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
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
use std::fmt;
use std::fmt::Debug;
use std::path::PathBuf;
use std::sync::Arc;

use hive::HivePartitions;
use polars_core::prelude::*;
use recursive::recursive;

use crate::prelude::*;

pub(crate) mod aexpr;
pub(crate) mod anonymous_scan;
pub(crate) mod ir;

mod apply;
mod builder_dsl;
mod builder_ir;
pub(crate) mod conversion;
#[cfg(feature = "debugging")]
pub(crate) mod debug;
pub mod expr_ir;
mod file_scan;
mod format;
mod functions;
pub mod hive;
pub(crate) mod iterator;
mod lit;
pub(crate) mod optimizer;
pub(crate) mod options;
#[cfg(feature = "python")]
mod pyarrow;
mod schema;
pub mod visitor;

pub use aexpr::*;
pub use anonymous_scan::*;
pub use apply::*;
pub use builder_dsl::*;
pub use builder_ir::*;
pub use conversion::*;
pub(crate) use expr_ir::*;
pub use file_scan::*;
pub use functions::*;
pub use ir::*;
pub use iterator::*;
pub use lit::*;
pub use optimizer::*;
pub use schema::*;
#[cfg(feature = "serde")]
use serde::{Deserialize, Serialize};
use strum_macros::IntoStaticStr;

pub type ColumnName = Arc<str>;

#[derive(Clone, Copy, Debug)]
pub enum Context {
    /// Any operation that is done on groups
    Aggregation,
    /// Any operation that is done while projection/ selection of data
    Default,
}

// https://stackoverflow.com/questions/1031076/what-are-projection-and-selection
#[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
pub enum DslPlan {
    #[cfg(feature = "python")]
    PythonScan { options: PythonOptions },
    /// Filter on a boolean mask
    Filter {
        input: Arc<DslPlan>,
        predicate: Expr,
    },
    /// Cache the input at this point in the LP
    Cache {
        input: Arc<DslPlan>,
        id: usize,
        cache_hits: u32,
    },
    Scan {
        paths: Arc<[PathBuf]>,
        // Option as this is mostly materialized on the IR phase.
        file_info: Option<FileInfo>,
        hive_parts: Option<Arc<[HivePartitions]>>,
        predicate: Option<Expr>,
        file_options: FileScanOptions,
        scan_type: FileScan,
    },
    // we keep track of the projection and selection as it is cheaper to first project and then filter
    /// In memory DataFrame
    DataFrameScan {
        df: Arc<DataFrame>,
        schema: SchemaRef,
        // schema of the projected file
        output_schema: Option<SchemaRef>,
        filter: Option<Expr>,
    },
    /// Polars' `select` operation, this can mean projection, but also full data access.
    Select {
        expr: Vec<Expr>,
        input: Arc<DslPlan>,
        options: ProjectionOptions,
    },
    /// Groupby aggregation
    GroupBy {
        input: Arc<DslPlan>,
        keys: Vec<Expr>,
        aggs: Vec<Expr>,
        #[cfg_attr(feature = "serde", serde(skip))]
        apply: Option<(Arc<dyn DataFrameUdf>, SchemaRef)>,
        maintain_order: bool,
        options: Arc<GroupbyOptions>,
    },
    /// Join operation
    Join {
        input_left: Arc<DslPlan>,
        input_right: Arc<DslPlan>,
        left_on: Vec<Expr>,
        right_on: Vec<Expr>,
        options: Arc<JoinOptions>,
    },
    /// Adding columns to the table without a Join
    HStack {
        input: Arc<DslPlan>,
        exprs: Vec<Expr>,
        options: ProjectionOptions,
    },
    /// Remove duplicates from the table
    Distinct {
        input: Arc<DslPlan>,
        options: DistinctOptions,
    },
    /// Sort the table
    Sort {
        input: Arc<DslPlan>,
        by_column: Vec<Expr>,
        slice: Option<(i64, usize)>,
        sort_options: SortMultipleOptions,
    },
    /// Slice the table
    Slice {
        input: Arc<DslPlan>,
        offset: i64,
        len: IdxSize,
    },
    /// A (User Defined) Function
    MapFunction {
        input: Arc<DslPlan>,
        function: DslFunction,
    },
    /// Vertical concatenation
    Union {
        inputs: Vec<DslPlan>,
        args: UnionArgs,
    },
    /// Horizontal concatenation of multiple plans
    HConcat {
        inputs: Vec<DslPlan>,
        options: HConcatOptions,
    },
    /// This allows expressions to access other tables
    ExtContext {
        input: Arc<DslPlan>,
        contexts: Vec<DslPlan>,
    },
    Sink {
        input: Arc<DslPlan>,
        payload: SinkType,
    },
    IR {
        #[cfg_attr(feature = "serde", serde(skip))]
        node: Option<Node>,
        version: u32,
        // Keep the original Dsl around as we need that for serialization.
        dsl: Arc<DslPlan>,
    },
}

impl Clone for DslPlan {
    // Autogenerated by rust-analyzer, don't care about it looking nice, it just
    // calls clone on every member of every enum variant.
    #[rustfmt::skip]
    #[allow(clippy::clone_on_copy)]
    #[recursive]
    fn clone(&self) -> Self {
        match self {
            #[cfg(feature = "python")]
            Self::PythonScan { options } => Self::PythonScan { options: options.clone() },
            Self::Filter { input, predicate } => Self::Filter { input: input.clone(), predicate: predicate.clone() },
            Self::Cache { input, id, cache_hits } => Self::Cache { input: input.clone(), id: id.clone(), cache_hits: cache_hits.clone() },
            Self::Scan { paths, file_info, hive_parts, predicate, file_options, scan_type } => Self::Scan { paths: paths.clone(), file_info: file_info.clone(), hive_parts: hive_parts.clone(), predicate: predicate.clone(), file_options: file_options.clone(), scan_type: scan_type.clone() },
            Self::DataFrameScan { df, schema, output_schema, filter: selection } => Self::DataFrameScan { df: df.clone(), schema: schema.clone(), output_schema: output_schema.clone(), filter: selection.clone() },
            Self::Select { expr, input, options } => Self::Select { expr: expr.clone(), input: input.clone(), options: options.clone() },
            Self::GroupBy { input, keys, aggs,  apply, maintain_order, options } => Self::GroupBy { input: input.clone(), keys: keys.clone(), aggs: aggs.clone(), apply: apply.clone(), maintain_order: maintain_order.clone(), options: options.clone() },
            Self::Join { input_left, input_right, left_on, right_on, options } => Self::Join { input_left: input_left.clone(), input_right: input_right.clone(), left_on: left_on.clone(), right_on: right_on.clone(), options: options.clone() },
            Self::HStack { input, exprs, options } => Self::HStack { input: input.clone(), exprs: exprs.clone(),  options: options.clone() },
            Self::Distinct { input, options } => Self::Distinct { input: input.clone(), options: options.clone() },
            Self::Sort {input,by_column, slice, sort_options } => Self::Sort { input: input.clone(), by_column: by_column.clone(), slice: slice.clone(), sort_options: sort_options.clone() },
            Self::Slice { input, offset, len } => Self::Slice { input: input.clone(), offset: offset.clone(), len: len.clone() },
            Self::MapFunction { input, function } => Self::MapFunction { input: input.clone(), function: function.clone() },
            Self::Union { inputs, args} => Self::Union { inputs: inputs.clone(), args: args.clone() },
            Self::HConcat { inputs, options } => Self::HConcat { inputs: inputs.clone(), options: options.clone() },
            Self::ExtContext { input, contexts, } => Self::ExtContext { input: input.clone(), contexts: contexts.clone() },
            Self::Sink { input, payload } => Self::Sink { input: input.clone(), payload: payload.clone() },
            Self::IR {node, dsl, version} => Self::IR {node: *node, dsl: dsl.clone(), version: *version}
        }
    }
}

impl Default for DslPlan {
    fn default() -> Self {
        let df = DataFrame::new::<Series>(vec![]).unwrap();
        let schema = df.schema();
        DslPlan::DataFrameScan {
            df: Arc::new(df),
            schema: Arc::new(schema),
            output_schema: None,
            filter: None,
        }
    }
}

impl DslPlan {
    pub fn describe(&self) -> PolarsResult<String> {
        Ok(self.clone().to_alp()?.describe())
    }

    pub fn describe_tree_format(&self) -> PolarsResult<String> {
        Ok(self.clone().to_alp()?.describe_tree_format())
    }

    pub fn display(&self) -> PolarsResult<impl fmt::Display> {
        struct DslPlanDisplay(IRPlan);
        impl fmt::Display for DslPlanDisplay {
            fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
                self.0.as_ref().display().fmt(f)
            }
        }
        Ok(DslPlanDisplay(self.clone().to_alp()?))
    }

    pub fn to_alp(self) -> PolarsResult<IRPlan> {
        let mut lp_arena = Arena::with_capacity(16);
        let mut expr_arena = Arena::with_capacity(16);

        let node = to_alp(self, &mut expr_arena, &mut lp_arena, true, true)?;
        let plan = IRPlan::new(node, lp_arena, expr_arena);

        Ok(plan)
    }
}