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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567
//! A lightweight Datalog engine in Rust //! //! The intended design is that one has static `Relation` types that are sets //! of tuples, and `Variable` types that represent monotonically increasing //! sets of tuples. //! //! The types are mostly wrappers around `Vec<Tuple>` indicating sorted-ness, //! and the intent is that this code can be dropped in the middle of an otherwise //! normal Rust program, run to completion, and then the results extracted as //! vectors again. #![forbid(missing_docs)] use std::cell::RefCell; use std::cmp::Ordering; use std::iter::FromIterator; use std::rc::Rc; mod join; mod map; mod test; mod treefrog; pub use crate::join::JoinInput; pub use crate::treefrog::{ extend_anti::ExtendAnti, extend_with::ExtendWith, filter_anti::FilterAnti, filter_with::FilterWith, filters::{PrefixFilter, ValueFilter}, Leaper, Leapers, RelationLeaper, }; /// A static, ordered list of key-value pairs. /// /// A relation represents a fixed set of key-value pairs. In many places in a /// Datalog computation we want to be sure that certain relations are not able /// to vary (for example, in antijoins). #[derive(Clone)] pub struct Relation<Tuple: Ord> { /// Sorted list of distinct tuples. pub elements: Vec<Tuple>, } impl<Tuple: Ord> Relation<Tuple> { /// Merges two relations into their union. pub fn merge(self, other: Self) -> Self { let Relation { elements: mut elements1, } = self; let Relation { elements: mut elements2, } = other; // If one of the element lists is zero-length, we don't need to do any work if elements1.is_empty() { return Relation { elements: elements2, }; } if elements2.is_empty() { return Relation { elements: elements1, }; } // Make sure that elements1 starts with the lower element // Will not panic since both collections must have at least 1 element at this point if elements1[0] > elements2[0] { std::mem::swap(&mut elements1, &mut elements2); } // Fast path for when all the new elements are after the exiting ones if elements1[elements1.len() - 1] < elements2[0] { elements1.extend(elements2.into_iter()); // println!("fast path"); return Relation { elements: elements1, }; } let mut elements = Vec::with_capacity(elements1.len() + elements2.len()); let mut elements1 = elements1.drain(..); let mut elements2 = elements2.drain(..).peekable(); elements.push(elements1.next().unwrap()); if elements.first() == elements2.peek() { elements2.next(); } for elem in elements1 { while elements2.peek().map(|x| x.cmp(&elem)) == Some(Ordering::Less) { elements.push(elements2.next().unwrap()); } if elements2.peek().map(|x| x.cmp(&elem)) == Some(Ordering::Equal) { elements2.next(); } elements.push(elem); } // Finish draining second list elements.extend(elements2); Relation { elements } } /// Creates a `Relation` from the elements of the `iterator`. /// /// Same as the `from_iter` method from `std::iter::FromIterator` trait. pub fn from_iter<I>(iterator: I) -> Self where I: IntoIterator<Item = Tuple>, { iterator.into_iter().collect() } /// Creates a `Relation` using the `leapjoin` logic; /// see [`Variable::from_leapjoin`] pub fn from_leapjoin<'leap, SourceTuple: Ord, Val: Ord + 'leap>( source: &Relation<SourceTuple>, leapers: impl Leapers<'leap, SourceTuple, Val>, logic: impl FnMut(&SourceTuple, &Val) -> Tuple, ) -> Self { treefrog::leapjoin(&source.elements, leapers, logic) } /// Creates a `Relation` by joining the values from `input1` and /// `input2` and then applying `logic`. Like /// [`Variable::from_join`] except for use where the inputs are /// not varying across iterations. pub fn from_join<Key: Ord, Val1: Ord, Val2: Ord>( input1: &Relation<(Key, Val1)>, input2: &Relation<(Key, Val2)>, logic: impl FnMut(&Key, &Val1, &Val2) -> Tuple, ) -> Self { join::join_into_relation(input1, input2, logic) } /// Creates a `Relation` by removing all values from `input1` that /// share a key with `input2`, and then transforming the resulting /// tuples with the `logic` closure. Like /// [`Variable::from_antijoin`] except for use where the inputs /// are not varying across iterations. pub fn from_antijoin<Key: Ord, Val1: Ord>( input1: &Relation<(Key, Val1)>, input2: &Relation<Key>, logic: impl FnMut(&Key, &Val1) -> Tuple, ) -> Self { join::antijoin(input1, input2, logic) } /// Construct a new relation by mapping another one. Equivalent to /// creating an iterator but perhaps more convenient. Analogous to /// `Variable::from_map`. pub fn from_map<T2: Ord>(input: &Relation<T2>, logic: impl FnMut(&T2) -> Tuple) -> Self { input.iter().map(logic).collect() } /// Creates a `Relation` from a vector of tuples. pub fn from_vec(mut elements: Vec<Tuple>) -> Self { elements.sort(); elements.dedup(); Relation { elements } } } impl<Tuple: Ord> From<Vec<Tuple>> for Relation<Tuple> { fn from(iterator: Vec<Tuple>) -> Self { Self::from_vec(iterator) } } impl<Tuple: Ord> FromIterator<Tuple> for Relation<Tuple> { fn from_iter<I>(iterator: I) -> Self where I: IntoIterator<Item = Tuple>, { Relation::from_vec(iterator.into_iter().collect()) } } impl<'tuple, Tuple: 'tuple + Copy + Ord> FromIterator<&'tuple Tuple> for Relation<Tuple> { fn from_iter<I>(iterator: I) -> Self where I: IntoIterator<Item = &'tuple Tuple>, { Relation::from_vec(iterator.into_iter().cloned().collect()) } } impl<Tuple: Ord> std::ops::Deref for Relation<Tuple> { type Target = [Tuple]; fn deref(&self) -> &Self::Target { &self.elements[..] } } /// An iterative context for recursive evaluation. /// /// An `Iteration` tracks monotonic variables, and monitors their progress. /// It can inform the user if they have ceased changing, at which point the /// computation should be done. pub struct Iteration { variables: Vec<Box<dyn VariableTrait>>, } impl Iteration { /// Create a new iterative context. pub fn new() -> Self { Iteration { variables: Vec::new(), } } /// Reports whether any of the monitored variables have changed since /// the most recent call. pub fn changed(&mut self) -> bool { let mut result = false; for variable in self.variables.iter_mut() { if variable.changed() { result = true; } } result } /// Creates a new named variable associated with the iterative context. pub fn variable<Tuple: Ord + 'static>(&mut self, name: &str) -> Variable<Tuple> { let variable = Variable::new(name); self.variables.push(Box::new(variable.clone())); variable } /// Creates a new named variable associated with the iterative context. /// /// This variable will not be maintained distinctly, and may advertise tuples as /// recent multiple times (perhaps unboundedly many times). pub fn variable_indistinct<Tuple: Ord + 'static>(&mut self, name: &str) -> Variable<Tuple> { let mut variable = Variable::new(name); variable.distinct = false; self.variables.push(Box::new(variable.clone())); variable } } /// A type that can report on whether it has changed. trait VariableTrait { /// Reports whether the variable has changed since it was last asked. fn changed(&mut self) -> bool; } /// An monotonically increasing set of `Tuple`s. /// /// There are three stages in the lifecycle of a tuple: /// /// 1. A tuple is added to `self.to_add`, but is not yet visible externally. /// 2. Newly added tuples are then promoted to `self.recent` for one iteration. /// 3. After one iteration, recent tuples are moved to `self.tuples` for posterity. /// /// Each time `self.changed()` is called, the `recent` relation is folded into `tuples`, /// and the `to_add` relations are merged, potentially deduplicated against `tuples`, and /// then made `recent`. This way, across calls to `changed()` all added tuples are in /// `recent` at least once and eventually all are in `tuples`. /// /// A `Variable` may optionally be instructed not to de-duplicate its tuples, for reasons /// of performance. Such a variable cannot be relied on to terminate iterative computation, /// and it is important that any cycle of derivations have at least one de-duplicating /// variable on it. pub struct Variable<Tuple: Ord> { /// Should the variable be maintained distinctly. distinct: bool, /// A useful name for the variable. name: String, /// A list of relations whose union are the accepted tuples. pub stable: Rc<RefCell<Vec<Relation<Tuple>>>>, /// A list of recent tuples, still to be processed. pub recent: Rc<RefCell<Relation<Tuple>>>, /// A list of future tuples, to be introduced. to_add: Rc<RefCell<Vec<Relation<Tuple>>>>, } // Operator implementations. impl<Tuple: Ord> Variable<Tuple> { /// Adds tuples that result from joining `input1` and `input2` -- /// each of the inputs must be a set of (Key, Value) tuples. Both /// `input1` and `input2` must have the same type of key (`K`) but /// they can have distinct value types (`V1` and `V2` /// respectively). The `logic` closure will be invoked for each /// key that appears in both inputs; it is also given the two /// values, and from those it should construct the resulting /// value. /// /// Note that `input1` must be a variable, but `input2` can be a /// relation or a variable. Therefore, you cannot join two /// relations with this method. This is not because the result /// would be wrong, but because it would be inefficient: the /// result from such a join cannot vary across iterations (as /// relations are fixed), so you should prefer to invoke `insert` /// on a relation created by `Relation::from_join` instead. /// /// # Examples /// /// This example starts a collection with the pairs (x, x+1) and (x+1, x) for x in 0 .. 10. /// It then adds pairs (y, z) for which (x, y) and (x, z) are present. Because the initial /// pairs are symmetric, this should result in all pairs (x, y) for x and y in 0 .. 11. /// /// ``` /// use datafrog::{Iteration, Relation}; /// /// let mut iteration = Iteration::new(); /// let variable = iteration.variable::<(usize, usize)>("source"); /// variable.extend((0 .. 10).map(|x| (x, x + 1))); /// variable.extend((0 .. 10).map(|x| (x + 1, x))); /// /// while iteration.changed() { /// variable.from_join(&variable, &variable, |&key, &val1, &val2| (val1, val2)); /// } /// /// let result = variable.complete(); /// assert_eq!(result.len(), 121); /// ``` pub fn from_join<'me, K: Ord, V1: Ord, V2: Ord>( &self, input1: &'me Variable<(K, V1)>, input2: impl JoinInput<'me, (K, V2)>, logic: impl FnMut(&K, &V1, &V2) -> Tuple, ) { join::join_into(input1, input2, self, logic) } /// Adds tuples from `input1` whose key is not present in `input2`. /// /// Note that `input1` must be a variable: if you have a relation /// instead, you can use `Relation::from_antijoin` and then /// `Variable::insert`. Note that the result will not vary during /// the iteration. /// /// # Examples /// /// This example starts a collection with the pairs (x, x+1) for x in 0 .. 10. It then /// adds any pairs (x+1,x) for which x is not a multiple of three. That excludes four /// pairs (for 0, 3, 6, and 9) which should leave us with 16 total pairs. /// /// ``` /// use datafrog::{Iteration, Relation}; /// /// let mut iteration = Iteration::new(); /// let variable = iteration.variable::<(usize, usize)>("source"); /// variable.extend((0 .. 10).map(|x| (x, x + 1))); /// /// let relation: Relation<_> = (0 .. 10).filter(|x| x % 3 == 0).collect(); /// /// while iteration.changed() { /// variable.from_antijoin(&variable, &relation, |&key, &val| (val, key)); /// } /// /// let result = variable.complete(); /// assert_eq!(result.len(), 16); /// ``` pub fn from_antijoin<K: Ord, V: Ord>( &self, input1: &Variable<(K, V)>, input2: &Relation<K>, logic: impl FnMut(&K, &V) -> Tuple, ) { self.insert(join::antijoin(input1, input2, logic)) } /// Adds tuples that result from mapping `input`. /// /// # Examples /// /// This example starts a collection with the pairs (x, x) for x in 0 .. 10. It then /// repeatedly adds any pairs (x, z) for (x, y) in the collection, where z is the Collatz /// step for y: it is y/2 if y is even, and 3*y + 1 if y is odd. This produces all of the /// pairs (x, y) where x visits y as part of its Collatz journey. /// /// ``` /// use datafrog::{Iteration, Relation}; /// /// let mut iteration = Iteration::new(); /// let variable = iteration.variable::<(usize, usize)>("source"); /// variable.extend((0 .. 10).map(|x| (x, x))); /// /// while iteration.changed() { /// variable.from_map(&variable, |&(key, val)| /// if val % 2 == 0 { /// (key, val/2) /// } /// else { /// (key, 3*val + 1) /// }); /// } /// /// let result = variable.complete(); /// assert_eq!(result.len(), 74); /// ``` pub fn from_map<T2: Ord>(&self, input: &Variable<T2>, logic: impl FnMut(&T2) -> Tuple) { map::map_into(input, self, logic) } /// Adds tuples that result from combining `source` with the /// relations given in `leapers`. This operation is very flexible /// and can be used to do a combination of joins and anti-joins. /// The main limitation is that the things being combined must /// consist of one dynamic variable (`source`) and then several /// fixed relations (`leapers`). /// /// The idea is as follows: /// /// - You will be inserting new tuples that result from joining (and anti-joining) /// some dynamic variable `source` of source tuples (`SourceTuple`) /// with some set of values (of type `Val`). /// - You provide these values by combining `source` with a set of leapers /// `leapers`, each of which is derived from a fixed relation. The `leapers` /// should be either a single leaper (of suitable type) or else a tuple of leapers. /// You can create a leaper in one of two ways: /// - Extension: In this case, you have a relation of type `(K, Val)` for some /// type `K`. You provide a closure that maps from `SourceTuple` to the key /// `K`. If you use `relation.extend_with`, then any `Val` values the /// relation provides will be added to the set of values; if you use /// `extend_anti`, then the `Val` values will be removed. /// - Filtering: In this case, you have a relation of type `K` for some /// type `K` and you provide a closure that maps from `SourceTuple` to /// the key `K`. Filters don't provide values but they remove source /// tuples. /// - Finally, you get a callback `logic` that accepts each `(SourceTuple, Val)` /// that was successfully joined (and not filtered) and which maps to the /// type of this variable. pub fn from_leapjoin<'leap, SourceTuple: Ord, Val: Ord + 'leap>( &self, source: &Variable<SourceTuple>, leapers: impl Leapers<'leap, SourceTuple, Val>, logic: impl FnMut(&SourceTuple, &Val) -> Tuple, ) { self.insert(treefrog::leapjoin(&source.recent.borrow(), leapers, logic)); } } impl<Tuple: Ord> Clone for Variable<Tuple> { fn clone(&self) -> Self { Variable { distinct: self.distinct, name: self.name.clone(), stable: self.stable.clone(), recent: self.recent.clone(), to_add: self.to_add.clone(), } } } impl<Tuple: Ord> Variable<Tuple> { fn new(name: &str) -> Self { Variable { distinct: true, name: name.to_string(), stable: Rc::new(RefCell::new(Vec::new())), recent: Rc::new(RefCell::new(Vec::new().into())), to_add: Rc::new(RefCell::new(Vec::new())), } } /// Inserts a relation into the variable. /// /// This is most commonly used to load initial values into a variable. /// it is not obvious that it should be commonly used otherwise, but /// it should not be harmful. pub fn insert(&self, relation: Relation<Tuple>) { if !relation.is_empty() { self.to_add.borrow_mut().push(relation); } } /// Extend the variable with values from the iterator. /// /// This is most commonly used to load initial values into a variable. /// it is not obvious that it should be commonly used otherwise, but /// it should not be harmful. pub fn extend<T>(&self, iterator: impl IntoIterator<Item = T>) where Relation<Tuple>: FromIterator<T>, { self.insert(iterator.into_iter().collect()); } /// Consumes the variable and returns a relation. /// /// This method removes the ability for the variable to develop, and /// flattens all internal tuples down to one relation. The method /// asserts that iteration has completed, in that `self.recent` and /// `self.to_add` should both be empty. pub fn complete(self) -> Relation<Tuple> { assert!(self.recent.borrow().is_empty()); assert!(self.to_add.borrow().is_empty()); let mut result: Relation<Tuple> = Vec::new().into(); while let Some(batch) = self.stable.borrow_mut().pop() { result = result.merge(batch); } result } } impl<Tuple: Ord> VariableTrait for Variable<Tuple> { fn changed(&mut self) -> bool { // 1. Merge self.recent into self.stable. if !self.recent.borrow().is_empty() { let mut recent = ::std::mem::replace(&mut (*self.recent.borrow_mut()), Vec::new().into()); while self .stable .borrow() .last() .map(|x| x.len() <= 2 * recent.len()) == Some(true) { let last = self.stable.borrow_mut().pop().unwrap(); recent = recent.merge(last); } self.stable.borrow_mut().push(recent); } // 2. Move self.to_add into self.recent. let to_add = self.to_add.borrow_mut().pop(); if let Some(mut to_add) = to_add { while let Some(to_add_more) = self.to_add.borrow_mut().pop() { to_add = to_add.merge(to_add_more); } // 2b. Restrict `to_add` to tuples not in `self.stable`. if self.distinct { for batch in self.stable.borrow().iter() { let mut slice = &batch[..]; // Only gallop if the slice is relatively large. if slice.len() > 4 * to_add.elements.len() { to_add.elements.retain(|x| { slice = join::gallop(slice, |y| y < x); slice.is_empty() || &slice[0] != x }); } else { to_add.elements.retain(|x| { while !slice.is_empty() && &slice[0] < x { slice = &slice[1..]; } slice.is_empty() || &slice[0] != x }); } } } *self.recent.borrow_mut() = to_add; } // let mut total = 0; // for tuple in self.stable.borrow().iter() { // total += tuple.len(); // } // println!("Variable\t{}\t{}\t{}", self.name, total, self.recent.borrow().len()); !self.recent.borrow().is_empty() } } // impl<Tuple: Ord> Drop for Variable<Tuple> { // fn drop(&mut self) { // let mut total = 0; // for batch in self.stable.borrow().iter() { // total += batch.len(); // } // println!("FINAL: {:?}\t{:?}", self.name, total); // } // }