Type inference for Columns API

[ezmiller]

This is a discussion thread relating to automatic type inference issues as it relates the new Column API that I'm working with. Right now this question issue relates to our "constructor" function for columns:

(tablecloth.column.api/column [1 2 3])

Currently, this function uses tech.v3.dataset.column/new-column for column creation. @genmeblog pointed out:

generateme: Hi. In the past (I don't know if it's valid currently) there was a difference in data type inference when different column creation methods were used. (ds/->dataset {:a any_seq}) and (ds/new-column :a any_seq) could have different data types for :a column. Worth to check how it behaves now.

I took a look at the two functions and didn't see a difference right now:

  (-> {:a [1 2 3 4]}
      (ds/->dataset)
      :a
      (dtype/elemwise-datatype));; => :int64

  (->> [1 2 3 4]
       (ds/new-column)
       (dtype/elemwise-datatype));; => :int64


  (-> {:a [1.0 2.0 3.0 4.0]}
      (ds/->dataset)
      :a
      (dtype/elemwise-datatype));; => :float64

  (->> [1.0 2.0 3.0 4.0]
       (ds/new-column)
       (dtype/elemwise-datatype));; => :float64

  (-> {:a ["a" "b" "c"]}
      ds/->dataset
      :a
      (dtype/elemwise-datatype));; => :string

  (->> ["a" "b" "c"]
       (ds/new-column)
       (dtype/elemwise-datatype));; => :string

@genmeblog are these the kind of calls where we used to see different type inference results?