1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196

use std::ops::{Add, Div, Mul, Rem, Sub};

use rayon::prelude::*;

use crate::prelude::*;
use crate::utils::try_get_supertype;
use crate::POOL;

/// Get the supertype that is valid for all columns in the [`DataFrame`].
/// This reduces casting of the rhs in arithmetic.
fn get_supertype_all(df: &DataFrame, rhs: &Series) -> PolarsResult<DataType> {
    df.columns.iter().try_fold(rhs.dtype().clone(), |dt, s| {
        try_get_supertype(s.dtype(), &dt)
    })
}

macro_rules! impl_arithmetic {
    ($self:expr, $rhs:expr, $operand: tt) => {{
        let st = get_supertype_all($self, $rhs)?;
        let rhs = $rhs.cast(&st)?;
        let cols = POOL.install(|| {$self.columns.par_iter().map(|s| {
            Ok(&s.cast(&st)? $operand &rhs)
        }).collect::<PolarsResult<_>>()})?;
        Ok(DataFrame::new_no_checks(cols))
    }}
}

impl Add<&Series> for &DataFrame {
    type Output = PolarsResult<DataFrame>;

    fn add(self, rhs: &Series) -> Self::Output {
        impl_arithmetic!(self, rhs, +)
    }
}

impl Add<&Series> for DataFrame {
    type Output = PolarsResult<DataFrame>;

    fn add(self, rhs: &Series) -> Self::Output {
        (&self).add(rhs)
    }
}

impl Sub<&Series> for &DataFrame {
    type Output = PolarsResult<DataFrame>;

    fn sub(self, rhs: &Series) -> Self::Output {
        impl_arithmetic!(self, rhs, -)
    }
}

impl Sub<&Series> for DataFrame {
    type Output = PolarsResult<DataFrame>;

    fn sub(self, rhs: &Series) -> Self::Output {
        (&self).sub(rhs)
    }
}

impl Mul<&Series> for &DataFrame {
    type Output = PolarsResult<DataFrame>;

    fn mul(self, rhs: &Series) -> Self::Output {
        impl_arithmetic!(self, rhs, *)
    }
}

impl Mul<&Series> for DataFrame {
    type Output = PolarsResult<DataFrame>;

    fn mul(self, rhs: &Series) -> Self::Output {
        (&self).mul(rhs)
    }
}

impl Div<&Series> for &DataFrame {
    type Output = PolarsResult<DataFrame>;

    fn div(self, rhs: &Series) -> Self::Output {
        impl_arithmetic!(self, rhs, /)
    }
}

impl Div<&Series> for DataFrame {
    type Output = PolarsResult<DataFrame>;

    fn div(self, rhs: &Series) -> Self::Output {
        (&self).div(rhs)
    }
}

impl Rem<&Series> for &DataFrame {
    type Output = PolarsResult<DataFrame>;

    fn rem(self, rhs: &Series) -> Self::Output {
        impl_arithmetic!(self, rhs, %)
    }
}

impl Rem<&Series> for DataFrame {
    type Output = PolarsResult<DataFrame>;

    fn rem(self, rhs: &Series) -> Self::Output {
        (&self).rem(rhs)
    }
}

impl DataFrame {
    fn binary_aligned(
        &self,
        other: &DataFrame,
        f: &(dyn Fn(&Series, &Series) -> PolarsResult<Series> + Sync + Send),
    ) -> PolarsResult<DataFrame> {
        let max_len = std::cmp::max(self.height(), other.height());
        let max_width = std::cmp::max(self.width(), other.width());
        let mut cols = self
            .get_columns()
            .par_iter()
            .zip(other.get_columns().par_iter())
            .map(|(l, r)| {
                let diff_l = max_len - l.len();
                let diff_r = max_len - r.len();

                let st = try_get_supertype(l.dtype(), r.dtype())?;
                let mut l = l.cast(&st)?;
                let mut r = r.cast(&st)?;

                if diff_l > 0 {
                    l = l.extend_constant(AnyValue::Null, diff_l)?;
                };
                if diff_r > 0 {
                    r = r.extend_constant(AnyValue::Null, diff_r)?;
                };

                f(&l, &r)
            })
            .collect::<PolarsResult<Vec<_>>>()?;

        let col_len = cols.len();
        if col_len < max_width {
            let df = if col_len < self.width() { self } else { other };

            for i in col_len..max_len {
                let s = &df.get_columns()[i];
                let name = s.name();
                let dtype = s.dtype();

                // trick to fill a series with nulls
                let vals: &[Option<i32>] = &[None];
                let s = Series::new(name, vals).cast(dtype)?;
                cols.push(s.new_from_index(0, max_len))
            }
        }
        DataFrame::new(cols)
    }
}

impl Add<&DataFrame> for &DataFrame {
    type Output = PolarsResult<DataFrame>;

    fn add(self, rhs: &DataFrame) -> Self::Output {
        self.binary_aligned(rhs, &|a, b| Ok(a + b))
    }
}

impl Sub<&DataFrame> for &DataFrame {
    type Output = PolarsResult<DataFrame>;

    fn sub(self, rhs: &DataFrame) -> Self::Output {
        self.binary_aligned(rhs, &|a, b| Ok(a - b))
    }
}

impl Div<&DataFrame> for &DataFrame {
    type Output = PolarsResult<DataFrame>;

    fn div(self, rhs: &DataFrame) -> Self::Output {
        self.binary_aligned(rhs, &|a, b| Ok(a / b))
    }
}

impl Mul<&DataFrame> for &DataFrame {
    type Output = PolarsResult<DataFrame>;

    fn mul(self, rhs: &DataFrame) -> Self::Output {
        self.binary_aligned(rhs, &|a, b| Ok(a * b))
    }
}

impl Rem<&DataFrame> for &DataFrame {
    type Output = PolarsResult<DataFrame>;

    fn rem(self, rhs: &DataFrame) -> Self::Output {
        self.binary_aligned(rhs, &|a, b| Ok(a % b))
    }
}