polars-expr 0.53.0

Physical expression implementation of the Polars project.
Documentation 
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

use polars_core::chunked_array::cast::CastOptions;
use polars_core::prelude::*;

use super::*;
use crate::expressions::{AggState, AggregationContext, PhysicalExpr};

pub struct CastExpr {
    pub(crate) input: Arc<dyn PhysicalExpr>,
    pub(crate) dtype: DataType,
    pub(crate) expr: Expr,
    pub(crate) options: CastOptions,
}

impl CastExpr {
    fn finish(&self, input: &Column) -> PolarsResult<Column> {
        input.cast_with_options(&self.dtype, self.options)
    }
}

impl PhysicalExpr for CastExpr {
    fn as_expression(&self) -> Option<&Expr> {
        Some(&self.expr)
    }

    fn evaluate(&self, df: &DataFrame, state: &ExecutionState) -> PolarsResult<Column> {
        let column = self.input.evaluate(df, state)?;
        self.finish(&column)
    }

    #[allow(clippy::ptr_arg)]
    fn evaluate_on_groups<'a>(
        &self,
        df: &DataFrame,
        groups: &'a GroupPositions,
        state: &ExecutionState,
    ) -> PolarsResult<AggregationContext<'a>> {
        let mut ac = self.input.evaluate_on_groups(df, groups, state)?;

        match ac.agg_state() {
            // this will not explode and potentially increase memory due to overlapping groups
            AggState::AggregatedList(s) => {
                let ca = s.list().unwrap();
                let casted = ca.apply_to_inner(&|s| {
                    self.finish(&s.into_column())
                        .map(|c| c.take_materialized_series())
                })?;
                ac.with_values(casted.into_column(), true, None)?;
            },
            AggState::AggregatedScalar(s) => {
                let s = self.finish(&s.clone().into_column())?;
                if ac.is_literal() {
                    ac.with_literal(s);
                } else {
                    ac.with_values(s, true, None)?;
                }
            },
            AggState::NotAggregated(_) => {
                if match self.options {
                    CastOptions::NonStrict | CastOptions::Overflowing => true,
                    CastOptions::Strict => ac.original_len,
                } {
                    // before we flatten, make sure that groups are updated
                    ac.groups();

                    let s = ac.flat_naive();
                    let s = self.finish(&s.as_ref().clone().into_column())?;

                    ac.with_values(s, false, None)?;
                } else {
                    // We need to perform aggregation only for strict mode, since if this is not done,
                    // filtered-out values may incorrectly cause a cast error.
                    let s = ac.aggregated();
                    let ca = s.list().unwrap();
                    let casted = ca.apply_to_inner(&|s| {
                        self.finish(&s.into_column())
                            .map(|c| c.take_materialized_series())
                    })?;
                    ac.with_values(casted.into_column(), true, None)?;
                }
            },

            AggState::LiteralScalar(s) => {
                let s = self.finish(s)?;
                ac.with_literal(s);
            },
        }

        Ok(ac)
    }

    fn to_field(&self, input_schema: &Schema) -> PolarsResult<Field> {
        self.input.to_field(input_schema).map(|mut fld| {
            fld.coerce(self.dtype.clone());
            fld
        })
    }

    fn is_scalar(&self) -> bool {
        self.input.is_scalar()
    }
}