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
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
// Copyright Materialize, Inc. All rights reserved.
//
// Use of this software is governed by the Business Source License
// included in the LICENSE file.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0.

//! Transformations based on pulling information about individual columns from sources.

use std::collections::HashMap;

use itertools::Itertools;

use expr::{RelationExpr, ScalarExpr, UnaryFunc};
use repr::Datum;
use repr::{ColumnType, RelationType, ScalarType};

use crate::TransformArgs;

/// Harvest and act upon per-column information.
#[derive(Debug)]
pub struct ColumnKnowledge;

impl crate::Transform for ColumnKnowledge {
    fn transform(
        &self,
        expr: &mut RelationExpr,
        _: TransformArgs,
    ) -> Result<(), crate::TransformError> {
        self.transform(expr)
    }
}

impl ColumnKnowledge {
    /// Transforms an expression through accumulated knowledge.
    pub fn transform(&self, expr: &mut RelationExpr) -> Result<(), crate::TransformError> {
        ColumnKnowledge::harvest(expr, &mut HashMap::new())?;
        Ok(())
    }

    /// Harvest per-column knowledge.
    fn harvest(
        expr: &mut RelationExpr,
        knowledge: &mut HashMap<expr::Id, Vec<DatumKnowledge>>,
    ) -> Result<Vec<DatumKnowledge>, crate::TransformError> {
        Ok(match expr {
            RelationExpr::ArrangeBy { input, .. } => ColumnKnowledge::harvest(input, knowledge)?,
            RelationExpr::Get { id, typ } => knowledge
                .get(id)
                .cloned()
                .unwrap_or_else(|| typ.column_types.iter().map(DatumKnowledge::from).collect()),
            RelationExpr::Constant { rows, typ } => {
                if rows.len() == 1 {
                    let mut row_packer = repr::RowPacker::new();
                    rows[0]
                        .0
                        .iter()
                        .zip(typ.column_types.iter())
                        .map(|(datum, typ)| DatumKnowledge {
                            value: Some((row_packer.pack(Some(datum.clone())), typ.clone())),
                            nullable: datum == Datum::Null,
                        })
                        .collect()
                } else {
                    typ.column_types.iter().map(DatumKnowledge::from).collect()
                }
            }
            RelationExpr::Let { id, value, body } => {
                let value_knowledge = ColumnKnowledge::harvest(value, knowledge)?;
                let prior_knowledge =
                    knowledge.insert(expr::Id::Local(id.clone()), value_knowledge);
                let body_knowledge = ColumnKnowledge::harvest(body, knowledge)?;
                knowledge.remove(&expr::Id::Local(id.clone()));
                if let Some(prior_knowledge) = prior_knowledge {
                    knowledge.insert(expr::Id::Local(id.clone()), prior_knowledge);
                }
                body_knowledge
            }
            RelationExpr::Project { input, outputs } => {
                let input_knowledge = ColumnKnowledge::harvest(input, knowledge)?;
                outputs
                    .iter()
                    .map(|i| input_knowledge[*i].clone())
                    .collect()
            }
            RelationExpr::Map { input, scalars } => {
                let mut input_knowledge = ColumnKnowledge::harvest(input, knowledge)?;
                for scalar in scalars.iter_mut() {
                    let know = optimize(scalar, &input.typ(), &input_knowledge[..])?;
                    input_knowledge.push(know);
                }
                input_knowledge
            }
            RelationExpr::FlatMap {
                input,
                func,
                exprs,
                demand: _,
            } => {
                let mut input_knowledge = ColumnKnowledge::harvest(input, knowledge)?;
                for expr in exprs {
                    optimize(expr, &input.typ(), &input_knowledge[..])?;
                }
                let func_typ = func.output_type();
                input_knowledge.extend(func_typ.column_types.iter().map(DatumKnowledge::from));
                input_knowledge
            }
            RelationExpr::Filter { input, predicates } => {
                let mut input_knowledge = ColumnKnowledge::harvest(input, knowledge)?;
                for predicate in predicates.iter_mut() {
                    optimize(predicate, &input.typ(), &input_knowledge[..])?;
                }
                // If any predicate tests a column for equality, truth, or is_null, we learn stuff.
                for predicate in predicates.iter() {
                    // Equality tests allow us to unify the column knowledge of each input.
                    if let ScalarExpr::CallBinary { func, expr1, expr2 } = predicate {
                        if func == &expr::BinaryFunc::Eq {
                            // Collect knowledge about the inputs (for columns and literals).
                            let mut knowledge = DatumKnowledge::default();
                            if let ScalarExpr::Column(c) = &**expr1 {
                                knowledge.absorb(&input_knowledge[*c]);
                            }
                            if let ScalarExpr::Column(c) = &**expr2 {
                                knowledge.absorb(&input_knowledge[*c]);
                            }
                            // Absorb literal knowledge about columns.
                            knowledge.absorb(&DatumKnowledge::from(&**expr1));
                            knowledge.absorb(&DatumKnowledge::from(&**expr2));

                            // Write back unified knowledge to each column.
                            if let ScalarExpr::Column(c) = &**expr1 {
                                input_knowledge[*c].absorb(&knowledge);
                            }
                            if let ScalarExpr::Column(c) = &**expr2 {
                                input_knowledge[*c].absorb(&knowledge);
                            }
                        }
                    }
                    if let ScalarExpr::CallUnary {
                        func: UnaryFunc::Not,
                        expr,
                    } = predicate
                    {
                        if let ScalarExpr::CallUnary {
                            func: UnaryFunc::IsNull,
                            expr,
                        } = &**expr
                        {
                            if let ScalarExpr::Column(c) = &**expr {
                                input_knowledge[*c].nullable = false;
                            }
                        }
                    }
                }

                input_knowledge
            }
            RelationExpr::Join {
                inputs,
                equivalences,
                ..
            } => {
                let mut knowledges = Vec::new();
                for input in inputs.iter_mut() {
                    for knowledge in ColumnKnowledge::harvest(input, knowledge)? {
                        knowledges.push(knowledge);
                    }
                }

                for equivalence in equivalences.iter_mut() {
                    let mut knowledge = DatumKnowledge::default();

                    // We can produce composite knowledge for everything in the equivalence class.
                    for expr in equivalence.iter_mut() {
                        if let ScalarExpr::Column(c) = expr {
                            knowledge.absorb(&knowledges[*c]);
                        }
                        knowledge.absorb(&DatumKnowledge::from(&*expr));
                    }
                    for expr in equivalence.iter_mut() {
                        if let ScalarExpr::Column(c) = expr {
                            knowledges[*c] = knowledge.clone();
                        }
                    }
                }

                knowledges
            }
            RelationExpr::Reduce {
                input,
                group_key,
                aggregates,
                monotonic: _,
                expected_group_size: _,
            } => {
                let input_knowledge = ColumnKnowledge::harvest(input, knowledge)?;
                let mut output = group_key
                    .iter_mut()
                    .map(|k| optimize(k, &input.typ(), &input_knowledge[..]))
                    .collect::<Result<Vec<_>, _>>()?;
                for aggregate in aggregates.iter_mut() {
                    use expr::AggregateFunc;
                    let knowledge =
                        optimize(&mut aggregate.expr, &input.typ(), &input_knowledge[..])?;
                    // This could be improved.
                    let knowledge = match aggregate.func {
                        AggregateFunc::MaxInt32
                        | AggregateFunc::MaxInt64
                        | AggregateFunc::MaxFloat32
                        | AggregateFunc::MaxFloat64
                        | AggregateFunc::MaxDecimal
                        | AggregateFunc::MaxBool
                        | AggregateFunc::MaxString
                        | AggregateFunc::MaxDate
                        | AggregateFunc::MaxTimestamp
                        | AggregateFunc::MaxTimestampTz
                        | AggregateFunc::MinInt32
                        | AggregateFunc::MinInt64
                        | AggregateFunc::MinFloat32
                        | AggregateFunc::MinFloat64
                        | AggregateFunc::MinDecimal
                        | AggregateFunc::MinBool
                        | AggregateFunc::MinString
                        | AggregateFunc::MinDate
                        | AggregateFunc::MinTimestamp
                        | AggregateFunc::MinTimestampTz
                        | AggregateFunc::Any
                        | AggregateFunc::All => {
                            // These methods propagate constant values exactly.
                            knowledge
                        }
                        _ => {
                            // All aggregates are non-null if their inputs are non-null.
                            DatumKnowledge {
                                value: None,
                                nullable: knowledge.nullable,
                            }
                        }
                    };
                    output.push(knowledge);
                }
                output
            }
            RelationExpr::TopK { input, .. } => ColumnKnowledge::harvest(input, knowledge)?,
            RelationExpr::Negate { input } => ColumnKnowledge::harvest(input, knowledge)?,
            RelationExpr::Threshold { input } => ColumnKnowledge::harvest(input, knowledge)?,
            RelationExpr::Union { base, inputs } => {
                let mut know = ColumnKnowledge::harvest(base, knowledge)?;
                for input in inputs {
                    know = know
                        .into_iter()
                        .zip_eq(ColumnKnowledge::harvest(input, knowledge)?)
                        .map(|(k1, k2)| DatumKnowledge {
                            value: if k1.value == k2.value {
                                k1.value.clone()
                            } else {
                                None
                            },
                            nullable: k1.nullable || k2.nullable,
                        })
                        .collect();
                }
                know
            }
        })
    }
}

/// Information about a specific column.
#[derive(Clone, Debug)]
pub struct DatumKnowledge {
    /// If set, a specific value for the column.
    value: Option<(repr::Row, ColumnType)>,
    /// If false, the value is not `Datum::Null`.
    nullable: bool,
}

impl DatumKnowledge {
    // Intersects the two knowledge about a column.
    fn absorb(&mut self, other: &Self) {
        self.nullable &= other.nullable;
        if self.value.is_none() {
            self.value = other.value.clone()
        }
    }
}

impl Default for DatumKnowledge {
    fn default() -> Self {
        Self {
            value: None,
            nullable: true,
        }
    }
}

impl From<&ScalarExpr> for DatumKnowledge {
    fn from(expr: &ScalarExpr) -> Self {
        if let ScalarExpr::Literal(Ok(l), t) = expr {
            Self {
                value: Some((l.clone(), t.clone())),
                nullable: expr.is_literal_null(),
            }
        } else {
            Self::default()
        }
    }
}

impl From<&ColumnType> for DatumKnowledge {
    fn from(typ: &ColumnType) -> Self {
        Self {
            value: None,
            nullable: typ.nullable,
        }
    }
}

/// Attempts to optimize
pub fn optimize(
    expr: &mut ScalarExpr,
    input_type: &RelationType,
    column_knowledge: &[DatumKnowledge],
) -> Result<DatumKnowledge, crate::TransformError> {
    Ok(match expr {
        ScalarExpr::Column(index) => {
            let index = *index;
            if let Some((datum, typ)) = &column_knowledge[index].value {
                *expr = ScalarExpr::Literal(Ok(datum.clone()), typ.clone());
            }
            column_knowledge[index].clone()
        }
        ScalarExpr::Literal(res, typ) => {
            let row = match res {
                Ok(row) => row,
                Err(err) => return Err(err.clone().into()),
            };
            DatumKnowledge {
                value: Some((row.clone(), typ.clone())),
                nullable: row.unpack_first() == Datum::Null,
            }
        }
        ScalarExpr::CallNullary(_) => {
            expr.reduce(input_type);
            optimize(expr, input_type, column_knowledge)?
        }
        ScalarExpr::CallUnary { func, expr: inner } => {
            let knowledge = optimize(inner, input_type, column_knowledge)?;
            if knowledge.value.is_some() {
                expr.reduce(input_type);
                optimize(expr, input_type, column_knowledge)?
            } else if func == &UnaryFunc::IsNull && !knowledge.nullable {
                *expr = ScalarExpr::literal_ok(Datum::False, ScalarType::Bool.nullable(false));
                optimize(expr, input_type, column_knowledge)?
            } else {
                DatumKnowledge::default()
            }
        }
        ScalarExpr::CallBinary {
            func: _,
            expr1,
            expr2,
        } => {
            let knowledge1 = optimize(expr1, input_type, column_knowledge)?;
            let knowledge2 = optimize(expr2, input_type, column_knowledge)?;
            if knowledge1.value.is_some() && knowledge2.value.is_some() {
                expr.reduce(input_type);
                optimize(expr, input_type, column_knowledge)?
            } else {
                DatumKnowledge::default()
            }
        }
        ScalarExpr::CallVariadic { func: _, exprs } => {
            let mut knows = Vec::new();
            for expr in exprs.iter_mut() {
                knows.push(optimize(expr, input_type, column_knowledge)?);
            }

            if knows.iter().all(|k| k.value.is_some()) {
                expr.reduce(input_type);
                optimize(expr, input_type, column_knowledge)?
            } else {
                DatumKnowledge::default()
            }
        }
        ScalarExpr::If { cond, then, els } => {
            if let Some((value, _typ)) = optimize(cond, input_type, column_knowledge)?.value {
                match value.unpack_first() {
                    Datum::True => *expr = (**then).clone(),
                    Datum::False | Datum::Null => *expr = (**els).clone(),
                    d => panic!("IF condition evaluated to non-boolean datum {:?}", d),
                }
                optimize(expr, input_type, column_knowledge)?
            } else {
                DatumKnowledge::default()
            }
        }
    })
}