Possible improvement of hcat

[bkamins]

Consider the following code:

 x = DataFrame([(i,j) for i in 1:3, j in 1:4])
hcat(x, [1,2,3])

works OK, but:

hcat([1,2,3], x)

fails.

Additionally:

hcat([1 2 3], x)

appends DataFrame to [1 2 3] in fourth position.

Is this intended, or should it be fixed.

[bkamins]

Also this is probably something we should avoid (x is a DataFrame):

hcat(x, "bad mistake")

which produces StackOverflow.

[nalimilan]

I don't think that's really intended, but it's difficult to fix in full generality since AFAIK there's no way to define a method which is called when at least one of the arguments is a DataFrame. See JuliaLang/julia#20815. I guess we could add methods to handle the two-argument case at least, since it's likely to be the most common one. PR welcome.

Regarding the stack overflow, it seems it could be fixed by adding a two-argument hcat! method to stop the recursive call of hcat!(a::DataFrame, b, c...).

[bkamins]

This is what I meant. I will make a PR (I was not so much into DataFrames development earlier so I do not know well what already has been considered and decided).

[bkamins]

closing as the PR is merged.