Polars

In memory DataFrames in Rust

This is my mock up of DataFrames implemented in Rust, using Apache Arrow as backend.

WIP

Series

• cast
• take by index/ boolean mask
• limit
• Rust iterators!
• append
• aggregation: min, max, sum
• arithmetic
• comparison
• find
• sorting

DataFrame

• take by index/ boolean mask
• limit
• join: inner, left
• column ops: drop, select, rename
• group by
• concat (horizontal)
• read csv
• write csv
• write json
• read json
• sorting

Data types

• null
• boolean
• u32
• i32
• i64
• f32
• f64
• utf-8
• date
• time

Example

use polars::prelude::*;

// Create first df.
let s0 = Series::init("days", [0, 1, 2, 3, 4].as_ref());
let s1 = Series::init("temp", [22.1, 19.9, 7., 2., 3.].as_ref());
let temp = DataFrame::new_from_columns(vec![s0, s1]).unwrap();

// Create second df.
let s0 = Series::init("days", [1, 2].as_ref());
let s1 = Series::init("rain", [0.1, 0.2].as_ref());
let rain = DataFrame::new_from_columns(vec![s0, s1]).unwrap();

// Left join on days column.
let joined = temp.left_join(&rain, "days", "days");
println!("{}", joined.unwrap())

           days           temp           rain
            i32            f64            f64
            ---            ---            ---

              0           22.1           null
              1           19.9            0.1
              2              7            0.2
              3              2           null
              4              3           nul