Tables.columns seem to flatten columns

[aviks]

julia> db = SQLite.DB()
SQLite.DB(":memory:")

julia> nt = (a=[1,2,3],b=[2,3,4])
(a = [1, 2, 3], b = [2, 3, 4])

julia> SQLite.load!(nt, db, "mytest")
"mytest"

julia> result  = DBInterface.execute(db, "select * from mytest")
SQLite.Query(SQLite.Stmt(SQLite.DB(":memory:"), 2), Base.RefValue{Int32}(100), [:a, :b], Type[Union{Missing, Int64}, Union{Missing, Int64}], Dict(:a => 1, :b => 2), Base.RefValue{Int64}(0))

julia>  [Tables.columns(result)...;;]
6-element Vector{Int64}:
 1
 2
 3
 2
 3
 4

This was surprising since this type of code seems to work for other types of Tables compatible objects

julia> [Tables.columns(nt)...;;]
3×2 Matrix{Int64}:
 1  2
 2  3

Reports on slack stated that this works on DataFrame and Arrow objects.

Note that Tables.matrix does work successfully on SQLite resultsets

julia> Tables.matrix(result)
3×2 Matrix{Int64}:
 1  2
 2  3
 3  4

Is this a misunderstanding of how Row based Tables should behave? In which case, should this kind of code fail silently? Or is this a bug in the Tables.columns implementation in this package?

[aviks]

OK, this was confusing due to the fact that apparently repeated semicolons are a new feature in 1.7, and the code above was executed party in 1.6, and partly in 1.7 .