Add isapprox support for dataframes.

[Sov-trotter]

Based on #2169
@bkamins For now I went on with the way Base.isapprox handles numeric values. And used broadcasting for column elements. Is this the right way to go?

[Sov-trotter]

In this latest commit I am using the isapprox method for individual columns defined https://github.com/JuliaLang/julia/blob/e58bc723ed7b1c10da28f55385c243f6df5499cf/stdlib/LinearAlgebra/src/generic.jl#L1631-L1642 here. One thing is that norm::Function=abs throws error since it's nowhere broadcasted in the isapprox for AbstractVector, in fact it's using the norm method by default, while in the commented out code(manual implementation), we were able to broadcast the abs function over to the elements.

[bkamins]

Also can you please update NEWS.md with an entry about this addition (it should be mentioned as new functionality). Thank you!

[Sov-trotter]

Yeah. Will do it. Thanks for helping out. :)

[Sov-trotter]

Should I add LinearAlgebra into the dependency changes in News.md?

[bkamins]

No, because it is a part of a standard library and in particular for LinearAlgebra you do not specify version number in [compat].

The idea of "Dependency changes" is to inform users of possible dependency conflicts, but in this case there is no such risk (everyone has LinearAlgebra installed anyway).

[bkamins]

Sorry - I have forgotten that the docstring is missing. Can you please add it before merging?

[Sov-trotter]

Oh. Really sorry, it skipped my mind.

[nalimilan]

Looks good apart from two minor things.

[nalimilan]

Ah, and it looks like you need to add LinearAlgebra to Project.toml to fix the CI failures. The line is:

LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"

[bkamins]

it looks like you need to add LinearAlgebra to Project.toml

It was added earlier - I wonder why it is now gone.

[Sov-trotter]

It must have gone, when I rebased the branch onto master. Sorry for that.

[bkamins]

Additionally functions.md in the documentation has to be updated. I think a new section Other is required for this function as it does not fit any section we have now.

[Sov-trotter]

Won't it be "equality"? Since I see isequal is also missing in functions.md?

[bkamins]

We do not have a docstring for isequal and == so that is why it is missing. Not sure if we should (they do what I guess people assume they do, in isapprox this is different as we compare columns)

[bkamins]

But we can add Equality section I think

[bkamins]

Thank you!