add nrow and ncol

[xiaodaigh]

No description provided.

[joshday]

I don't see a benefit to having this in Tables at the cost of adding more methods to the interface.

[pdeffebach]

Is the canonical recommendation to do this nrow = length(Tables.rows(t)) and ncol = length(Tables.columns(t))? if t = Tables.columns(df)?

[xiaodaigh]

length(Tables.rows(t))

Works because many iterators has the has length traits

[joshday]

To expand, "most" tables in practice have a length, but the Tables interface doesn't guarantee it. This is what I would do to get nrows and ncols from column/row iterators:

t = Tables.columns(df)

nrows = length(first(t))

ncols = length(t) 

t2 = Tables.rows(df)

nrows = length(t2)

ncols = length(first(t2))

[pdeffebach]

Thanks for the note.

Just to be clear, the iteration itself is determined by DataFrames (for example), so length may not be guaranteed either.

I was under the impression iteration was implemented by Tables.jl via some sort of

getproperty(t, s) for s in propertynames(t)

[pdeffebach]

Also your second example doesn't work for generic tables because iteration on rows isn't defined, even as a fallback. Ref #75

[joshday]

Just to be clear, the iteration itself is determined by DataFrames (for example), so length may not be guaranteed either.

Correct. A DataFrame probably has length defined, but other things may not. Tables doesn't need the length method in order to function, but most things probably have a length.

iteration on rows isn't defined

Iteration over a single row isn't guaranteed, but rows(t) returns an iterator. In both my examples, there's no guarantee that any of those length methods exist, but in practice they probably do.

[xiaodaigh]

https://docs.julialang.org/en/v1/manual/interfaces/

There is an IteratorSize() type checker and the default is HasLength() trait. Mostly, likely rows would return an iterator that has length

[ablaom]

You can get the number of columns from length(Tables.schema(X).names):

julia> using Tables

julia> X = (x1=rand(3), x2=rand(3)) 
(x1 = [0.893945, 0.696079, 0.544089],
 x2 = [0.162807, 0.0328988, 0.567906],)

julia> length(Tables.schema(X).names)
2

[quinnj]

I think there have been enough clarifications and suggestions here; but yeah, I dont' think we need anything in the interface for this.