map(f,x::DNDSparse) produces wrong parametric type if f changes value type

[gkappler]

A = ndsparse((numid=1:10,), (a=1:10,b=1:10); chunks=2)
B = map(v->v.a+1,A)

seems to work, but assigns wrong type parameter typeof(B)

JuliaDB.DNDSparse{NamedTuple{(:numid,),Tuple{Int64}},NamedTuple{(:a, :b),Tuple{Int64,Int64}}}

which makes

C = filter(isodd, B)

fail with

ERROR: column a not found.

[gkappler]

I investigated a bit and think the minor fix in pull request #316 suffices.

typeof(B) in the example is now

JuliaDB.DNDSparse{NamedTuple{(:numid,),Tuple{Int64}},Int64}