1 //! Note: most of the tests relevant to this file can be found (at the time of writing) in
2 //! src/tests/ui/pattern/usefulness.
4 //! This file includes the logic for exhaustiveness and usefulness checking for
5 //! pattern-matching. Specifically, given a list of patterns for a type, we can
7 //! (a) the patterns cover every possible constructor for the type (exhaustiveness)
8 //! (b) each pattern is necessary (usefulness)
10 //! The algorithm implemented here is a modified version of the one described in:
11 //! http://moscova.inria.fr/~maranget/papers/warn/index.html
12 //! However, to save future implementors from reading the original paper, we
13 //! summarise the algorithm here to hopefully save time and be a little clearer
14 //! (without being so rigorous).
18 //! The core of the algorithm revolves about a "usefulness" check. In particular, we
19 //! are trying to compute a predicate `U(P, p)` where `P` is a list of patterns (we refer to this as
20 //! a matrix). `U(P, p)` represents whether, given an existing list of patterns
21 //! `P_1 ..= P_m`, adding a new pattern `p` will be "useful" (that is, cover previously-
22 //! uncovered values of the type).
24 //! If we have this predicate, then we can easily compute both exhaustiveness of an
25 //! entire set of patterns and the individual usefulness of each one.
26 //! (a) the set of patterns is exhaustive iff `U(P, _)` is false (i.e., adding a wildcard
27 //! match doesn't increase the number of values we're matching)
28 //! (b) a pattern `P_i` is not useful if `U(P[0..=(i-1), P_i)` is false (i.e., adding a
29 //! pattern to those that have come before it doesn't increase the number of values
34 //! The idea that powers everything that is done in this file is the following: a value is made
35 //! from a constructor applied to some fields. Examples of constructors are `Some`, `None`, `(,)`
36 //! (the 2-tuple constructor), `Foo {..}` (the constructor for a struct `Foo`), and `2` (the
37 //! constructor for the number `2`). Fields are just a (possibly empty) list of values.
39 //! Some of the constructors listed above might feel weird: `None` and `2` don't take any
40 //! arguments. This is part of what makes constructors so general: we will consider plain values
41 //! like numbers and string literals to be constructors that take no arguments, also called "0-ary
42 //! constructors"; they are the simplest case of constructors. This allows us to see any value as
43 //! made up from a tree of constructors, each having a given number of children. For example:
44 //! `(None, Ok(0))` is made from 4 different constructors.
46 //! This idea can be extended to patterns: a pattern captures a set of possible values, and we can
47 //! describe this set using constructors. For example, `Err(_)` captures all values of the type
48 //! `Result<T, E>` that start with the `Err` constructor (for some choice of `T` and `E`). The
49 //! wildcard `_` captures all values of the given type starting with any of the constructors for
52 //! We use this to compute whether different patterns might capture a same value. Do the patterns
53 //! `Ok("foo")` and `Err(_)` capture a common value? The answer is no, because the first pattern
54 //! captures only values starting with the `Ok` constructor and the second only values starting
55 //! with the `Err` constructor. Do the patterns `Some(42)` and `Some(1..10)` intersect? They might,
56 //! since they both capture values starting with `Some`. To be certain, we need to dig under the
57 //! `Some` constructor and continue asking the question. This is the main idea behind the
58 //! exhaustiveness algorithm: by looking at patterns constructor-by-constructor, we can efficiently
59 //! figure out if some new pattern might capture a value that hadn't been captured by previous
62 //! Constructors are represented by the `Constructor` enum, and its fields by the `Fields` enum.
63 //! Most of the complexity of this file resides in transforming between patterns and
64 //! (`Constructor`, `Fields`) pairs, handling all the special cases correctly.
66 //! Caveat: this constructors/fields distinction doesn't quite cover every Rust value. For example
67 //! a value of type `Rc<u64>` doesn't fit this idea very well, nor do various other things.
68 //! However, this idea covers most of the cases that are relevant to exhaustiveness checking.
73 //! Recall that `U(P, p)` represents whether, given an existing list of patterns (aka matrix) `P`,
74 //! adding a new pattern `p` will cover previously-uncovered values of the type.
75 //! During the course of the algorithm, the rows of the matrix won't just be individual patterns,
76 //! but rather partially-deconstructed patterns in the form of a list of fields. The paper
77 //! calls those pattern-vectors, and we will call them pattern-stacks. The same holds for the
80 //! For example, say we have the following:
83 //! // x: (Option<bool>, Result<()>)
85 //! (Some(true), _) => {}
86 //! (None, Err(())) => {}
87 //! (None, Err(_)) => {}
91 //! Here, the matrix `P` starts as:
95 //! [(Some(true), _)],
96 //! [(None, Err(()))],
101 //! We can tell it's not exhaustive, because `U(P, _)` is true (we're not covering
102 //! `[(Some(false), _)]`, for instance). In addition, row 3 is not useful, because
103 //! all the values it covers are already covered by row 2.
105 //! A list of patterns can be thought of as a stack, because we are mainly interested in the top of
106 //! the stack at any given point, and we can pop or apply constructors to get new pattern-stacks.
107 //! To match the paper, the top of the stack is at the beginning / on the left.
109 //! There are two important operations on pattern-stacks necessary to understand the algorithm:
111 //! 1. We can pop a given constructor off the top of a stack. This operation is called
112 //! `specialize`, and is denoted `S(c, p)` where `c` is a constructor (like `Some` or
113 //! `None`) and `p` a pattern-stack.
114 //! If the pattern on top of the stack can cover `c`, this removes the constructor and
115 //! pushes its arguments onto the stack. It also expands OR-patterns into distinct patterns.
116 //! Otherwise the pattern-stack is discarded.
117 //! This essentially filters those pattern-stacks whose top covers the constructor `c` and
118 //! discards the others.
120 //! For example, the first pattern above initially gives a stack `[(Some(true), _)]`. If we
121 //! pop the tuple constructor, we are left with `[Some(true), _]`, and if we then pop the
122 //! `Some` constructor we get `[true, _]`. If we had popped `None` instead, we would get
125 //! This returns zero or more new pattern-stacks, as follows. We look at the pattern `p_1`
126 //! on top of the stack, and we have four cases:
127 //! 1.1. `p_1 = c(r_1, .., r_a)`, i.e. the top of the stack has constructor `c`. We
128 //! push onto the stack the arguments of this constructor, and return the result:
129 //! r_1, .., r_a, p_2, .., p_n
130 //! 1.2. `p_1 = c'(r_1, .., r_a')` where `c ≠ c'`. We discard the current stack and
132 //! 1.3. `p_1 = _`. We push onto the stack as many wildcards as the constructor `c` has
133 //! arguments (its arity), and return the resulting stack:
134 //! _, .., _, p_2, .., p_n
135 //! 1.4. `p_1 = r_1 | r_2`. We expand the OR-pattern and then recurse on each resulting
137 //! S(c, (r_1, p_2, .., p_n))
138 //! S(c, (r_2, p_2, .., p_n))
140 //! 2. We can pop a wildcard off the top of the stack. This is called `S(_, p)`, where `p` is
141 //! a pattern-stack. Note: the paper calls this `D(p)`.
142 //! This is used when we know there are missing constructor cases, but there might be
143 //! existing wildcard patterns, so to check the usefulness of the matrix, we have to check
144 //! all its *other* components.
146 //! It is computed as follows. We look at the pattern `p_1` on top of the stack,
147 //! and we have three cases:
148 //! 2.1. `p_1 = c(r_1, .., r_a)`. We discard the current stack and return nothing.
149 //! 2.2. `p_1 = _`. We return the rest of the stack:
151 //! 2.3. `p_1 = r_1 | r_2`. We expand the OR-pattern and then recurse on each resulting
153 //! S(_, (r_1, p_2, .., p_n))
154 //! S(_, (r_2, p_2, .., p_n))
156 //! Note that the OR-patterns are not always used directly in Rust, but are used to derive the
157 //! exhaustive integer matching rules, so they're written here for posterity.
159 //! Both those operations extend straightforwardly to a list or pattern-stacks, i.e. a matrix, by
160 //! working row-by-row. Popping a constructor ends up keeping only the matrix rows that start with
161 //! the given constructor, and popping a wildcard keeps those rows that start with a wildcard.
164 //! The algorithm for computing `U`
165 //! -------------------------------
166 //! The algorithm is inductive (on the number of columns: i.e., components of tuple patterns).
167 //! That means we're going to check the components from left-to-right, so the algorithm
168 //! operates principally on the first component of the matrix and new pattern-stack `p`.
169 //! This algorithm is realised in the `is_useful` function.
171 //! Base case. (`n = 0`, i.e., an empty tuple pattern)
172 //! - If `P` already contains an empty pattern (i.e., if the number of patterns `m > 0`),
173 //! then `U(P, p)` is false.
174 //! - Otherwise, `P` must be empty, so `U(P, p)` is true.
176 //! Inductive step. (`n > 0`, i.e., whether there's at least one column
177 //! [which may then be expanded into further columns later])
178 //! We're going to match on the top of the new pattern-stack, `p_1`.
179 //! - If `p_1 == c(r_1, .., r_a)`, i.e. we have a constructor pattern.
180 //! Then, the usefulness of `p_1` can be reduced to whether it is useful when
181 //! we ignore all the patterns in the first column of `P` that involve other constructors.
182 //! This is where `S(c, P)` comes in:
183 //! `U(P, p) := U(S(c, P), S(c, p))`
184 //! This special case is handled in `is_useful_specialized`.
186 //! For example, if `P` is:
195 //! and `p` is [Some(false), 0], then we don't care about row 2 since we know `p` only
196 //! matches values that row 2 doesn't. For row 1 however, we need to dig into the
197 //! arguments of `Some` to know whether some new value is covered. So we compute
198 //! `U([[true, _]], [false, 0])`.
200 //! - If `p_1 == _`, then we look at the list of constructors that appear in the first
201 //! component of the rows of `P`:
202 //! + If there are some constructors that aren't present, then we might think that the
203 //! wildcard `_` is useful, since it covers those constructors that weren't covered
205 //! That's almost correct, but only works if there were no wildcards in those first
206 //! components. So we need to check that `p` is useful with respect to the rows that
207 //! start with a wildcard, if there are any. This is where `D` comes in:
208 //! `U(P, p) := U(S(_, P), S(_, p))`
210 //! For example, if `P` is:
215 //! [None, false, 1],
219 //! and `p` is [_, false, _], the `Some` constructor doesn't appear in `P`. So if we
220 //! only had row 2, we'd know that `p` is useful. However row 1 starts with a
221 //! wildcard, so we need to check whether `U([[true, _]], [false, 1])`.
223 //! + Otherwise, all possible constructors (for the relevant type) are present. In this
224 //! case we must check whether the wildcard pattern covers any unmatched value. For
225 //! that, we can think of the `_` pattern as a big OR-pattern that covers all
226 //! possible constructors. For `Option`, that would mean `_ = None | Some(_)` for
227 //! example. The wildcard pattern is useful in this case if it is useful when
228 //! specialized to one of the possible constructors. So we compute:
229 //! `U(P, p) := ∃(k ϵ constructors) U(S(k, P), S(k, p))`
231 //! For example, if `P` is:
240 //! and `p` is [_, false], both `None` and `Some` constructors appear in the first
241 //! components of `P`. We will therefore try popping both constructors in turn: we
242 //! compute `U([[true, _]], [_, false])` for the `Some` constructor, and `U([[false]],
243 //! [false])` for the `None` constructor. The first case returns true, so we know that
244 //! `p` is useful for `P`. Indeed, it matches `[Some(false), _]` that wasn't matched
247 //! - If `p_1 == r_1 | r_2`, then the usefulness depends on each `r_i` separately:
248 //! `U(P, p) := U(P, (r_1, p_2, .., p_n))
249 //! || U(P, (r_2, p_2, .., p_n))`
251 //! Modifications to the algorithm
252 //! ------------------------------
253 //! The algorithm in the paper doesn't cover some of the special cases that arise in Rust, for
254 //! example uninhabited types and variable-length slice patterns. These are drawn attention to
255 //! throughout the code below. I'll make a quick note here about how exhaustive integer matching is
256 //! accounted for, though.
258 //! Exhaustive integer matching
259 //! ---------------------------
260 //! An integer type can be thought of as a (huge) sum type: 1 | 2 | 3 | ...
261 //! So to support exhaustive integer matching, we can make use of the logic in the paper for
262 //! OR-patterns. However, we obviously can't just treat ranges x..=y as individual sums, because
263 //! they are likely gigantic. So we instead treat ranges as constructors of the integers. This means
264 //! that we have a constructor *of* constructors (the integers themselves). We then need to work
265 //! through all the inductive step rules above, deriving how the ranges would be treated as
266 //! OR-patterns, and making sure that they're treated in the same way even when they're ranges.
267 //! There are really only four special cases here:
268 //! - When we match on a constructor that's actually a range, we have to treat it as if we would
270 //! + It turns out that we can simply extend the case for single-value patterns in
271 //! `specialize` to either be *equal* to a value constructor, or *contained within* a range
273 //! + When the pattern itself is a range, you just want to tell whether any of the values in
274 //! the pattern range coincide with values in the constructor range, which is precisely
276 //! Since when encountering a range pattern for a value constructor, we also use inclusion, it
277 //! means that whenever the constructor is a value/range and the pattern is also a value/range,
278 //! we can simply use intersection to test usefulness.
279 //! - When we're testing for usefulness of a pattern and the pattern's first component is a
281 //! + If all the constructors appear in the matrix, we have a slight complication. By default,
282 //! the behaviour (i.e., a disjunction over specialised matrices for each constructor) is
283 //! invalid, because we want a disjunction over every *integer* in each range, not just a
284 //! disjunction over every range. This is a bit more tricky to deal with: essentially we need
285 //! to form equivalence classes of subranges of the constructor range for which the behaviour
286 //! of the matrix `P` and new pattern `p` are the same. This is described in more
287 //! detail in `Constructor::split`.
288 //! + If some constructors are missing from the matrix, it turns out we don't need to do
289 //! anything special (because we know none of the integers are actually wildcards: i.e., we
290 //! can't span wildcards using ranges).
291 use self::Constructor::*;
292 use self::SliceKind::*;
293 use self::Usefulness::*;
294 use self::WitnessPreference::*;
296 use rustc_data_structures::captures::Captures;
297 use rustc_data_structures::fx::{FxHashMap, FxHashSet};
298 use rustc_data_structures::sync::OnceCell;
299 use rustc_index::vec::Idx;
301 use super::{compare_const_vals, PatternFoldable, PatternFolder};
302 use super::{FieldPat, Pat, PatKind, PatRange};
304 use rustc_arena::TypedArena;
305 use rustc_attr::{SignedInt, UnsignedInt};
306 use rustc_hir::def_id::DefId;
307 use rustc_hir::{HirId, RangeEnd};
308 use rustc_middle::mir::interpret::{truncate, ConstValue};
309 use rustc_middle::mir::Field;
310 use rustc_middle::ty::layout::IntegerExt;
311 use rustc_middle::ty::{self, Const, Ty, TyCtxt};
312 use rustc_session::lint;
313 use rustc_span::{Span, DUMMY_SP};
314 use rustc_target::abi::{Integer, Size, VariantIdx};
316 use smallvec::{smallvec, SmallVec};
317 use std::cmp::{self, max, min, Ordering};
319 use std::iter::{FromIterator, IntoIterator};
320 use std::ops::RangeInclusive;
322 crate fn expand_pattern<'tcx>(pat: Pat<'tcx>) -> Pat<'tcx> {
323 LiteralExpander.fold_pattern(&pat)
326 struct LiteralExpander;
328 impl<'tcx> PatternFolder<'tcx> for LiteralExpander {
329 fn fold_pattern(&mut self, pat: &Pat<'tcx>) -> Pat<'tcx> {
330 debug!("fold_pattern {:?} {:?} {:?}", pat, pat.ty.kind(), pat.kind);
331 match (pat.ty.kind(), &*pat.kind) {
332 (_, &PatKind::Binding { subpattern: Some(ref s), .. }) => s.fold_with(self),
333 (_, &PatKind::AscribeUserType { subpattern: ref s, .. }) => s.fold_with(self),
334 _ => pat.super_fold_with(self),
339 impl<'tcx> Pat<'tcx> {
340 pub(super) fn is_wildcard(&self) -> bool {
342 PatKind::Binding { subpattern: None, .. } | PatKind::Wild => true,
348 /// A row of a matrix. Rows of len 1 are very common, which is why `SmallVec[_; 2]`
350 #[derive(Debug, Clone)]
351 crate struct PatStack<'p, 'tcx> {
352 pats: SmallVec<[&'p Pat<'tcx>; 2]>,
353 /// Cache for the constructor of the head
354 head_ctor: OnceCell<Constructor<'tcx>>,
357 impl<'p, 'tcx> PatStack<'p, 'tcx> {
358 crate fn from_pattern(pat: &'p Pat<'tcx>) -> Self {
359 Self::from_vec(smallvec![pat])
362 fn from_vec(vec: SmallVec<[&'p Pat<'tcx>; 2]>) -> Self {
363 PatStack { pats: vec, head_ctor: OnceCell::new() }
366 fn is_empty(&self) -> bool {
370 fn len(&self) -> usize {
374 fn head(&self) -> &'p Pat<'tcx> {
378 fn head_ctor<'a>(&'a self, cx: &MatchCheckCtxt<'p, 'tcx>) -> &'a Constructor<'tcx> {
379 self.head_ctor.get_or_init(|| pat_constructor(cx, self.head()))
382 fn iter(&self) -> impl Iterator<Item = &Pat<'tcx>> {
383 self.pats.iter().copied()
386 // If the first pattern is an or-pattern, expand this pattern. Otherwise, return `None`.
387 fn expand_or_pat(&self) -> Option<Vec<Self>> {
390 } else if let PatKind::Or { pats } = &*self.head().kind {
394 let mut new_patstack = PatStack::from_pattern(pat);
395 new_patstack.pats.extend_from_slice(&self.pats[1..]);
405 /// This computes `S(constructor, self)`. See top of the file for explanations.
407 /// This is the main specialization step. It expands the pattern
408 /// into `arity` patterns based on the constructor. For most patterns, the step is trivial,
409 /// for instance tuple patterns are flattened and box patterns expand into their inner pattern.
410 /// Returns `None` if the pattern does not have the given constructor.
412 /// OTOH, slice patterns with a subslice pattern (tail @ ..) can be expanded into multiple
413 /// different patterns.
414 /// Structure patterns with a partial wild pattern (Foo { a: 42, .. }) have their missing
415 /// fields filled with wild patterns.
417 /// This is roughly the inverse of `Constructor::apply`.
418 fn specialize_constructor(
420 cx: &MatchCheckCtxt<'p, 'tcx>,
421 ctor: &Constructor<'tcx>,
422 ctor_wild_subpatterns: &Fields<'p, 'tcx>,
423 is_my_head_ctor: bool,
424 ) -> Option<PatStack<'p, 'tcx>> {
425 // We return `None` if `ctor` is not covered by `self.head()`. If `ctor` is known to be
426 // derived from `self.head()`, then we don't need to check; otherwise, we check for
427 // constructor inclusion.
428 // Note that this shortcut is also necessary for correctness: a pattern should always be
429 // specializable with its own constructor, even in cases where we refuse to inspect values like
431 if !is_my_head_ctor && !ctor.is_covered_by(cx, self.head_ctor(cx), self.head().ty) {
434 let new_fields = ctor_wild_subpatterns.replace_with_pattern_arguments(self.head());
437 "specialize_constructor({:#?}, {:#?}, {:#?}) = {:#?}",
440 ctor_wild_subpatterns,
444 // We pop the head pattern and push the new fields extracted from the arguments of
446 Some(new_fields.push_on_patstack(&self.pats[1..]))
450 impl<'p, 'tcx> Default for PatStack<'p, 'tcx> {
451 fn default() -> Self {
452 Self::from_vec(smallvec![])
456 impl<'p, 'tcx> PartialEq for PatStack<'p, 'tcx> {
457 fn eq(&self, other: &Self) -> bool {
458 self.pats == other.pats
462 impl<'p, 'tcx> FromIterator<&'p Pat<'tcx>> for PatStack<'p, 'tcx> {
463 fn from_iter<T>(iter: T) -> Self
465 T: IntoIterator<Item = &'p Pat<'tcx>>,
467 Self::from_vec(iter.into_iter().collect())
471 /// Depending on the match patterns, the specialization process might be able to use a fast path.
472 /// Tracks whether we can use the fast path and the lookup table needed in those cases.
473 #[derive(Clone, Debug, PartialEq)]
474 enum SpecializationCache {
475 /// Patterns consist of only enum variants.
476 /// Variant patterns does not intersect with each other (in contrast to range patterns),
477 /// so it is possible to precompute the result of `Matrix::specialize_constructor` at a
478 /// lower computational complexity.
479 /// `lookup` is responsible for holding the precomputed result of
480 /// specialization, while `wilds` is used for two purposes: the first one is
481 /// the precomputed result of specialization with a wildcard, and the second is to be used as a
482 /// fallback for `Matrix::specialize_constructor` when it tries to apply a constructor that
483 /// has not been seen in the `Matrix`. See `update_cache` for further explanations.
484 Variants { lookup: FxHashMap<DefId, SmallVec<[usize; 1]>>, wilds: SmallVec<[usize; 1]> },
485 /// Does not belong to the cases above, use the slow path.
490 #[derive(Clone, PartialEq)]
491 crate struct Matrix<'p, 'tcx> {
492 patterns: Vec<PatStack<'p, 'tcx>>,
493 cache: SpecializationCache,
496 impl<'p, 'tcx> Matrix<'p, 'tcx> {
497 crate fn empty() -> Self {
498 // Use `SpecializationCache::Incompatible` as a placeholder; we will initialize it on the
499 // first call to `push`. See the first half of `update_cache`.
500 Matrix { patterns: vec![], cache: SpecializationCache::Incompatible }
503 /// Pushes a new row to the matrix. If the row starts with an or-pattern, this expands it.
504 crate fn push(&mut self, row: PatStack<'p, 'tcx>) {
505 if let Some(rows) = row.expand_or_pat() {
507 // We recursively expand the or-patterns of the new rows.
508 // This is necessary as we might have `0 | (1 | 2)` or e.g., `x @ 0 | x @ (1 | 2)`.
512 self.patterns.push(row);
513 self.update_cache(self.patterns.len() - 1);
517 fn update_cache(&mut self, idx: usize) {
518 let row = &self.patterns[idx];
519 // We don't know which kind of cache could be used until we see the first row; therefore an
520 // empty `Matrix` is initialized with `SpecializationCache::Empty`, then the cache is
521 // assigned the appropriate variant below on the first call to `push`.
522 if self.patterns.is_empty() {
523 self.cache = if row.is_empty() {
524 SpecializationCache::Incompatible
526 match *row.head().kind {
527 PatKind::Variant { .. } => SpecializationCache::Variants {
528 lookup: FxHashMap::default(),
529 wilds: SmallVec::new(),
531 // Note: If the first pattern is a wildcard, then all patterns after that is not
532 // useful. The check is simple enough so we treat it as the same as unsupported
534 _ => SpecializationCache::Incompatible,
539 match &mut self.cache {
540 SpecializationCache::Variants { ref mut lookup, ref mut wilds } => {
541 let head = row.head();
543 _ if head.is_wildcard() => {
544 // Per rule 1.3 in the top-level comments, a wildcard pattern is included in
545 // the result of `specialize_constructor` for *any* `Constructor`.
546 // We push the wildcard pattern to the precomputed result for constructors
547 // that we have seen before; results for constructors we have not yet seen
548 // defaults to `wilds`, which is updated right below.
549 for (_, v) in lookup.iter_mut() {
552 // Per rule 2.1 and 2.2 in the top-level comments, only wildcard patterns
553 // are included in the result of specialization with a wildcard.
554 // What we do here is to track the wildcards we have seen; so in addition to
555 // acting as the precomputed result of specialization with a wildcard, `wilds` also
556 // serves as the default value of `specialize_constructor` for constructors
557 // that are not in `lookup`.
560 PatKind::Variant { adt_def, variant_index, .. } => {
561 // Handle the cases of rule 1.1 and 1.2 in the top-level comments.
562 // A variant pattern can only be included in the results of
563 // `specialize_constructor` for a particular constructor, therefore we are
564 // using a HashMap to track that.
566 .entry(adt_def.variants[variant_index].def_id)
567 // Default to `wilds` for absent keys. See above for an explanation.
568 .or_insert_with(|| wilds.clone())
572 self.cache = SpecializationCache::Incompatible;
576 SpecializationCache::Incompatible => {}
580 /// Iterate over the first component of each row
581 fn heads<'a>(&'a self) -> impl Iterator<Item = &'a Pat<'tcx>> + Captures<'p> {
582 self.patterns.iter().map(|r| r.head())
585 /// Iterate over the first constructor of each row
588 cx: &'a MatchCheckCtxt<'p, 'tcx>,
589 ) -> impl Iterator<Item = &'a Constructor<'tcx>> + Captures<'a> + Captures<'p> {
590 self.patterns.iter().map(move |r| r.head_ctor(cx))
593 /// This computes `S(constructor, self)`. See top of the file for explanations.
594 fn specialize_constructor(
596 cx: &MatchCheckCtxt<'p, 'tcx>,
597 constructor: &Constructor<'tcx>,
598 ctor_wild_subpatterns: &Fields<'p, 'tcx>,
599 ) -> Matrix<'p, 'tcx> {
601 SpecializationCache::Variants { lookup, wilds } => {
602 let cached = if let Constructor::Variant(id) = constructor {
605 // Default to `wilds` for absent keys. See `update_cache` for an explanation.
607 } else if let Wildcard = constructor {
611 "unexpected constructor encountered while dealing with matrix cache: {:?}",
615 let result: Self = cached
618 self.patterns[i].specialize_constructor(
621 ctor_wild_subpatterns,
626 // When debug assertions are enabled, check the results against the "slow path"
631 patterns: self.patterns.clone(),
632 cache: SpecializationCache::Incompatible
634 .specialize_constructor(
637 ctor_wild_subpatterns
642 SpecializationCache::Incompatible => self
646 r.specialize_constructor(cx, constructor, ctor_wild_subpatterns, false)
653 /// Pretty-printer for matrices of patterns, example:
656 /// +++++++++++++++++++++++++++++
658 /// +++++++++++++++++++++++++++++
659 /// + true + [First] +
660 /// +++++++++++++++++++++++++++++
661 /// + true + [Second(true)] +
662 /// +++++++++++++++++++++++++++++
664 /// +++++++++++++++++++++++++++++
665 /// + _ + [_, _, tail @ ..] +
666 /// +++++++++++++++++++++++++++++
668 impl<'p, 'tcx> fmt::Debug for Matrix<'p, 'tcx> {
669 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
672 let Matrix { patterns: m, .. } = self;
673 let pretty_printed_matrix: Vec<Vec<String>> =
674 m.iter().map(|row| row.iter().map(|pat| format!("{:?}", pat)).collect()).collect();
676 let column_count = m.iter().map(|row| row.len()).max().unwrap_or(0);
677 assert!(m.iter().all(|row| row.len() == column_count));
678 let column_widths: Vec<usize> = (0..column_count)
679 .map(|col| pretty_printed_matrix.iter().map(|row| row[col].len()).max().unwrap_or(0))
682 let total_width = column_widths.iter().cloned().sum::<usize>() + column_count * 3 + 1;
683 let br = "+".repeat(total_width);
684 write!(f, "{}\n", br)?;
685 for row in pretty_printed_matrix {
687 for (column, pat_str) in row.into_iter().enumerate() {
689 write!(f, "{:1$}", pat_str, column_widths[column])?;
693 write!(f, "{}\n", br)?;
699 impl<'p, 'tcx> FromIterator<PatStack<'p, 'tcx>> for Matrix<'p, 'tcx> {
700 fn from_iter<T>(iter: T) -> Self
702 T: IntoIterator<Item = PatStack<'p, 'tcx>>,
704 let mut matrix = Matrix::empty();
706 // Using `push` ensures we correctly expand or-patterns.
713 crate struct MatchCheckCtxt<'a, 'tcx> {
714 crate tcx: TyCtxt<'tcx>,
715 /// The module in which the match occurs. This is necessary for
716 /// checking inhabited-ness of types because whether a type is (visibly)
717 /// inhabited can depend on whether it was defined in the current module or
718 /// not. E.g., `struct Foo { _private: ! }` cannot be seen to be empty
719 /// outside it's module and should not be matchable with an empty match
722 crate param_env: ty::ParamEnv<'tcx>,
723 crate pattern_arena: &'a TypedArena<Pat<'tcx>>,
726 impl<'a, 'tcx> MatchCheckCtxt<'a, 'tcx> {
727 fn is_uninhabited(&self, ty: Ty<'tcx>) -> bool {
728 if self.tcx.features().exhaustive_patterns {
729 self.tcx.is_ty_uninhabited_from(self.module, ty, self.param_env)
735 /// Returns whether the given type is an enum from another crate declared `#[non_exhaustive]`.
736 crate fn is_foreign_non_exhaustive_enum(&self, ty: Ty<'tcx>) -> bool {
738 ty::Adt(def, ..) => {
739 def.is_enum() && def.is_variant_list_non_exhaustive() && !def.did.is_local()
746 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
748 /// Patterns of length `n` (`[x, y]`).
750 /// Patterns using the `..` notation (`[x, .., y]`).
751 /// Captures any array constructor of `length >= i + j`.
752 /// In the case where `array_len` is `Some(_)`,
753 /// this indicates that we only care about the first `i` and the last `j` values of the array,
754 /// and everything in between is a wildcard `_`.
759 fn arity(self) -> u64 {
761 FixedLen(length) => length,
762 VarLen(prefix, suffix) => prefix + suffix,
766 /// Whether this pattern includes patterns of length `other_len`.
767 fn covers_length(self, other_len: u64) -> bool {
769 FixedLen(len) => len == other_len,
770 VarLen(prefix, suffix) => prefix + suffix <= other_len,
774 /// Returns a collection of slices that spans the values covered by `self`, subtracted by the
775 /// values covered by `other`: i.e., `self \ other` (in set notation).
776 fn subtract(self, other: Self) -> SmallVec<[Self; 1]> {
777 // Remember, `VarLen(i, j)` covers the union of `FixedLen` from `i + j` to infinity.
778 // Naming: we remove the "neg" constructors from the "pos" ones.
780 FixedLen(pos_len) => {
781 if other.covers_length(pos_len) {
787 VarLen(pos_prefix, pos_suffix) => {
788 let pos_len = pos_prefix + pos_suffix;
790 FixedLen(neg_len) => {
791 if neg_len < pos_len {
796 // We know that `neg_len + 1 >= pos_len >= pos_suffix`.
797 .chain(Some(VarLen(neg_len + 1 - pos_suffix, pos_suffix)))
801 VarLen(neg_prefix, neg_suffix) => {
802 let neg_len = neg_prefix + neg_suffix;
803 if neg_len <= pos_len {
806 (pos_len..neg_len).map(FixedLen).collect()
815 /// A constructor for array and slice patterns.
816 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
818 /// `None` if the matched value is a slice, `Some(n)` if it is an array of size `n`.
819 array_len: Option<u64>,
820 /// The kind of pattern it is: fixed-length `[x, y]` or variable length `[x, .., y]`.
825 /// Returns what patterns this constructor covers: either fixed-length patterns or
826 /// variable-length patterns.
827 fn pattern_kind(self) -> SliceKind {
829 Slice { array_len: Some(len), kind: VarLen(prefix, suffix) }
830 if prefix + suffix == len =>
838 /// Returns what values this constructor covers: either values of only one given length, or
839 /// values of length above a given length.
840 /// This is different from `pattern_kind()` because in some cases the pattern only takes into
841 /// account a subset of the entries of the array, but still only captures values of a given
843 fn value_kind(self) -> SliceKind {
845 Slice { array_len: Some(len), kind: VarLen(_, _) } => FixedLen(len),
850 fn arity(self) -> u64 {
851 self.pattern_kind().arity()
854 /// The exhaustiveness-checking paper does not include any details on
855 /// checking variable-length slice patterns. However, they are matched
856 /// by an infinite collection of fixed-length array patterns.
858 /// Checking the infinite set directly would take an infinite amount
859 /// of time. However, it turns out that for each finite set of
860 /// patterns `P`, all sufficiently large array lengths are equivalent:
862 /// Each slice `s` with a "sufficiently-large" length `l ≥ L` that applies
863 /// to exactly the subset `Pₜ` of `P` can be transformed to a slice
864 /// `sₘ` for each sufficiently-large length `m` that applies to exactly
865 /// the same subset of `P`.
867 /// Because of that, each witness for reachability-checking from one
868 /// of the sufficiently-large lengths can be transformed to an
869 /// equally-valid witness from any other length, so we only have
870 /// to check slice lengths from the "minimal sufficiently-large length"
873 /// Note that the fact that there is a *single* `sₘ` for each `m`
874 /// not depending on the specific pattern in `P` is important: if
875 /// you look at the pair of patterns
878 /// Then any slice of length ≥1 that matches one of these two
879 /// patterns can be trivially turned to a slice of any
880 /// other length ≥1 that matches them and vice-versa - for
881 /// but the slice from length 2 `[false, true]` that matches neither
882 /// of these patterns can't be turned to a slice from length 1 that
883 /// matches neither of these patterns, so we have to consider
884 /// slices from length 2 there.
886 /// Now, to see that that length exists and find it, observe that slice
887 /// patterns are either "fixed-length" patterns (`[_, _, _]`) or
888 /// "variable-length" patterns (`[_, .., _]`).
890 /// For fixed-length patterns, all slices with lengths *longer* than
891 /// the pattern's length have the same outcome (of not matching), so
892 /// as long as `L` is greater than the pattern's length we can pick
893 /// any `sₘ` from that length and get the same result.
895 /// For variable-length patterns, the situation is more complicated,
896 /// because as seen above the precise value of `sₘ` matters.
898 /// However, for each variable-length pattern `p` with a prefix of length
899 /// `plₚ` and suffix of length `slₚ`, only the first `plₚ` and the last
900 /// `slₚ` elements are examined.
902 /// Therefore, as long as `L` is positive (to avoid concerns about empty
903 /// types), all elements after the maximum prefix length and before
904 /// the maximum suffix length are not examined by any variable-length
905 /// pattern, and therefore can be added/removed without affecting
906 /// them - creating equivalent patterns from any sufficiently-large
909 /// Of course, if fixed-length patterns exist, we must be sure
910 /// that our length is large enough to miss them all, so
911 /// we can pick `L = max(max(FIXED_LEN)+1, max(PREFIX_LEN) + max(SUFFIX_LEN))`
913 /// for example, with the above pair of patterns, all elements
914 /// but the first and last can be added/removed, so any
915 /// witness of length ≥2 (say, `[false, false, true]`) can be
916 /// turned to a witness from any other length ≥2.
919 cx: &MatchCheckCtxt<'p, 'tcx>,
920 matrix: &Matrix<'p, 'tcx>,
921 ) -> SmallVec<[Constructor<'tcx>; 1]> {
922 let (array_len, self_prefix, self_suffix) = match self {
923 Slice { array_len, kind: VarLen(self_prefix, self_suffix) } => {
924 (array_len, self_prefix, self_suffix)
926 _ => return smallvec![Slice(self)],
929 let head_ctors = matrix.head_ctors(cx).filter(|c| !c.is_wildcard());
931 let mut max_prefix_len = self_prefix;
932 let mut max_suffix_len = self_suffix;
933 let mut max_fixed_len = 0;
935 for ctor in head_ctors {
936 if let Slice(slice) = ctor {
937 match slice.pattern_kind() {
939 max_fixed_len = cmp::max(max_fixed_len, len);
941 VarLen(prefix, suffix) => {
942 max_prefix_len = cmp::max(max_prefix_len, prefix);
943 max_suffix_len = cmp::max(max_suffix_len, suffix);
949 // For diagnostics, we keep the prefix and suffix lengths separate, so in the case
950 // where `max_fixed_len + 1` is the largest, we adapt `max_prefix_len` accordingly,
951 // so that `L = max_prefix_len + max_suffix_len`.
952 if max_fixed_len + 1 >= max_prefix_len + max_suffix_len {
953 // The subtraction can't overflow thanks to the above check.
954 // The new `max_prefix_len` is also guaranteed to be larger than its previous
956 max_prefix_len = max_fixed_len + 1 - max_suffix_len;
961 let kind = if max_prefix_len + max_suffix_len < len {
962 VarLen(max_prefix_len, max_suffix_len)
966 smallvec![Slice(Slice { array_len, kind })]
969 // `ctor` originally covered the range `(self_prefix +
970 // self_suffix..infinity)`. We now split it into two: lengths smaller than
971 // `max_prefix_len + max_suffix_len` are treated independently as
972 // fixed-lengths slices, and lengths above are captured by a final VarLen
974 let smaller_lengths =
975 (self_prefix + self_suffix..max_prefix_len + max_suffix_len).map(FixedLen);
976 let final_slice = VarLen(max_prefix_len, max_suffix_len);
978 .chain(Some(final_slice))
979 .map(|kind| Slice { array_len, kind })
987 /// A value can be decomposed into a constructor applied to some fields. This struct represents
988 /// the constructor. See also `Fields`.
990 /// `pat_constructor` retrieves the constructor corresponding to a pattern.
991 /// `specialize_constructor` returns the list of fields corresponding to a pattern, given a
992 /// constructor. `Constructor::apply` reconstructs the pattern from a pair of `Constructor` and
994 #[derive(Clone, Debug, PartialEq)]
995 enum Constructor<'tcx> {
996 /// The constructor for patterns that have a single constructor, like tuples, struct patterns
997 /// and fixed-length arrays.
1001 /// Ranges of integer literal values (`2`, `2..=5` or `2..5`).
1002 IntRange(IntRange<'tcx>),
1003 /// Ranges of floating-point literal values (`2.0..=5.2`).
1004 FloatRange(&'tcx ty::Const<'tcx>, &'tcx ty::Const<'tcx>, RangeEnd),
1005 /// String literals. Strings are not quite the same as `&[u8]` so we treat them separately.
1006 Str(&'tcx ty::Const<'tcx>),
1007 /// Array and slice patterns.
1009 /// Constants that must not be matched structurally. They are treated as black
1010 /// boxes for the purposes of exhaustiveness: we must not inspect them, and they
1011 /// don't count towards making a match exhaustive.
1013 /// Fake extra constructor for enums that aren't allowed to be matched exhaustively.
1015 /// Wildcard pattern.
1019 impl<'tcx> Constructor<'tcx> {
1020 fn is_wildcard(&self) -> bool {
1027 fn variant_index_for_adt(&self, adt: &'tcx ty::AdtDef) -> VariantIdx {
1029 Variant(id) => adt.variant_index_with_id(id),
1031 assert!(!adt.is_enum());
1034 _ => bug!("bad constructor {:?} for adt {:?}", self, adt),
1038 // Returns the set of constructors covered by `self` but not by
1039 // anything in `other_ctors`.
1040 fn subtract_ctors(&self, other_ctors: &Vec<Constructor<'tcx>>) -> Vec<Constructor<'tcx>> {
1041 if other_ctors.is_empty() {
1042 return vec![self.clone()];
1046 // Those constructors can only match themselves.
1047 Single | Variant(_) | Str(..) | FloatRange(..) => {
1048 if other_ctors.iter().any(|c| c == self) { vec![] } else { vec![self.clone()] }
1051 let mut other_slices = other_ctors
1053 .filter_map(|c: &Constructor<'_>| match c {
1054 Slice(slice) => Some(*slice),
1055 _ => bug!("bad slice pattern constructor {:?}", c),
1057 .map(Slice::value_kind);
1059 match slice.value_kind() {
1060 FixedLen(self_len) => {
1061 if other_slices.any(|other_slice| other_slice.covers_length(self_len)) {
1067 kind @ VarLen(..) => {
1068 let mut remaining_slices = vec![kind];
1070 // For each used slice, subtract from the current set of slices.
1071 for other_slice in other_slices {
1072 remaining_slices = remaining_slices
1074 .flat_map(|remaining_slice| remaining_slice.subtract(other_slice))
1077 // If the constructors that have been considered so far already cover
1078 // the entire range of `self`, no need to look at more constructors.
1079 if remaining_slices.is_empty() {
1086 .map(|kind| Slice { array_len: slice.array_len, kind })
1092 IntRange(self_range) => {
1093 let mut remaining_ranges = vec![self_range.clone()];
1094 for other_ctor in other_ctors {
1095 if let IntRange(other_range) = other_ctor {
1096 if other_range == self_range {
1097 // If the `self` range appears directly in a `match` arm, we can
1098 // eliminate it straight away.
1099 remaining_ranges = vec![];
1101 // Otherwise explicitly compute the remaining ranges.
1102 remaining_ranges = other_range.subtract_from(remaining_ranges);
1105 // If the ranges that have been considered so far already cover the entire
1106 // range of values, we can return early.
1107 if remaining_ranges.is_empty() {
1113 // Convert the ranges back into constructors.
1114 remaining_ranges.into_iter().map(IntRange).collect()
1116 // This constructor is never covered by anything else
1117 NonExhaustive => vec![NonExhaustive],
1118 Opaque => bug!("found unexpected opaque ctor in all_ctors"),
1119 Wildcard => bug!("found unexpected wildcard ctor in all_ctors"),
1123 /// Some constructors (namely IntRange and Slice) actually stand for a set of actual
1124 /// constructors (integers and fixed-sized slices). When specializing for these
1125 /// constructors, we want to be specialising for the actual underlying constructors.
1126 /// Naively, we would simply return the list of constructors they correspond to. We instead are
1127 /// more clever: if there are constructors that we know will behave the same wrt the current
1128 /// matrix, we keep them grouped. For example, all slices of a sufficiently large length
1129 /// will either be all useful or all non-useful with a given matrix.
1131 /// See the branches for details on how the splitting is done.
1133 /// This function may discard some irrelevant constructors if this preserves behavior and
1134 /// diagnostics. Eg. for the `_` case, we ignore the constructors already present in the
1135 /// matrix, unless all of them are.
1137 /// `hir_id` is `None` when we're evaluating the wildcard pattern. In that case we do not want
1138 /// to lint for overlapping ranges.
1141 cx: &MatchCheckCtxt<'p, 'tcx>,
1143 matrix: &Matrix<'p, 'tcx>,
1144 hir_id: Option<HirId>,
1145 ) -> SmallVec<[Self; 1]> {
1146 debug!("Constructor::split({:#?}, {:#?})", self, matrix);
1149 // Fast-track if the range is trivial. In particular, we don't do the overlapping
1151 IntRange(ctor_range)
1152 if ctor_range.treat_exhaustively(cx.tcx) && !ctor_range.is_singleton() =>
1154 ctor_range.split(cx, pcx, matrix, hir_id)
1156 Slice(slice @ Slice { kind: VarLen(..), .. }) => slice.split(cx, matrix),
1157 // Any other constructor can be used unchanged.
1158 _ => smallvec![self.clone()],
1162 /// Returns whether `self` is covered by `other`, ie whether `self` is a subset of `other`. For
1163 /// the simple cases, this is simply checking for equality. For the "grouped" constructors,
1164 /// this checks for inclusion.
1165 fn is_covered_by<'p>(
1167 cx: &MatchCheckCtxt<'p, 'tcx>,
1168 other: &Constructor<'tcx>,
1171 match (self, other) {
1172 // Wildcards cover anything
1173 (_, Wildcard) => true,
1174 // Wildcards are only covered by wildcards
1175 (Wildcard, _) => false,
1177 (Single, Single) => true,
1178 (Variant(self_id), Variant(other_id)) => self_id == other_id,
1180 (IntRange(self_range), IntRange(other_range)) => {
1181 if self_range.intersection(cx.tcx, other_range).is_some() {
1182 // Constructor splitting should ensure that all intersections we encounter
1183 // are actually inclusions.
1184 assert!(self_range.is_subrange(other_range));
1191 FloatRange(self_from, self_to, self_end),
1192 FloatRange(other_from, other_to, other_end),
1195 compare_const_vals(cx.tcx, self_to, other_to, cx.param_env, ty),
1196 compare_const_vals(cx.tcx, self_from, other_from, cx.param_env, ty),
1198 (Some(to), Some(from)) => {
1199 (from == Ordering::Greater || from == Ordering::Equal)
1200 && (to == Ordering::Less
1201 || (other_end == self_end && to == Ordering::Equal))
1206 (Str(self_val), Str(other_val)) => {
1207 // FIXME: there's probably a more direct way of comparing for equality
1208 match compare_const_vals(cx.tcx, self_val, other_val, cx.param_env, ty) {
1209 Some(comparison) => comparison == Ordering::Equal,
1214 (Slice(self_slice), Slice(other_slice)) => {
1215 other_slice.pattern_kind().covers_length(self_slice.arity())
1218 // We are trying to inspect an opaque constant. Thus we skip the row.
1219 (Opaque, _) | (_, Opaque) => false,
1220 // Only a wildcard pattern can match the special extra constructor.
1221 (NonExhaustive, _) => false,
1223 _ => bug!("trying to compare incompatible constructors {:?} and {:?}", self, other),
1227 /// Apply a constructor to a list of patterns, yielding a new pattern. `pats`
1228 /// must have as many elements as this constructor's arity.
1230 /// This is roughly the inverse of `specialize_constructor`.
1233 /// `self`: `Constructor::Single`
1234 /// `ty`: `(u32, u32, u32)`
1235 /// `pats`: `[10, 20, _]`
1236 /// returns `(10, 20, _)`
1238 /// `self`: `Constructor::Variant(Option::Some)`
1239 /// `ty`: `Option<bool>`
1240 /// `pats`: `[false]`
1241 /// returns `Some(false)`
1244 cx: &MatchCheckCtxt<'p, 'tcx>,
1246 fields: Fields<'p, 'tcx>,
1248 let mut subpatterns = fields.all_patterns();
1250 let pat = match self {
1251 Single | Variant(_) => match ty.kind() {
1252 ty::Adt(..) | ty::Tuple(..) => {
1253 let subpatterns = subpatterns
1255 .map(|(i, p)| FieldPat { field: Field::new(i), pattern: p })
1258 if let ty::Adt(adt, substs) = ty.kind() {
1263 variant_index: self.variant_index_for_adt(adt),
1267 PatKind::Leaf { subpatterns }
1270 PatKind::Leaf { subpatterns }
1273 ty::Ref(..) => PatKind::Deref { subpattern: subpatterns.next().unwrap() },
1274 ty::Slice(_) | ty::Array(..) => bug!("bad slice pattern {:?} {:?}", self, ty),
1277 Slice(slice) => match slice.pattern_kind() {
1279 PatKind::Slice { prefix: subpatterns.collect(), slice: None, suffix: vec![] }
1281 VarLen(prefix, _) => {
1282 let mut prefix: Vec<_> = subpatterns.by_ref().take(prefix as usize).collect();
1283 if slice.array_len.is_some() {
1284 // Improves diagnostics a bit: if the type is a known-size array, instead
1285 // of reporting `[x, _, .., _, y]`, we prefer to report `[x, .., y]`.
1286 // This is incorrect if the size is not known, since `[_, ..]` captures
1287 // arrays of lengths `>= 1` whereas `[..]` captures any length.
1288 while !prefix.is_empty() && prefix.last().unwrap().is_wildcard() {
1292 let suffix: Vec<_> = if slice.array_len.is_some() {
1294 subpatterns.skip_while(Pat::is_wildcard).collect()
1296 subpatterns.collect()
1298 let wild = Pat::wildcard_from_ty(ty);
1299 PatKind::Slice { prefix, slice: Some(wild), suffix }
1302 &Str(value) => PatKind::Constant { value },
1303 &FloatRange(lo, hi, end) => PatKind::Range(PatRange { lo, hi, end }),
1304 IntRange(range) => return range.to_pat(cx.tcx),
1305 NonExhaustive => PatKind::Wild,
1306 Opaque => bug!("we should not try to apply an opaque constructor"),
1307 Wildcard => bug!("we should not try to apply a wildcard constructor"),
1310 Pat { ty, span: DUMMY_SP, kind: Box::new(pat) }
1313 /// Like `apply`, but where all the subpatterns are wildcards `_`.
1314 fn apply_wildcards<'a>(&self, cx: &MatchCheckCtxt<'a, 'tcx>, ty: Ty<'tcx>) -> Pat<'tcx> {
1315 self.apply(cx, ty, Fields::wildcards(cx, self, ty))
1319 /// Some fields need to be explicitly hidden away in certain cases; see the comment above the
1320 /// `Fields` struct. This struct represents such a potentially-hidden field. When a field is hidden
1321 /// we still keep its type around.
1322 #[derive(Debug, Copy, Clone)]
1323 enum FilteredField<'p, 'tcx> {
1324 Kept(&'p Pat<'tcx>),
1328 impl<'p, 'tcx> FilteredField<'p, 'tcx> {
1329 fn kept(self) -> Option<&'p Pat<'tcx>> {
1331 FilteredField::Kept(p) => Some(p),
1332 FilteredField::Hidden(_) => None,
1336 fn to_pattern(self) -> Pat<'tcx> {
1338 FilteredField::Kept(p) => p.clone(),
1339 FilteredField::Hidden(ty) => Pat::wildcard_from_ty(ty),
1344 /// A value can be decomposed into a constructor applied to some fields. This struct represents
1345 /// those fields, generalized to allow patterns in each field. See also `Constructor`.
1347 /// If a private or `non_exhaustive` field is uninhabited, the code mustn't observe that it is
1348 /// uninhabited. For that, we filter these fields out of the matrix. This is subtle because we
1349 /// still need to have those fields back when going to/from a `Pat`. Most of this is handled
1350 /// automatically in `Fields`, but when constructing or deconstructing `Fields` you need to be
1351 /// careful. As a rule, when going to/from the matrix, use the filtered field list; when going
1352 /// to/from `Pat`, use the full field list.
1353 /// This filtering is uncommon in practice, because uninhabited fields are rarely used, so we avoid
1354 /// it when possible to preserve performance.
1355 #[derive(Debug, Clone)]
1356 enum Fields<'p, 'tcx> {
1357 /// Lists of patterns that don't contain any filtered fields.
1358 /// `Slice` and `Vec` behave the same; the difference is only to avoid allocating and
1359 /// triple-dereferences when possible. Frankly this is premature optimization, I (Nadrieril)
1360 /// have not measured if it really made a difference.
1361 Slice(&'p [Pat<'tcx>]),
1362 Vec(SmallVec<[&'p Pat<'tcx>; 2]>),
1363 /// Patterns where some of the fields need to be hidden. `kept_count` caches the number of
1364 /// non-hidden fields.
1366 fields: SmallVec<[FilteredField<'p, 'tcx>; 2]>,
1371 impl<'p, 'tcx> Fields<'p, 'tcx> {
1372 fn empty() -> Self {
1376 /// Construct a new `Fields` from the given pattern. Must not be used if the pattern is a field
1377 /// of a struct/tuple/variant.
1378 fn from_single_pattern(pat: &'p Pat<'tcx>) -> Self {
1379 Fields::Slice(std::slice::from_ref(pat))
1382 /// Convenience; internal use.
1383 fn wildcards_from_tys(
1384 cx: &MatchCheckCtxt<'p, 'tcx>,
1385 tys: impl IntoIterator<Item = Ty<'tcx>>,
1387 let wilds = tys.into_iter().map(Pat::wildcard_from_ty);
1388 let pats = cx.pattern_arena.alloc_from_iter(wilds);
1392 /// Creates a new list of wildcard fields for a given constructor.
1394 cx: &MatchCheckCtxt<'p, 'tcx>,
1395 constructor: &Constructor<'tcx>,
1398 let wildcard_from_ty = |ty| &*cx.pattern_arena.alloc(Pat::wildcard_from_ty(ty));
1400 let ret = match constructor {
1401 Single | Variant(_) => match ty.kind() {
1402 ty::Tuple(ref fs) => {
1403 Fields::wildcards_from_tys(cx, fs.into_iter().map(|ty| ty.expect_ty()))
1405 ty::Ref(_, rty, _) => Fields::from_single_pattern(wildcard_from_ty(rty)),
1406 ty::Adt(adt, substs) => {
1408 // Use T as the sub pattern type of Box<T>.
1409 Fields::from_single_pattern(wildcard_from_ty(substs.type_at(0)))
1411 let variant = &adt.variants[constructor.variant_index_for_adt(adt)];
1412 // Whether we must not match the fields of this variant exhaustively.
1413 let is_non_exhaustive =
1414 variant.is_field_list_non_exhaustive() && !adt.did.is_local();
1415 let field_tys = variant.fields.iter().map(|field| field.ty(cx.tcx, substs));
1416 // In the following cases, we don't need to filter out any fields. This is
1417 // the vast majority of real cases, since uninhabited fields are uncommon.
1418 let has_no_hidden_fields = (adt.is_enum() && !is_non_exhaustive)
1419 || !field_tys.clone().any(|ty| cx.is_uninhabited(ty));
1421 if has_no_hidden_fields {
1422 Fields::wildcards_from_tys(cx, field_tys)
1424 let mut kept_count = 0;
1425 let fields = variant
1429 let ty = field.ty(cx.tcx, substs);
1430 let is_visible = adt.is_enum()
1431 || field.vis.is_accessible_from(cx.module, cx.tcx);
1432 let is_uninhabited = cx.is_uninhabited(ty);
1434 // In the cases of either a `#[non_exhaustive]` field list
1435 // or a non-public field, we hide uninhabited fields in
1436 // order not to reveal the uninhabitedness of the whole
1438 if is_uninhabited && (!is_visible || is_non_exhaustive) {
1439 FilteredField::Hidden(ty)
1442 FilteredField::Kept(wildcard_from_ty(ty))
1446 Fields::Filtered { fields, kept_count }
1450 _ => Fields::empty(),
1452 Slice(slice) => match *ty.kind() {
1453 ty::Slice(ty) | ty::Array(ty, _) => {
1454 let arity = slice.arity();
1455 Fields::wildcards_from_tys(cx, (0..arity).map(|_| ty))
1457 _ => bug!("bad slice pattern {:?} {:?}", constructor, ty),
1459 Str(..) | FloatRange(..) | IntRange(..) | NonExhaustive | Opaque | Wildcard => {
1463 debug!("Fields::wildcards({:?}, {:?}) = {:#?}", constructor, ty, ret);
1467 /// Returns the number of patterns from the viewpoint of match-checking, i.e. excluding hidden
1468 /// fields. This is what we want in most cases in this file, the only exception being
1469 /// conversion to/from `Pat`.
1470 fn len(&self) -> usize {
1472 Fields::Slice(pats) => pats.len(),
1473 Fields::Vec(pats) => pats.len(),
1474 Fields::Filtered { kept_count, .. } => *kept_count,
1478 /// Returns the complete list of patterns, including hidden fields.
1479 fn all_patterns(self) -> impl Iterator<Item = Pat<'tcx>> {
1480 let pats: SmallVec<[_; 2]> = match self {
1481 Fields::Slice(pats) => pats.iter().cloned().collect(),
1482 Fields::Vec(pats) => pats.into_iter().cloned().collect(),
1483 Fields::Filtered { fields, .. } => {
1484 // We don't skip any fields here.
1485 fields.into_iter().map(|p| p.to_pattern()).collect()
1491 /// Overrides some of the fields with the provided patterns. Exactly like
1492 /// `replace_fields_indexed`, except that it takes `FieldPat`s as input.
1493 fn replace_with_fieldpats(
1495 new_pats: impl IntoIterator<Item = &'p FieldPat<'tcx>>,
1497 self.replace_fields_indexed(
1498 new_pats.into_iter().map(|pat| (pat.field.index(), &pat.pattern)),
1502 /// Overrides some of the fields with the provided patterns. This is used when a pattern
1503 /// defines some fields but not all, for example `Foo { field1: Some(_), .. }`: here we start with a
1504 /// `Fields` that is just one wildcard per field of the `Foo` struct, and override the entry
1505 /// corresponding to `field1` with the pattern `Some(_)`. This is also used for slice patterns
1506 /// for the same reason.
1507 fn replace_fields_indexed(
1509 new_pats: impl IntoIterator<Item = (usize, &'p Pat<'tcx>)>,
1511 let mut fields = self.clone();
1512 if let Fields::Slice(pats) = fields {
1513 fields = Fields::Vec(pats.iter().collect());
1517 Fields::Vec(pats) => {
1518 for (i, pat) in new_pats {
1522 Fields::Filtered { fields, .. } => {
1523 for (i, pat) in new_pats {
1524 if let FilteredField::Kept(p) = &mut fields[i] {
1529 Fields::Slice(_) => unreachable!(),
1534 /// Replaces contained fields with the given filtered list of patterns, e.g. taken from the
1535 /// matrix. There must be `len()` patterns in `pats`.
1538 cx: &MatchCheckCtxt<'p, 'tcx>,
1539 pats: impl IntoIterator<Item = Pat<'tcx>>,
1541 let pats: &[_] = cx.pattern_arena.alloc_from_iter(pats);
1544 Fields::Filtered { fields, kept_count } => {
1545 let mut pats = pats.iter();
1546 let mut fields = fields.clone();
1547 for f in &mut fields {
1548 if let FilteredField::Kept(p) = f {
1549 // We take one input pattern for each `Kept` field, in order.
1550 *p = pats.next().unwrap();
1553 Fields::Filtered { fields, kept_count: *kept_count }
1555 _ => Fields::Slice(pats),
1559 /// Replaces contained fields with the arguments of the given pattern. Only use on a pattern
1560 /// that is compatible with the constructor used to build `self`.
1561 /// This is meant to be used on the result of `Fields::wildcards()`. The idea is that
1562 /// `wildcards` constructs a list of fields where all entries are wildcards, and the pattern
1563 /// provided to this function fills some of the fields with non-wildcards.
1564 /// In the following example `Fields::wildcards` would return `[_, _, _, _]`. If we call
1565 /// `replace_with_pattern_arguments` on it with the pattern, the result will be `[Some(0), _,
1568 /// let x: [Option<u8>; 4] = foo();
1570 /// [Some(0), ..] => {}
1573 fn replace_with_pattern_arguments(&self, pat: &'p Pat<'tcx>) -> Self {
1574 match pat.kind.as_ref() {
1575 PatKind::Deref { subpattern } => Self::from_single_pattern(subpattern),
1576 PatKind::Leaf { subpatterns } | PatKind::Variant { subpatterns, .. } => {
1577 self.replace_with_fieldpats(subpatterns)
1579 PatKind::Array { prefix, suffix, .. } | PatKind::Slice { prefix, suffix, .. } => {
1580 // Number of subpatterns for the constructor
1581 let ctor_arity = self.len();
1583 // Replace the prefix and the suffix with the given patterns, leaving wildcards in
1584 // the middle if there was a subslice pattern `..`.
1585 let prefix = prefix.iter().enumerate();
1587 suffix.iter().enumerate().map(|(i, p)| (ctor_arity - suffix.len() + i, p));
1588 self.replace_fields_indexed(prefix.chain(suffix))
1594 fn push_on_patstack(self, stack: &[&'p Pat<'tcx>]) -> PatStack<'p, 'tcx> {
1595 let pats: SmallVec<_> = match self {
1596 Fields::Slice(pats) => pats.iter().chain(stack.iter().copied()).collect(),
1597 Fields::Vec(mut pats) => {
1598 pats.extend_from_slice(stack);
1601 Fields::Filtered { fields, .. } => {
1602 // We skip hidden fields here
1603 fields.into_iter().filter_map(|p| p.kept()).chain(stack.iter().copied()).collect()
1606 PatStack::from_vec(pats)
1610 #[derive(Clone, Debug)]
1611 crate enum Usefulness<'tcx> {
1612 /// Carries a list of unreachable subpatterns. Used only in the presence of or-patterns.
1614 /// Carries a list of witnesses of non-exhaustiveness.
1615 UsefulWithWitness(Vec<Witness<'tcx>>),
1619 impl<'tcx> Usefulness<'tcx> {
1620 fn new_useful(preference: WitnessPreference) -> Self {
1622 ConstructWitness => UsefulWithWitness(vec![Witness(vec![])]),
1623 LeaveOutWitness => Useful(vec![]),
1627 fn is_useful(&self) -> bool {
1634 fn apply_constructor<'p>(
1636 cx: &MatchCheckCtxt<'p, 'tcx>,
1637 ctor: &Constructor<'tcx>,
1639 ctor_wild_subpatterns: &Fields<'p, 'tcx>,
1642 UsefulWithWitness(witnesses) => UsefulWithWitness(
1645 .map(|witness| witness.apply_constructor(cx, &ctor, ty, ctor_wild_subpatterns))
1652 fn apply_wildcard(self, ty: Ty<'tcx>) -> Self {
1654 UsefulWithWitness(witnesses) => {
1655 let wild = Pat::wildcard_from_ty(ty);
1659 .map(|mut witness| {
1660 witness.0.push(wild.clone());
1670 fn apply_missing_ctors(
1672 cx: &MatchCheckCtxt<'_, 'tcx>,
1674 missing_ctors: &MissingConstructors<'tcx>,
1677 UsefulWithWitness(witnesses) => {
1678 let new_patterns: Vec<_> =
1679 missing_ctors.iter().map(|ctor| ctor.apply_wildcards(cx, ty)).collect();
1680 // Add the new patterns to each witness
1684 .flat_map(|witness| {
1685 new_patterns.iter().map(move |pat| {
1686 let mut witness = witness.clone();
1687 witness.0.push(pat.clone());
1699 #[derive(Copy, Clone, Debug)]
1700 crate enum WitnessPreference {
1705 #[derive(Copy, Clone, Debug)]
1706 struct PatCtxt<'tcx> {
1711 /// A witness of non-exhaustiveness for error reporting, represented
1712 /// as a list of patterns (in reverse order of construction) with
1713 /// wildcards inside to represent elements that can take any inhabitant
1714 /// of the type as a value.
1716 /// A witness against a list of patterns should have the same types
1717 /// and length as the pattern matched against. Because Rust `match`
1718 /// is always against a single pattern, at the end the witness will
1719 /// have length 1, but in the middle of the algorithm, it can contain
1720 /// multiple patterns.
1722 /// For example, if we are constructing a witness for the match against
1725 /// struct Pair(Option<(u32, u32)>, bool);
1727 /// match (p: Pair) {
1728 /// Pair(None, _) => {}
1729 /// Pair(_, false) => {}
1733 /// We'll perform the following steps:
1734 /// 1. Start with an empty witness
1735 /// `Witness(vec![])`
1736 /// 2. Push a witness `Some(_)` against the `None`
1737 /// `Witness(vec![Some(_)])`
1738 /// 3. Push a witness `true` against the `false`
1739 /// `Witness(vec![Some(_), true])`
1740 /// 4. Apply the `Pair` constructor to the witnesses
1741 /// `Witness(vec![Pair(Some(_), true)])`
1743 /// The final `Pair(Some(_), true)` is then the resulting witness.
1744 #[derive(Clone, Debug)]
1745 crate struct Witness<'tcx>(Vec<Pat<'tcx>>);
1747 impl<'tcx> Witness<'tcx> {
1748 crate fn single_pattern(self) -> Pat<'tcx> {
1749 assert_eq!(self.0.len(), 1);
1750 self.0.into_iter().next().unwrap()
1753 /// Constructs a partial witness for a pattern given a list of
1754 /// patterns expanded by the specialization step.
1756 /// When a pattern P is discovered to be useful, this function is used bottom-up
1757 /// to reconstruct a complete witness, e.g., a pattern P' that covers a subset
1758 /// of values, V, where each value in that set is not covered by any previously
1759 /// used patterns and is covered by the pattern P'. Examples:
1761 /// left_ty: tuple of 3 elements
1762 /// pats: [10, 20, _] => (10, 20, _)
1764 /// left_ty: struct X { a: (bool, &'static str), b: usize}
1765 /// pats: [(false, "foo"), 42] => X { a: (false, "foo"), b: 42 }
1766 fn apply_constructor<'p>(
1768 cx: &MatchCheckCtxt<'p, 'tcx>,
1769 ctor: &Constructor<'tcx>,
1771 ctor_wild_subpatterns: &Fields<'p, 'tcx>,
1774 let len = self.0.len();
1775 let arity = ctor_wild_subpatterns.len();
1776 let pats = self.0.drain((len - arity)..).rev();
1777 let fields = ctor_wild_subpatterns.replace_fields(cx, pats);
1778 ctor.apply(cx, ty, fields)
1787 /// This determines the set of all possible constructors of a pattern matching
1788 /// values of type `left_ty`. For vectors, this would normally be an infinite set
1789 /// but is instead bounded by the maximum fixed length of slice patterns in
1790 /// the column of patterns being analyzed.
1792 /// We make sure to omit constructors that are statically impossible. E.g., for
1793 /// `Option<!>`, we do not include `Some(_)` in the returned list of constructors.
1794 /// Invariant: this returns an empty `Vec` if and only if the type is uninhabited (as determined by
1795 /// `cx.is_uninhabited()`).
1796 fn all_constructors<'a, 'tcx>(
1797 cx: &MatchCheckCtxt<'a, 'tcx>,
1799 ) -> Vec<Constructor<'tcx>> {
1800 debug!("all_constructors({:?})", pcx.ty);
1801 let make_range = |start, end| {
1803 // `unwrap()` is ok because we know the type is an integer.
1804 IntRange::from_range(cx.tcx, start, end, pcx.ty, &RangeEnd::Included, pcx.span)
1808 match *pcx.ty.kind() {
1809 ty::Bool => vec![make_range(0, 1)],
1810 ty::Array(ref sub_ty, len) if len.try_eval_usize(cx.tcx, cx.param_env).is_some() => {
1811 let len = len.eval_usize(cx.tcx, cx.param_env);
1812 if len != 0 && cx.is_uninhabited(sub_ty) {
1815 vec![Slice(Slice { array_len: Some(len), kind: VarLen(0, 0) })]
1818 // Treat arrays of a constant but unknown length like slices.
1819 ty::Array(ref sub_ty, _) | ty::Slice(ref sub_ty) => {
1820 let kind = if cx.is_uninhabited(sub_ty) { FixedLen(0) } else { VarLen(0, 0) };
1821 vec![Slice(Slice { array_len: None, kind })]
1823 ty::Adt(def, substs) if def.is_enum() => {
1824 let ctors: Vec<_> = if cx.tcx.features().exhaustive_patterns {
1825 // If `exhaustive_patterns` is enabled, we exclude variants known to be
1830 !v.uninhabited_from(cx.tcx, substs, def.adt_kind(), cx.param_env)
1831 .contains(cx.tcx, cx.module)
1833 .map(|v| Variant(v.def_id))
1836 def.variants.iter().map(|v| Variant(v.def_id)).collect()
1839 // If the enum is declared as `#[non_exhaustive]`, we treat it as if it had an
1840 // additional "unknown" constructor.
1841 // There is no point in enumerating all possible variants, because the user can't
1842 // actually match against them all themselves. So we always return only the fictitious
1844 // E.g., in an example like:
1847 // let err: io::ErrorKind = ...;
1849 // io::ErrorKind::NotFound => {},
1853 // we don't want to show every possible IO error, but instead have only `_` as the
1855 let is_declared_nonexhaustive = cx.is_foreign_non_exhaustive_enum(pcx.ty);
1857 // If `exhaustive_patterns` is disabled and our scrutinee is an empty enum, we treat it
1858 // as though it had an "unknown" constructor to avoid exposing its emptyness. Note that
1859 // an empty match will still be considered exhaustive because that case is handled
1860 // separately in `check_match`.
1861 let is_secretly_empty =
1862 def.variants.is_empty() && !cx.tcx.features().exhaustive_patterns;
1864 if is_secretly_empty || is_declared_nonexhaustive { vec![NonExhaustive] } else { ctors }
1868 // The valid Unicode Scalar Value ranges.
1869 make_range('\u{0000}' as u128, '\u{D7FF}' as u128),
1870 make_range('\u{E000}' as u128, '\u{10FFFF}' as u128),
1873 ty::Int(_) | ty::Uint(_)
1874 if pcx.ty.is_ptr_sized_integral()
1875 && !cx.tcx.features().precise_pointer_size_matching =>
1877 // `usize`/`isize` are not allowed to be matched exhaustively unless the
1878 // `precise_pointer_size_matching` feature is enabled. So we treat those types like
1879 // `#[non_exhaustive]` enums by returning a special unmatcheable constructor.
1883 let bits = Integer::from_attr(&cx.tcx, SignedInt(ity)).size().bits() as u128;
1884 let min = 1u128 << (bits - 1);
1886 vec![make_range(min, max)]
1889 let size = Integer::from_attr(&cx.tcx, UnsignedInt(uty)).size();
1890 let max = truncate(u128::MAX, size);
1891 vec![make_range(0, max)]
1894 if cx.is_uninhabited(pcx.ty) {
1903 /// An inclusive interval, used for precise integer exhaustiveness checking.
1904 /// `IntRange`s always store a contiguous range. This means that values are
1905 /// encoded such that `0` encodes the minimum value for the integer,
1906 /// regardless of the signedness.
1907 /// For example, the pattern `-128..=127i8` is encoded as `0..=255`.
1908 /// This makes comparisons and arithmetic on interval endpoints much more
1909 /// straightforward. See `signed_bias` for details.
1911 /// `IntRange` is never used to encode an empty range or a "range" that wraps
1912 /// around the (offset) space: i.e., `range.lo <= range.hi`.
1913 #[derive(Clone, Debug)]
1914 struct IntRange<'tcx> {
1915 range: RangeInclusive<u128>,
1920 impl<'tcx> IntRange<'tcx> {
1922 fn is_integral(ty: Ty<'_>) -> bool {
1924 ty::Char | ty::Int(_) | ty::Uint(_) | ty::Bool => true,
1929 fn is_singleton(&self) -> bool {
1930 self.range.start() == self.range.end()
1933 fn boundaries(&self) -> (u128, u128) {
1934 (*self.range.start(), *self.range.end())
1937 /// Don't treat `usize`/`isize` exhaustively unless the `precise_pointer_size_matching` feature
1939 fn treat_exhaustively(&self, tcx: TyCtxt<'tcx>) -> bool {
1940 !self.ty.is_ptr_sized_integral() || tcx.features().precise_pointer_size_matching
1944 fn integral_size_and_signed_bias(tcx: TyCtxt<'tcx>, ty: Ty<'_>) -> Option<(Size, u128)> {
1946 ty::Bool => Some((Size::from_bytes(1), 0)),
1947 ty::Char => Some((Size::from_bytes(4), 0)),
1949 let size = Integer::from_attr(&tcx, SignedInt(ity)).size();
1950 Some((size, 1u128 << (size.bits() as u128 - 1)))
1952 ty::Uint(uty) => Some((Integer::from_attr(&tcx, UnsignedInt(uty)).size(), 0)),
1960 param_env: ty::ParamEnv<'tcx>,
1961 value: &Const<'tcx>,
1963 ) -> Option<IntRange<'tcx>> {
1964 if let Some((target_size, bias)) = Self::integral_size_and_signed_bias(tcx, value.ty) {
1967 if let ty::ConstKind::Value(ConstValue::Scalar(scalar)) = value.val {
1968 // For this specific pattern we can skip a lot of effort and go
1969 // straight to the result, after doing a bit of checking. (We
1970 // could remove this branch and just fall through, which
1971 // is more general but much slower.)
1972 if let Ok(bits) = scalar.to_bits_or_ptr(target_size, &tcx) {
1976 // This is a more general form of the previous case.
1977 value.try_eval_bits(tcx, param_env, ty)
1979 let val = val ^ bias;
1980 Some(IntRange { range: val..=val, ty, span })
1994 ) -> Option<IntRange<'tcx>> {
1995 if Self::is_integral(ty) {
1996 // Perform a shift if the underlying types are signed,
1997 // which makes the interval arithmetic simpler.
1998 let bias = IntRange::signed_bias(tcx, ty);
1999 let (lo, hi) = (lo ^ bias, hi ^ bias);
2000 let offset = (*end == RangeEnd::Excluded) as u128;
2001 if lo > hi || (lo == hi && *end == RangeEnd::Excluded) {
2002 // This should have been caught earlier by E0030.
2003 bug!("malformed range pattern: {}..={}", lo, (hi - offset));
2005 Some(IntRange { range: lo..=(hi - offset), ty, span })
2011 fn from_ctor<'a>(ctor: &'a Constructor<'tcx>) -> Option<&'a IntRange<'tcx>> {
2013 IntRange(range) => Some(range),
2018 // The return value of `signed_bias` should be XORed with an endpoint to encode/decode it.
2019 fn signed_bias(tcx: TyCtxt<'tcx>, ty: Ty<'tcx>) -> u128 {
2022 let bits = Integer::from_attr(&tcx, SignedInt(ity)).size().bits() as u128;
2029 /// Returns a collection of ranges that spans the values covered by `ranges`, subtracted
2030 /// by the values covered by `self`: i.e., `ranges \ self` (in set notation).
2031 fn subtract_from(&self, ranges: Vec<IntRange<'tcx>>) -> Vec<IntRange<'tcx>> {
2032 let mut remaining_ranges = vec![];
2034 let span = self.span;
2035 let (lo, hi) = self.boundaries();
2036 for subrange in ranges {
2037 let (subrange_lo, subrange_hi) = subrange.range.into_inner();
2038 if lo > subrange_hi || subrange_lo > hi {
2039 // The pattern doesn't intersect with the subrange at all,
2040 // so the subrange remains untouched.
2041 remaining_ranges.push(IntRange { range: subrange_lo..=subrange_hi, ty, span });
2043 if lo > subrange_lo {
2044 // The pattern intersects an upper section of the
2045 // subrange, so a lower section will remain.
2046 remaining_ranges.push(IntRange { range: subrange_lo..=(lo - 1), ty, span });
2048 if hi < subrange_hi {
2049 // The pattern intersects a lower section of the
2050 // subrange, so an upper section will remain.
2051 remaining_ranges.push(IntRange { range: (hi + 1)..=subrange_hi, ty, span });
2058 fn is_subrange(&self, other: &Self) -> bool {
2059 other.range.start() <= self.range.start() && self.range.end() <= other.range.end()
2062 fn intersection(&self, tcx: TyCtxt<'tcx>, other: &Self) -> Option<Self> {
2064 let (lo, hi) = self.boundaries();
2065 let (other_lo, other_hi) = other.boundaries();
2066 if self.treat_exhaustively(tcx) {
2067 if lo <= other_hi && other_lo <= hi {
2068 let span = other.span;
2069 Some(IntRange { range: max(lo, other_lo)..=min(hi, other_hi), ty, span })
2074 // If the range should not be treated exhaustively, fallback to checking for inclusion.
2075 if self.is_subrange(other) { Some(self.clone()) } else { None }
2079 fn suspicious_intersection(&self, other: &Self) -> bool {
2080 // `false` in the following cases:
2081 // 1 ---- // 1 ---------- // 1 ---- // 1 ----
2082 // 2 ---------- // 2 ---- // 2 ---- // 2 ----
2084 // The following are currently `false`, but could be `true` in the future (#64007):
2085 // 1 --------- // 1 ---------
2086 // 2 ---------- // 2 ----------
2088 // `true` in the following cases:
2089 // 1 ------- // 1 -------
2090 // 2 -------- // 2 -------
2091 let (lo, hi) = self.boundaries();
2092 let (other_lo, other_hi) = other.boundaries();
2093 lo == other_hi || hi == other_lo
2096 fn to_pat(&self, tcx: TyCtxt<'tcx>) -> Pat<'tcx> {
2097 let (lo, hi) = self.boundaries();
2099 let bias = IntRange::signed_bias(tcx, self.ty);
2100 let (lo, hi) = (lo ^ bias, hi ^ bias);
2102 let ty = ty::ParamEnv::empty().and(self.ty);
2103 let lo_const = ty::Const::from_bits(tcx, lo, ty);
2104 let hi_const = ty::Const::from_bits(tcx, hi, ty);
2106 let kind = if lo == hi {
2107 PatKind::Constant { value: lo_const }
2109 PatKind::Range(PatRange { lo: lo_const, hi: hi_const, end: RangeEnd::Included })
2112 // This is a brand new pattern, so we don't reuse `self.span`.
2113 Pat { ty: self.ty, span: DUMMY_SP, kind: Box::new(kind) }
2116 /// For exhaustive integer matching, some constructors are grouped within other constructors
2117 /// (namely integer typed values are grouped within ranges). However, when specialising these
2118 /// constructors, we want to be specialising for the underlying constructors (the integers), not
2119 /// the groups (the ranges). Thus we need to split the groups up. Splitting them up naïvely would
2120 /// mean creating a separate constructor for every single value in the range, which is clearly
2121 /// impractical. However, observe that for some ranges of integers, the specialisation will be
2122 /// identical across all values in that range (i.e., there are equivalence classes of ranges of
2123 /// constructors based on their `U(S(c, P), S(c, p))` outcome). These classes are grouped by
2124 /// the patterns that apply to them (in the matrix `P`). We can split the range whenever the
2125 /// patterns that apply to that range (specifically: the patterns that *intersect* with that range)
2127 /// Our solution, therefore, is to split the range constructor into subranges at every single point
2128 /// the group of intersecting patterns changes (using the method described below).
2129 /// And voilà! We're testing precisely those ranges that we need to, without any exhaustive matching
2130 /// on actual integers. The nice thing about this is that the number of subranges is linear in the
2131 /// number of rows in the matrix (i.e., the number of cases in the `match` statement), so we don't
2132 /// need to be worried about matching over gargantuan ranges.
2134 /// Essentially, given the first column of a matrix representing ranges, looking like the following:
2136 /// |------| |----------| |-------| ||
2137 /// |-------| |-------| |----| ||
2140 /// We split the ranges up into equivalence classes so the ranges are no longer overlapping:
2142 /// |--|--|||-||||--||---|||-------| |-|||| ||
2144 /// The logic for determining how to split the ranges is fairly straightforward: we calculate
2145 /// boundaries for each interval range, sort them, then create constructors for each new interval
2146 /// between every pair of boundary points. (This essentially sums up to performing the intuitive
2147 /// merging operation depicted above.)
2150 cx: &MatchCheckCtxt<'p, 'tcx>,
2152 matrix: &Matrix<'p, 'tcx>,
2153 hir_id: Option<HirId>,
2154 ) -> SmallVec<[Constructor<'tcx>; 1]> {
2157 /// Represents a border between 2 integers. Because the intervals spanning borders
2158 /// must be able to cover every integer, we need to be able to represent
2159 /// 2^128 + 1 such borders.
2160 #[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Debug)]
2166 // A function for extracting the borders of an integer interval.
2167 fn range_borders(r: IntRange<'_>) -> impl Iterator<Item = Border> {
2168 let (lo, hi) = r.range.into_inner();
2169 let from = Border::JustBefore(lo);
2170 let to = match hi.checked_add(1) {
2171 Some(m) => Border::JustBefore(m),
2172 None => Border::AfterMax,
2174 vec![from, to].into_iter()
2177 // Collect the span and range of all the intersecting ranges to lint on likely
2178 // incorrect range patterns. (#63987)
2179 let mut overlaps = vec![];
2180 let row_len = matrix.patterns.get(0).map(|r| r.len()).unwrap_or(0);
2181 // `borders` is the set of borders between equivalence classes: each equivalence
2182 // class lies between 2 borders.
2183 let row_borders = matrix
2185 .filter_map(|ctor| IntRange::from_ctor(ctor))
2186 .filter_map(|range| {
2187 let intersection = self.intersection(cx.tcx, &range);
2188 let should_lint = self.suspicious_intersection(&range);
2189 if let (Some(range), 1, true) = (&intersection, row_len, should_lint) {
2190 // FIXME: for now, only check for overlapping ranges on simple range
2191 // patterns. Otherwise with the current logic the following is detected
2193 // match (10u8, true) {
2194 // (0 ..= 125, false) => {}
2195 // (126 ..= 255, false) => {}
2196 // (0 ..= 255, true) => {}
2198 overlaps.push(range.clone());
2202 .flat_map(range_borders);
2203 let self_borders = range_borders(self.clone());
2204 let mut borders: Vec<_> = row_borders.chain(self_borders).collect();
2205 borders.sort_unstable();
2207 self.lint_overlapping_patterns(cx.tcx, hir_id, ty, overlaps);
2209 // We're going to iterate through every adjacent pair of borders, making sure that
2210 // each represents an interval of nonnegative length, and convert each such
2211 // interval into a constructor.
2214 .filter_map(|&pair| match pair {
2215 [Border::JustBefore(n), Border::JustBefore(m)] => {
2222 [Border::JustBefore(n), Border::AfterMax] => Some(n..=u128::MAX),
2223 [Border::AfterMax, _] => None,
2225 .map(|range| IntRange { range, ty, span: pcx.span })
2230 fn lint_overlapping_patterns(
2233 hir_id: Option<HirId>,
2235 overlaps: Vec<IntRange<'tcx>>,
2237 if let (true, Some(hir_id)) = (!overlaps.is_empty(), hir_id) {
2238 tcx.struct_span_lint_hir(
2239 lint::builtin::OVERLAPPING_PATTERNS,
2243 let mut err = lint.build("multiple patterns covering the same range");
2244 err.span_label(self.span, "overlapping patterns");
2245 for int_range in overlaps {
2246 // Use the real type for user display of the ranges:
2250 "this range overlaps on `{}`",
2251 IntRange { range: int_range.range, ty, span: DUMMY_SP }.to_pat(tcx),
2262 /// Ignore spans when comparing, they don't carry semantic information as they are only for lints.
2263 impl<'tcx> std::cmp::PartialEq for IntRange<'tcx> {
2264 fn eq(&self, other: &Self) -> bool {
2265 self.range == other.range && self.ty == other.ty
2269 // A struct to compute a set of constructors equivalent to `all_ctors \ used_ctors`.
2270 struct MissingConstructors<'tcx> {
2271 all_ctors: Vec<Constructor<'tcx>>,
2272 used_ctors: Vec<Constructor<'tcx>>,
2275 impl<'tcx> MissingConstructors<'tcx> {
2276 fn new(all_ctors: Vec<Constructor<'tcx>>, used_ctors: Vec<Constructor<'tcx>>) -> Self {
2277 MissingConstructors { all_ctors, used_ctors }
2280 fn into_inner(self) -> (Vec<Constructor<'tcx>>, Vec<Constructor<'tcx>>) {
2281 (self.all_ctors, self.used_ctors)
2284 fn is_empty(&self) -> bool {
2285 self.iter().next().is_none()
2287 /// Whether this contains all the constructors for the given type or only a
2289 fn all_ctors_are_missing(&self) -> bool {
2290 self.used_ctors.is_empty()
2293 /// Iterate over all_ctors \ used_ctors
2294 fn iter<'a>(&'a self) -> impl Iterator<Item = Constructor<'tcx>> + Captures<'a> {
2295 self.all_ctors.iter().flat_map(move |req_ctor| req_ctor.subtract_ctors(&self.used_ctors))
2299 impl<'tcx> fmt::Debug for MissingConstructors<'tcx> {
2300 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
2301 let ctors: Vec<_> = self.iter().collect();
2302 write!(f, "{:?}", ctors)
2306 /// Algorithm from http://moscova.inria.fr/~maranget/papers/warn/index.html.
2307 /// The algorithm from the paper has been modified to correctly handle empty
2308 /// types. The changes are:
2309 /// (0) We don't exit early if the pattern matrix has zero rows. We just
2310 /// continue to recurse over columns.
2311 /// (1) all_constructors will only return constructors that are statically
2312 /// possible. E.g., it will only return `Ok` for `Result<T, !>`.
2314 /// This finds whether a (row) vector `v` of patterns is 'useful' in relation
2315 /// to a set of such vectors `m` - this is defined as there being a set of
2316 /// inputs that will match `v` but not any of the sets in `m`.
2318 /// All the patterns at each column of the `matrix ++ v` matrix must have the same type.
2320 /// This is used both for reachability checking (if a pattern isn't useful in
2321 /// relation to preceding patterns, it is not reachable) and exhaustiveness
2322 /// checking (if a wildcard pattern is useful in relation to a matrix, the
2323 /// matrix isn't exhaustive).
2325 /// `is_under_guard` is used to inform if the pattern has a guard. If it
2326 /// has one it must not be inserted into the matrix. This shouldn't be
2327 /// relied on for soundness.
2328 crate fn is_useful<'p, 'tcx>(
2329 cx: &MatchCheckCtxt<'p, 'tcx>,
2330 matrix: &Matrix<'p, 'tcx>,
2331 v: &PatStack<'p, 'tcx>,
2332 witness_preference: WitnessPreference,
2334 is_under_guard: bool,
2336 ) -> Usefulness<'tcx> {
2337 let Matrix { patterns: rows, .. } = matrix;
2338 debug!("is_useful({:#?}, {:#?})", matrix, v);
2340 // The base case. We are pattern-matching on () and the return value is
2341 // based on whether our matrix has a row or not.
2342 // NOTE: This could potentially be optimized by checking rows.is_empty()
2343 // first and then, if v is non-empty, the return value is based on whether
2344 // the type of the tuple we're checking is inhabited or not.
2346 return if rows.is_empty() {
2347 Usefulness::new_useful(witness_preference)
2353 assert!(rows.iter().all(|r| r.len() == v.len()));
2355 // If the first pattern is an or-pattern, expand it.
2356 if let Some(vs) = v.expand_or_pat() {
2357 // We need to push the already-seen patterns into the matrix in order to detect redundant
2358 // branches like `Some(_) | Some(0)`. We also keep track of the unreachable subpatterns.
2359 let mut matrix = matrix.clone();
2360 // `Vec` of all the unreachable branches of the current or-pattern.
2361 let mut unreachable_branches = Vec::new();
2362 // Subpatterns that are unreachable from all branches. E.g. in the following case, the last
2363 // `true` is unreachable only from one branch, so it is overall reachable.
2366 // match (true, true) {
2367 // (true, true) => {}
2368 // (false | true, false | true) => {}
2371 let mut unreachable_subpats = FxHashSet::default();
2372 // Whether any branch at all is useful.
2373 let mut any_is_useful = false;
2376 let res = is_useful(cx, &matrix, &v, witness_preference, hir_id, is_under_guard, false);
2380 any_is_useful = true;
2381 // Initialize with the first set of unreachable subpatterns encountered.
2382 unreachable_subpats = pats.into_iter().collect();
2384 // Keep the patterns unreachable from both this and previous branches.
2385 unreachable_subpats =
2386 pats.into_iter().filter(|p| unreachable_subpats.contains(p)).collect();
2389 NotUseful => unreachable_branches.push(v.head().span),
2390 UsefulWithWitness(_) => {
2391 bug!("Encountered or-pat in `v` during exhaustiveness checking")
2394 // If pattern has a guard don't add it to the matrix
2395 if !is_under_guard {
2400 // Collect all the unreachable patterns.
2401 unreachable_branches.extend(unreachable_subpats);
2402 return Useful(unreachable_branches);
2408 // FIXME(Nadrieril): Hack to work around type normalization issues (see #72476).
2409 let ty = matrix.heads().next().map(|r| r.ty).unwrap_or(v.head().ty);
2410 let pcx = PatCtxt { ty, span: v.head().span };
2412 debug!("is_useful_expand_first_col: pcx={:#?}, expanding {:#?}", pcx, v.head());
2414 let constructor = v.head_ctor(cx);
2415 let ret = if !constructor.is_wildcard() {
2416 debug!("is_useful - expanding constructor: {:#?}", constructor);
2418 .split(cx, pcx, matrix, Some(hir_id))
2421 is_useful_specialized(
2432 .find(|result| result.is_useful())
2433 .unwrap_or(NotUseful)
2435 debug!("is_useful - expanding wildcard");
2437 let used_ctors: Vec<Constructor<'_>> =
2438 matrix.head_ctors(cx).cloned().filter(|c| !c.is_wildcard()).collect();
2439 debug!("is_useful_used_ctors = {:#?}", used_ctors);
2440 // `all_ctors` are all the constructors for the given type, which
2441 // should all be represented (or caught with the wild pattern `_`).
2442 let all_ctors = all_constructors(cx, pcx);
2443 debug!("is_useful_all_ctors = {:#?}", all_ctors);
2445 // `missing_ctors` is the set of constructors from the same type as the
2446 // first column of `matrix` that are matched only by wildcard patterns
2447 // from the first column.
2449 // Therefore, if there is some pattern that is unmatched by `matrix`,
2450 // it will still be unmatched if the first constructor is replaced by
2451 // any of the constructors in `missing_ctors`
2453 // Missing constructors are those that are not matched by any non-wildcard patterns in the
2454 // current column. We only fully construct them on-demand, because they're rarely used and
2456 let missing_ctors = MissingConstructors::new(all_ctors, used_ctors);
2458 debug!("is_useful_missing_ctors.empty()={:#?}", missing_ctors.is_empty(),);
2460 if missing_ctors.is_empty() {
2461 let (all_ctors, _) = missing_ctors.into_inner();
2464 .flat_map(|ctor| ctor.split(cx, pcx, matrix, None))
2466 is_useful_specialized(
2477 .find(|result| result.is_useful())
2478 .unwrap_or(NotUseful)
2480 let ctor_wild_subpatterns = Fields::empty();
2481 let matrix = matrix.specialize_constructor(cx, &constructor, &ctor_wild_subpatterns);
2482 // Unwrap is ok: v can always be specialized with its own constructor.
2484 v.specialize_constructor(cx, &constructor, &ctor_wild_subpatterns, true).unwrap();
2486 is_useful(cx, &matrix, &v, witness_preference, hir_id, is_under_guard, false);
2488 // In this case, there's at least one "free"
2489 // constructor that is only matched against by
2490 // wildcard patterns.
2492 // There are 2 ways we can report a witness here.
2493 // Commonly, we can report all the "free"
2494 // constructors as witnesses, e.g., if we have:
2497 // enum Direction { N, S, E, W }
2498 // let Direction::N = ...;
2501 // we can report 3 witnesses: `S`, `E`, and `W`.
2503 // However, there is a case where we don't want
2504 // to do this and instead report a single `_` witness:
2505 // if the user didn't actually specify a constructor
2506 // in this arm, e.g., in
2509 // let x: (Direction, Direction, bool) = ...;
2510 // let (_, _, false) = x;
2513 // we don't want to show all 16 possible witnesses
2514 // `(<direction-1>, <direction-2>, true)` - we are
2515 // satisfied with `(_, _, true)`. In this case,
2516 // `used_ctors` is empty.
2517 // The exception is: if we are at the top-level, for example in an empty match, we
2518 // sometimes prefer reporting the list of constructors instead of just `_`.
2519 let report_ctors_rather_than_wildcard = is_top_level && !IntRange::is_integral(pcx.ty);
2520 if missing_ctors.all_ctors_are_missing() && !report_ctors_rather_than_wildcard {
2521 // All constructors are unused. Add a wild pattern
2522 // rather than each individual constructor.
2523 usefulness.apply_wildcard(pcx.ty)
2525 // Construct for each missing constructor a "wild" version of this
2526 // constructor, that matches everything that can be built with
2527 // it. For example, if `ctor` is a `Constructor::Variant` for
2528 // `Option::Some`, we get the pattern `Some(_)`.
2529 usefulness.apply_missing_ctors(cx, pcx.ty, &missing_ctors)
2533 debug!("is_useful::returns({:#?}, {:#?}) = {:?}", matrix, v, ret);
2537 /// A shorthand for the `U(S(c, P), S(c, q))` operation from the paper. I.e., `is_useful` applied
2538 /// to the specialised version of both the pattern matrix `P` and the new pattern `q`.
2539 fn is_useful_specialized<'p, 'tcx>(
2540 cx: &MatchCheckCtxt<'p, 'tcx>,
2541 matrix: &Matrix<'p, 'tcx>,
2542 v: &PatStack<'p, 'tcx>,
2543 ctor: Constructor<'tcx>,
2545 witness_preference: WitnessPreference,
2547 is_under_guard: bool,
2548 ) -> Usefulness<'tcx> {
2549 debug!("is_useful_specialized({:#?}, {:#?}, {:?})", v, ctor, ty);
2551 // We cache the result of `Fields::wildcards` because it is used a lot.
2552 let ctor_wild_subpatterns = Fields::wildcards(cx, &ctor, ty);
2553 let matrix = matrix.specialize_constructor(cx, &ctor, &ctor_wild_subpatterns);
2554 v.specialize_constructor(cx, &ctor, &ctor_wild_subpatterns, true)
2555 .map(|v| is_useful(cx, &matrix, &v, witness_preference, hir_id, is_under_guard, false))
2556 .map(|u| u.apply_constructor(cx, &ctor, ty, &ctor_wild_subpatterns))
2557 .unwrap_or(NotUseful)
2560 /// Determines the constructor that the given pattern can be specialized to.
2561 /// Returns `None` in case of a catch-all, which can't be specialized.
2562 fn pat_constructor<'p, 'tcx>(
2563 cx: &MatchCheckCtxt<'p, 'tcx>,
2565 ) -> Constructor<'tcx> {
2567 PatKind::AscribeUserType { .. } => bug!(), // Handled by `expand_pattern`
2568 PatKind::Binding { .. } | PatKind::Wild => Wildcard,
2569 PatKind::Leaf { .. } | PatKind::Deref { .. } => Single,
2570 PatKind::Variant { adt_def, variant_index, .. } => {
2571 Variant(adt_def.variants[variant_index].def_id)
2573 PatKind::Constant { value } => {
2574 if let Some(int_range) = IntRange::from_const(cx.tcx, cx.param_env, value, pat.span) {
2577 match value.ty.kind() {
2578 ty::Float(_) => FloatRange(value, value, RangeEnd::Included),
2579 ty::Ref(_, t, _) if t.is_str() => Str(value),
2580 // All constants that can be structurally matched have already been expanded
2581 // into the corresponding `Pat`s by `const_to_pat`. Constants that remain are
2587 PatKind::Range(PatRange { lo, hi, end }) => {
2589 if let Some(int_range) = IntRange::from_range(
2591 lo.eval_bits(cx.tcx, cx.param_env, lo.ty),
2592 hi.eval_bits(cx.tcx, cx.param_env, hi.ty),
2599 FloatRange(lo, hi, end)
2602 PatKind::Array { ref prefix, ref slice, ref suffix }
2603 | PatKind::Slice { ref prefix, ref slice, ref suffix } => {
2604 let array_len = match pat.ty.kind() {
2605 ty::Array(_, length) => Some(length.eval_usize(cx.tcx, cx.param_env)),
2606 ty::Slice(_) => None,
2607 _ => span_bug!(pat.span, "bad ty {:?} for slice pattern", pat.ty),
2609 let prefix = prefix.len() as u64;
2610 let suffix = suffix.len() as u64;
2612 if slice.is_some() { VarLen(prefix, suffix) } else { FixedLen(prefix + suffix) };
2613 Slice(Slice { array_len, kind })
2615 PatKind::Or { .. } => bug!("Or-pattern should have been expanded earlier on."),