Takafumi Arakaki

I think I understand. You want to modify a set of relating options rather than the selected option, right? I think it is a great idea. I hope you to...

That's an interesting point. I think I agree. But a pain point is that there is no easy way to query if `iterate(table)` is row-wise or not. We can use...

> `append!(Table(a=[1], b=[2]), Tables.rows((a=[3], b=[4])))` should still work and it's efficient as it uses `append_columnaccess!` path This already works. But we need https://github.com/JuliaData/Tables.jl/pull/126 for the optimization (column-oriented memory access).

> still `vcat`, but column-by-column? Yes, `vcat` done column-by-column as an optimization. > Also note that: > > ```julia > Tables.istable(t) && Tables.rowaccess(t) && Tables.rows(t) === t > ``` >...

I thought it's nice to avoid allocating unnecessary objects. If performance is not a concern, I think the cleanest approach here would be to add things row-by-row as there is...

I initially had something like that. But then @andyferris pointed out that it might be better to strictly require that `src` in `append!(::TypedTable.Table, src)` to be (semantically) an iterator over...

Naively thinking, I thought `Table{NamedTuple{(), Tuple{}}}` would be equivalent eto `Vector{Union{}}`. But maybe it should just be `Table{Union{}}`? I figured out that ```julia Table{Union{}, 1, NamedTuple{(), Tuple{}}}(NamedTuple()) ``` works (although...

Oh, I forgot about `Vector{Union{}}(undef, 3)`. Thanks for pointing it out. > Internally `Set{T}` uses `Dict{T, Union{}}` and there a `Vector{Union{}}` pops up; not sure who/what else uses them in...

Reading a bit about relational algebra (BTW, thanks for the remark on joining with the empty relation), I'm more confident that I need `Table{Union{}}`, not `Table{NamedTuple{()}}` (at least within the...

I think I explained my implementation details too much. A simple question is: how do you write a function for materializing a `Table` from an arbitrary container (with possibly unknown...