the column container, if necessary.