SQL interface to DataFrame?

[tk3369]

Has anyone thought about taking SQL as query interface for data frames?

For example, doing something like this:

julia> df = DataFrame(x = rand(1:3, 10), y = rand(10))
10×2 DataFrame
 Row │ x      y         
     │ Int64  Float64   
─────┼──────────────────
   1 │     2  0.589515
   2 │     1  0.72848
   3 │     2  0.344321
   4 │     1  0.900515
   5 │     2  0.667395
   6 │     2  0.0749464
   7 │     2  0.916652
   8 │     1  0.460028
   9 │     3  0.284116
  10 │     3  0.662666

julia> sql("select x, count(*) from df group by 1", (:df => df))

[bkamins]

This could be doable and I was thinking about it some time ago. Someone would need to write parser of SQL. The point is that I think most people nowadays started using DuckDB interface for writing SQL queries in Julia (but I might be wrong here).

[drizk1]

SQLdf.jl exists - I haven't tried it yet.

A parser that converted df.jl to sql would be amazing, and would lead me to rewrite TidierDB. I've toyed with writing a tables.jl to sql interface, but there's some nuance to sql that has made it tricky.

I think as far as writing sql to run on df.jl, I wonder what the benefit of that vs registering the table in duckdb and using their existing engine?