README.md

DataToolkit

Pure Rust crate allowing the manipulation of indexed data structures like timeseries:

extern crate datatoolkit;
extern crate chrono;

use datatoolkit::{DataPoint,TimeSeries};
use chrono::{Utc, TimeZone};

let dps = vec![ 
    DataPoint::new(Utc.ymd(2008, 1, 1).and_hms(0, 0, 0), 122),
    DataPoint::new(Utc.ymd(2008, 1, 1).and_hms(0, 1, 0), 120),
    DataPoint::new(Utc.ymd(2008, 1, 1).and_hms(0, 2, 0), 118),
    DataPoint::new(Utc.ymd(2008, 1, 1).and_hms(0, 3, 0), 114),
    DataPoint::new(Utc.ymd(2008, 1, 1).and_hms(0, 5, 0), 116),
    DataPoint::new(Utc.ymd(2008, 1, 1).and_hms(0, 4, 0), 117)
];
Series::from_vec( "Test", dps )

// Get method
assert_eq!( ts.at(&Utc.ymd(2008, 1, 1).and_hms(0, 2, 0), 0).unwrap().get(), &118 );
assert_eq!( ts.at(&Utc.ymd(2008, 1, 1).and_hms(0, 2, 0), -1).unwrap().get(), &120 );
assert_eq!( ts.at(&Utc.ymd(2008, 1, 1).and_hms(0, 3, 0), 1).unwrap().get(), &117 );
assert_eq!( ts.at(&Utc.ymd(2008, 1, 1).and_hms(0, 6, 0), 0), None );
// Latest range method
let res = ts.range(-3, -1);
assert_eq!( res[0].get(), &114 );
assert_eq!( res[1].get(), &117 );
assert_eq!( res[2].get(), &116 );
// Range method
let res = ts.range_at(&Utc.ymd(2008, 1, 1).and_hms(0, 5, 0), 2, -2);
assert_eq!( res[0].get(), &114 );
assert_eq!( res[1].get(), &117 );
// Index 
assert_eq!( ts[-1].get(), &116 ); // Last element
assert_eq!( ts[0].get(), &122 ); // First element

Similarly to Pandas in Python, it also handles data from multiple types thanks to flexible data structures like FlexTable. This example uses the 1920 season data from the English League 2 division from football-data.co.uk:

// Pandas Equivalent:
// df = pd.read_csv('./tests/E3.csv')
// df = df[["Div","Date","Time","HomeTeam","AwayTeam","FTHG","FTAG","B365H","B365D","B365A"]]

let headers = vec!["Div","Date","Time","HomeTeam","AwayTeam","FTHG","FTAG","B365H","B365D","B365A"];
let datatypes = vec![
    FlexDataType::Str,
    FlexDataType::Str,
    FlexDataType::Str,
    FlexDataType::Str,
    FlexDataType::Str,
    FlexDataType::Uint,
    FlexDataType::Uint,
    FlexDataType::Dbl,
    FlexDataType::Dbl,
    FlexDataType::Dbl
];
let table = FlexTable::from_csv("./tests/E3.csv", headers, datatypes);

All data missing or not fitting the type requirements are assigned a type of FlexDataType:NA. Here are some examples on generating new series using series in the FlexTable.

// All games where one team scored more than 3 goals
    // Pandas equivalent: df.where((df['FTHG'] > 3) | (df['FTAG'] > 3))
    let f = |x: &FlexData| x > &FlexData::Uint(3);
    table.filter_any(&["FTHG","FTAG"], f).print( Some(20) );

    // All games where no goals were scored
    // Pandas equivalent: df.where((df['FTHG'] == 0) & (df['FTAG'] == 0))
    let f = |x: &FlexData| x == &FlexData::Uint(0);
    table.filter_all(&["FTHG","FTAG"], f).print( Some(20) );

    // Create new series as function of others
    // using helper functions to condense the code
    // Pandas equivalent: df['GoalDiff'] = df['FTHG'] - df['FTAG']
    let series = table.extract_series(&["FTHG","FTAG"]);
    let gd_series = series[0].sub( "GoalDiff", &FlexDataType::Int, &series[1] );
    table.add_series( gd_series );
    
    // Pandas equivalent: print( df.head(10) )
    table.print( Some(10) ); // print first 10 records only

    // Pandas equivalent: print( df.iloc[24,:] )
    table[24].print();

    // Subset selection
    table.get_subset( vec![FlexIndex::Uint(12), FlexIndex::Uint(30)]).print( None );

    // Group by Hometeams
    for (k,v) in FlexTable::group_by(&table, "HomeTeam") {
        println!("{}", k);
        v.print( Some(5) );
        break;
    }

Please refer to the tests folder for more usage examples.

Bear in mind that this library is in early development so the interface could vary significantly over time.