Weak types in Tables.jl conversion

[juliohm]

using XLSX
using DataFrames

XLSX.writetable("foo.xlsx", "TABLE" => DataFrame(A=rand(3), B=["foo","bar","baz"]))

xlsx   = XLSX.readxlsx("foo.xlsx")
XLSX.eachtablerow(xlsx["TABLE"]) |> DataFrame

3×2 DataFrame
 Row │ A         B   
     │ Any       Any 
─────┼───────────────
   1 │ 0.753115  foo
   2 │ 0.447895  bar
   3 │ 0.236192  baz

[juliohm]

ping @felipenoris

[felipenoris]

Hi @juliohm !
Life has been difficult lately. We need another beer. 🍺

[TimG1964]

At present, infer_eltypes is invoked only after the table iteration has completed as part of gettable. This way, the inference can be made on entire columns. To do it as part of the iterator seems to imply a two-pass process.

In conversation with Claude, it seems like this is achievable at low cost, but it needs the gettable process to be restructured to do the inference earlier (ie by the iterator itself).

I'm going to mark this as an enhancement which I may explore "later".