You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Tables.jl uses a lazy row type (basically a custom type that has the original table and the index of the row) instead of materializing the whole NamedTuple. I was wondering whether we should do the same here. It has the advantage that map(i -> i.x+i.y, t) would already be efficient (at least in the non distributed case) without having to explicitly having to type map(i -> i.x+i.y, t, select = (:x, :y)). We could then drop our collect_columns code and use the mechanism at Tables (see here).
The text was updated successfully, but these errors were encountered:
Tables.jl uses a lazy row type (basically a custom type that has the original table and the index of the row) instead of materializing the whole
NamedTuple
. I was wondering whether we should do the same here. It has the advantage thatmap(i -> i.x+i.y, t)
would already be efficient (at least in the non distributed case) without having to explicitly having to typemap(i -> i.x+i.y, t, select = (:x, :y))
. We could then drop ourcollect_columns
code and use the mechanism at Tables (see here).The text was updated successfully, but these errors were encountered: