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))`
185 //! For example, if `P` is:
194 //! and `p` is [Some(false), 0], then we don't care about row 2 since we know `p` only
195 //! matches values that row 2 doesn't. For row 1 however, we need to dig into the
196 //! arguments of `Some` to know whether some new value is covered. So we compute
197 //! `U([[true, _]], [false, 0])`.
199 //! - If `p_1 == _`, then we look at the list of constructors that appear in the first
200 //! component of the rows of `P`:
201 //! + If there are some constructors that aren't present, then we might think that the
202 //! wildcard `_` is useful, since it covers those constructors that weren't covered
204 //! That's almost correct, but only works if there were no wildcards in those first
205 //! components. So we need to check that `p` is useful with respect to the rows that
206 //! start with a wildcard, if there are any. This is where `S(_, x)` comes in:
207 //! `U(P, p) := U(S(_, P), S(_, p))`
209 //! For example, if `P` is:
214 //! [None, false, 1],
218 //! and `p` is [_, false, _], the `Some` constructor doesn't appear in `P`. So if we
219 //! only had row 2, we'd know that `p` is useful. However row 1 starts with a
220 //! wildcard, so we need to check whether `U([[true, _]], [false, 1])`.
222 //! + Otherwise, all possible constructors (for the relevant type) are present. In this
223 //! case we must check whether the wildcard pattern covers any unmatched value. For
224 //! that, we can think of the `_` pattern as a big OR-pattern that covers all
225 //! possible constructors. For `Option`, that would mean `_ = None | Some(_)` for
226 //! example. The wildcard pattern is useful in this case if it is useful when
227 //! specialized to one of the possible constructors. So we compute:
228 //! `U(P, p) := ∃(k ϵ constructors) U(S(k, P), S(k, p))`
230 //! For example, if `P` is:
239 //! and `p` is [_, false], both `None` and `Some` constructors appear in the first
240 //! components of `P`. We will therefore try popping both constructors in turn: we
241 //! compute `U([[true, _]], [_, false])` for the `Some` constructor, and `U([[false]],
242 //! [false])` for the `None` constructor. The first case returns true, so we know that
243 //! `p` is useful for `P`. Indeed, it matches `[Some(false), _]` that wasn't matched
246 //! - If `p_1 == r_1 | r_2`, then the usefulness depends on each `r_i` separately:
247 //! `U(P, p) := U(P, (r_1, p_2, .., p_n))
248 //! || U(P, (r_2, p_2, .., p_n))`
250 //! Modifications to the algorithm
251 //! ------------------------------
252 //! The algorithm in the paper doesn't cover some of the special cases that arise in Rust, for
253 //! example uninhabited types and variable-length slice patterns. These are drawn attention to
254 //! throughout the code below. I'll make a quick note here about how exhaustive integer matching is
255 //! accounted for, though.
257 //! Exhaustive integer matching
258 //! ---------------------------
259 //! An integer type can be thought of as a (huge) sum type: 1 | 2 | 3 | ...
260 //! So to support exhaustive integer matching, we can make use of the logic in the paper for
261 //! OR-patterns. However, we obviously can't just treat ranges x..=y as individual sums, because
262 //! they are likely gigantic. So we instead treat ranges as constructors of the integers. This means
263 //! that we have a constructor *of* constructors (the integers themselves). We then need to work
264 //! through all the inductive step rules above, deriving how the ranges would be treated as
265 //! OR-patterns, and making sure that they're treated in the same way even when they're ranges.
266 //! There are really only four special cases here:
267 //! - When we match on a constructor that's actually a range, we have to treat it as if we would
269 //! + It turns out that we can simply extend the case for single-value patterns in
270 //! `specialize` to either be *equal* to a value constructor, or *contained within* a range
272 //! + When the pattern itself is a range, you just want to tell whether any of the values in
273 //! the pattern range coincide with values in the constructor range, which is precisely
275 //! Since when encountering a range pattern for a value constructor, we also use inclusion, it
276 //! means that whenever the constructor is a value/range and the pattern is also a value/range,
277 //! we can simply use intersection to test usefulness.
278 //! - When we're testing for usefulness of a pattern and the pattern's first component is a
280 //! + If all the constructors appear in the matrix, we have a slight complication. By default,
281 //! the behaviour (i.e., a disjunction over specialised matrices for each constructor) is
282 //! invalid, because we want a disjunction over every *integer* in each range, not just a
283 //! disjunction over every range. This is a bit more tricky to deal with: essentially we need
284 //! to form equivalence classes of subranges of the constructor range for which the behaviour
285 //! of the matrix `P` and new pattern `p` are the same. This is described in more
286 //! detail in `Constructor::split`.
287 //! + If some constructors are missing from the matrix, it turns out we don't need to do
288 //! anything special (because we know none of the integers are actually wildcards: i.e., we
289 //! can't span wildcards using ranges).
290 use self::Constructor::*;
291 use self::SliceKind::*;
292 use self::Usefulness::*;
293 use self::WitnessPreference::*;
295 use rustc_data_structures::captures::Captures;
296 use rustc_data_structures::fx::FxHashSet;
297 use rustc_data_structures::sync::OnceCell;
298 use rustc_index::vec::Idx;
300 use super::{compare_const_vals, PatternFoldable, PatternFolder};
301 use super::{FieldPat, Pat, PatKind, PatRange};
303 use rustc_arena::TypedArena;
304 use rustc_attr::{SignedInt, UnsignedInt};
305 use rustc_hir::def_id::DefId;
306 use rustc_hir::{HirId, RangeEnd};
307 use rustc_middle::mir::interpret::{truncate, ConstValue};
308 use rustc_middle::mir::Field;
309 use rustc_middle::ty::layout::IntegerExt;
310 use rustc_middle::ty::{self, Const, Ty, TyCtxt};
311 use rustc_session::lint;
312 use rustc_span::{Span, DUMMY_SP};
313 use rustc_target::abi::{Integer, Size, VariantIdx};
315 use smallvec::{smallvec, SmallVec};
316 use std::cmp::{self, max, min, Ordering};
318 use std::iter::{FromIterator, IntoIterator};
319 use std::ops::RangeInclusive;
321 crate fn expand_pattern<'tcx>(pat: Pat<'tcx>) -> Pat<'tcx> {
322 LiteralExpander.fold_pattern(&pat)
325 struct LiteralExpander;
327 impl<'tcx> PatternFolder<'tcx> for LiteralExpander {
328 fn fold_pattern(&mut self, pat: &Pat<'tcx>) -> Pat<'tcx> {
329 debug!("fold_pattern {:?} {:?} {:?}", pat, pat.ty.kind(), pat.kind);
330 match (pat.ty.kind(), pat.kind.as_ref()) {
331 (_, PatKind::Binding { subpattern: Some(s), .. }) => s.fold_with(self),
332 (_, PatKind::AscribeUserType { subpattern: s, .. }) => s.fold_with(self),
333 (ty::Ref(_, t, _), PatKind::Constant { .. }) if t.is_str() => {
334 // Treat string literal patterns as deref patterns to a `str` constant, i.e.
335 // `&CONST`. This expands them like other const patterns. This could have been done
336 // in `const_to_pat`, but that causes issues with the rest of the matching code.
337 let mut new_pat = pat.super_fold_with(self);
338 // Make a fake const pattern of type `str` (instead of `&str`). That the carried
339 // constant value still knows it is of type `&str`.
342 kind: Box::new(PatKind::Deref { subpattern: new_pat }),
347 _ => pat.super_fold_with(self),
352 impl<'tcx> Pat<'tcx> {
353 pub(super) fn is_wildcard(&self) -> bool {
355 PatKind::Binding { subpattern: None, .. } | PatKind::Wild => true,
361 /// A row of a matrix. Rows of len 1 are very common, which is why `SmallVec[_; 2]`
363 #[derive(Debug, Clone)]
364 crate struct PatStack<'p, 'tcx> {
365 pats: SmallVec<[&'p Pat<'tcx>; 2]>,
366 /// Cache for the constructor of the head
367 head_ctor: OnceCell<Constructor<'tcx>>,
370 impl<'p, 'tcx> PatStack<'p, 'tcx> {
371 crate fn from_pattern(pat: &'p Pat<'tcx>) -> Self {
372 Self::from_vec(smallvec![pat])
375 fn from_vec(vec: SmallVec<[&'p Pat<'tcx>; 2]>) -> Self {
376 PatStack { pats: vec, head_ctor: OnceCell::new() }
379 fn is_empty(&self) -> bool {
383 fn len(&self) -> usize {
387 fn head(&self) -> &'p Pat<'tcx> {
391 fn head_ctor<'a>(&'a self, cx: &MatchCheckCtxt<'p, 'tcx>) -> &'a Constructor<'tcx> {
392 self.head_ctor.get_or_init(|| pat_constructor(cx, self.head()))
395 fn iter(&self) -> impl Iterator<Item = &Pat<'tcx>> {
396 self.pats.iter().copied()
399 // If the first pattern is an or-pattern, expand this pattern. Otherwise, return `None`.
400 fn expand_or_pat(&self) -> Option<Vec<Self>> {
403 } else if let PatKind::Or { pats } = &*self.head().kind {
407 let mut new_patstack = PatStack::from_pattern(pat);
408 new_patstack.pats.extend_from_slice(&self.pats[1..]);
418 /// This computes `S(self.head_ctor(), self)`. See top of the file for explanations.
420 /// Structure patterns with a partial wild pattern (Foo { a: 42, .. }) have their missing
421 /// fields filled with wild patterns.
423 /// This is roughly the inverse of `Constructor::apply`.
424 fn pop_head_constructor(&self, ctor_wild_subpatterns: &Fields<'p, 'tcx>) -> PatStack<'p, 'tcx> {
425 // We pop the head pattern and push the new fields extracted from the arguments of
427 let new_fields = ctor_wild_subpatterns.replace_with_pattern_arguments(self.head());
428 new_fields.push_on_patstack(&self.pats[1..])
432 impl<'p, 'tcx> Default for PatStack<'p, 'tcx> {
433 fn default() -> Self {
434 Self::from_vec(smallvec![])
438 impl<'p, 'tcx> PartialEq for PatStack<'p, 'tcx> {
439 fn eq(&self, other: &Self) -> bool {
440 self.pats == other.pats
444 impl<'p, 'tcx> FromIterator<&'p Pat<'tcx>> for PatStack<'p, 'tcx> {
445 fn from_iter<T>(iter: T) -> Self
447 T: IntoIterator<Item = &'p Pat<'tcx>>,
449 Self::from_vec(iter.into_iter().collect())
454 #[derive(Clone, PartialEq)]
455 crate struct Matrix<'p, 'tcx> {
456 patterns: Vec<PatStack<'p, 'tcx>>,
459 impl<'p, 'tcx> Matrix<'p, 'tcx> {
460 crate fn empty() -> Self {
461 Matrix { patterns: vec![] }
464 /// Pushes a new row to the matrix. If the row starts with an or-pattern, this expands it.
465 crate fn push(&mut self, row: PatStack<'p, 'tcx>) {
466 if let Some(rows) = row.expand_or_pat() {
468 // We recursively expand the or-patterns of the new rows.
469 // This is necessary as we might have `0 | (1 | 2)` or e.g., `x @ 0 | x @ (1 | 2)`.
473 self.patterns.push(row);
477 /// Iterate over the first component of each row
478 fn heads<'a>(&'a self) -> impl Iterator<Item = &'a Pat<'tcx>> + Captures<'p> {
479 self.patterns.iter().map(|r| r.head())
482 /// Iterate over the first constructor of each row
485 cx: &'a MatchCheckCtxt<'p, 'tcx>,
486 ) -> impl Iterator<Item = &'a Constructor<'tcx>> + Captures<'a> + Captures<'p> {
487 self.patterns.iter().map(move |r| r.head_ctor(cx))
490 /// This computes `S(constructor, self)`. See top of the file for explanations.
491 fn specialize_constructor(
493 pcx: PatCtxt<'_, 'p, 'tcx>,
494 ctor: &Constructor<'tcx>,
495 ctor_wild_subpatterns: &Fields<'p, 'tcx>,
496 ) -> Matrix<'p, 'tcx> {
499 .filter(|r| ctor.is_covered_by(pcx, r.head_ctor(pcx.cx)))
500 .map(|r| r.pop_head_constructor(ctor_wild_subpatterns))
505 /// Pretty-printer for matrices of patterns, example:
508 /// +++++++++++++++++++++++++++++
510 /// +++++++++++++++++++++++++++++
511 /// + true + [First] +
512 /// +++++++++++++++++++++++++++++
513 /// + true + [Second(true)] +
514 /// +++++++++++++++++++++++++++++
516 /// +++++++++++++++++++++++++++++
517 /// + _ + [_, _, tail @ ..] +
518 /// +++++++++++++++++++++++++++++
520 impl<'p, 'tcx> fmt::Debug for Matrix<'p, 'tcx> {
521 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
524 let Matrix { patterns: m, .. } = self;
525 let pretty_printed_matrix: Vec<Vec<String>> =
526 m.iter().map(|row| row.iter().map(|pat| format!("{:?}", pat)).collect()).collect();
528 let column_count = m.iter().map(|row| row.len()).max().unwrap_or(0);
529 assert!(m.iter().all(|row| row.len() == column_count));
530 let column_widths: Vec<usize> = (0..column_count)
531 .map(|col| pretty_printed_matrix.iter().map(|row| row[col].len()).max().unwrap_or(0))
534 let total_width = column_widths.iter().cloned().sum::<usize>() + column_count * 3 + 1;
535 let br = "+".repeat(total_width);
536 write!(f, "{}\n", br)?;
537 for row in pretty_printed_matrix {
539 for (column, pat_str) in row.into_iter().enumerate() {
541 write!(f, "{:1$}", pat_str, column_widths[column])?;
545 write!(f, "{}\n", br)?;
551 impl<'p, 'tcx> FromIterator<PatStack<'p, 'tcx>> for Matrix<'p, 'tcx> {
552 fn from_iter<T>(iter: T) -> Self
554 T: IntoIterator<Item = PatStack<'p, 'tcx>>,
556 let mut matrix = Matrix::empty();
558 // Using `push` ensures we correctly expand or-patterns.
565 crate struct MatchCheckCtxt<'a, 'tcx> {
566 crate tcx: TyCtxt<'tcx>,
567 /// The module in which the match occurs. This is necessary for
568 /// checking inhabited-ness of types because whether a type is (visibly)
569 /// inhabited can depend on whether it was defined in the current module or
570 /// not. E.g., `struct Foo { _private: ! }` cannot be seen to be empty
571 /// outside it's module and should not be matchable with an empty match
574 crate param_env: ty::ParamEnv<'tcx>,
575 crate pattern_arena: &'a TypedArena<Pat<'tcx>>,
578 impl<'a, 'tcx> MatchCheckCtxt<'a, 'tcx> {
579 fn is_uninhabited(&self, ty: Ty<'tcx>) -> bool {
580 if self.tcx.features().exhaustive_patterns {
581 self.tcx.is_ty_uninhabited_from(self.module, ty, self.param_env)
587 /// Returns whether the given type is an enum from another crate declared `#[non_exhaustive]`.
588 crate fn is_foreign_non_exhaustive_enum(&self, ty: Ty<'tcx>) -> bool {
590 ty::Adt(def, ..) => {
591 def.is_enum() && def.is_variant_list_non_exhaustive() && !def.did.is_local()
598 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
600 /// Patterns of length `n` (`[x, y]`).
602 /// Patterns using the `..` notation (`[x, .., y]`).
603 /// Captures any array constructor of `length >= i + j`.
604 /// In the case where `array_len` is `Some(_)`,
605 /// this indicates that we only care about the first `i` and the last `j` values of the array,
606 /// and everything in between is a wildcard `_`.
611 fn arity(self) -> u64 {
613 FixedLen(length) => length,
614 VarLen(prefix, suffix) => prefix + suffix,
618 /// Whether this pattern includes patterns of length `other_len`.
619 fn covers_length(self, other_len: u64) -> bool {
621 FixedLen(len) => len == other_len,
622 VarLen(prefix, suffix) => prefix + suffix <= other_len,
627 /// A constructor for array and slice patterns.
628 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
630 /// `None` if the matched value is a slice, `Some(n)` if it is an array of size `n`.
631 array_len: Option<u64>,
632 /// The kind of pattern it is: fixed-length `[x, y]` or variable length `[x, .., y]`.
637 fn new(array_len: Option<u64>, kind: SliceKind) -> Self {
638 let kind = match (array_len, kind) {
639 // If the middle `..` is empty, we effectively have a fixed-length pattern.
640 (Some(len), VarLen(prefix, suffix)) if prefix + suffix >= len => FixedLen(len),
643 Slice { array_len, kind }
646 fn arity(self) -> u64 {
650 /// The exhaustiveness-checking paper does not include any details on
651 /// checking variable-length slice patterns. However, they may be
652 /// matched by an infinite collection of fixed-length array patterns.
654 /// Checking the infinite set directly would take an infinite amount
655 /// of time. However, it turns out that for each finite set of
656 /// patterns `P`, all sufficiently large array lengths are equivalent:
658 /// Each slice `s` with a "sufficiently-large" length `l ≥ L` that applies
659 /// to exactly the subset `Pₜ` of `P` can be transformed to a slice
660 /// `sₘ` for each sufficiently-large length `m` that applies to exactly
661 /// the same subset of `P`.
663 /// Because of that, each witness for reachability-checking of one
664 /// of the sufficiently-large lengths can be transformed to an
665 /// equally-valid witness of any other length, so we only have
666 /// to check slices of the "minimal sufficiently-large length"
669 /// Note that the fact that there is a *single* `sₘ` for each `m`
670 /// not depending on the specific pattern in `P` is important: if
671 /// you look at the pair of patterns
674 /// Then any slice of length ≥1 that matches one of these two
675 /// patterns can be trivially turned to a slice of any
676 /// other length ≥1 that matches them and vice-versa,
677 /// but the slice of length 2 `[false, true]` that matches neither
678 /// of these patterns can't be turned to a slice from length 1 that
679 /// matches neither of these patterns, so we have to consider
680 /// slices from length 2 there.
682 /// Now, to see that that length exists and find it, observe that slice
683 /// patterns are either "fixed-length" patterns (`[_, _, _]`) or
684 /// "variable-length" patterns (`[_, .., _]`).
686 /// For fixed-length patterns, all slices with lengths *longer* than
687 /// the pattern's length have the same outcome (of not matching), so
688 /// as long as `L` is greater than the pattern's length we can pick
689 /// any `sₘ` from that length and get the same result.
691 /// For variable-length patterns, the situation is more complicated,
692 /// because as seen above the precise value of `sₘ` matters.
694 /// However, for each variable-length pattern `p` with a prefix of length
695 /// `plₚ` and suffix of length `slₚ`, only the first `plₚ` and the last
696 /// `slₚ` elements are examined.
698 /// Therefore, as long as `L` is positive (to avoid concerns about empty
699 /// types), all elements after the maximum prefix length and before
700 /// the maximum suffix length are not examined by any variable-length
701 /// pattern, and therefore can be added/removed without affecting
702 /// them - creating equivalent patterns from any sufficiently-large
705 /// Of course, if fixed-length patterns exist, we must be sure
706 /// that our length is large enough to miss them all, so
707 /// we can pick `L = max(max(FIXED_LEN)+1, max(PREFIX_LEN) + max(SUFFIX_LEN))`
709 /// for example, with the above pair of patterns, all elements
710 /// but the first and last can be added/removed, so any
711 /// witness of length ≥2 (say, `[false, false, true]`) can be
712 /// turned to a witness from any other length ≥2.
713 fn split<'p, 'tcx>(self, pcx: PatCtxt<'_, 'p, 'tcx>) -> SmallVec<[Constructor<'tcx>; 1]> {
714 let (self_prefix, self_suffix) = match self.kind {
715 VarLen(self_prefix, self_suffix) => (self_prefix, self_suffix),
716 _ => return smallvec![Slice(self)],
719 let head_ctors = pcx.matrix.head_ctors(pcx.cx).filter(|c| !c.is_wildcard());
721 let mut max_prefix_len = self_prefix;
722 let mut max_suffix_len = self_suffix;
723 let mut max_fixed_len = 0;
725 for ctor in head_ctors {
726 if let Slice(slice) = ctor {
729 max_fixed_len = cmp::max(max_fixed_len, len);
731 VarLen(prefix, suffix) => {
732 max_prefix_len = cmp::max(max_prefix_len, prefix);
733 max_suffix_len = cmp::max(max_suffix_len, suffix);
737 bug!("unexpected ctor for slice type: {:?}", ctor);
741 // For diagnostics, we keep the prefix and suffix lengths separate, so in the case
742 // where `max_fixed_len + 1` is the largest, we adapt `max_prefix_len` accordingly,
743 // so that `L = max_prefix_len + max_suffix_len`.
744 if max_fixed_len + 1 >= max_prefix_len + max_suffix_len {
745 // The subtraction can't overflow thanks to the above check.
746 // The new `max_prefix_len` is also guaranteed to be larger than its previous
748 max_prefix_len = max_fixed_len + 1 - max_suffix_len;
751 let final_slice = VarLen(max_prefix_len, max_suffix_len);
752 let final_slice = Slice::new(self.array_len, final_slice);
753 match self.array_len {
754 Some(_) => smallvec![Slice(final_slice)],
756 // `self` originally covered the range `(self.arity()..infinity)`. We split that
757 // range into two: lengths smaller than `final_slice.arity()` are treated
758 // independently as fixed-lengths slices, and lengths above are captured by
760 let smaller_lengths = (self.arity()..final_slice.arity()).map(FixedLen);
762 .map(|kind| Slice::new(self.array_len, kind))
763 .chain(Some(final_slice))
770 /// See `Constructor::is_covered_by`
771 fn is_covered_by(self, other: Self) -> bool {
772 other.kind.covers_length(self.arity())
776 /// A value can be decomposed into a constructor applied to some fields. This struct represents
777 /// the constructor. See also `Fields`.
779 /// `pat_constructor` retrieves the constructor corresponding to a pattern.
780 /// `specialize_constructor` returns the list of fields corresponding to a pattern, given a
781 /// constructor. `Constructor::apply` reconstructs the pattern from a pair of `Constructor` and
783 #[derive(Clone, Debug, PartialEq)]
784 enum Constructor<'tcx> {
785 /// The constructor for patterns that have a single constructor, like tuples, struct patterns
786 /// and fixed-length arrays.
790 /// Ranges of integer literal values (`2`, `2..=5` or `2..5`).
791 IntRange(IntRange<'tcx>),
792 /// Ranges of floating-point literal values (`2.0..=5.2`).
793 FloatRange(&'tcx ty::Const<'tcx>, &'tcx ty::Const<'tcx>, RangeEnd),
794 /// String literals. Strings are not quite the same as `&[u8]` so we treat them separately.
795 Str(&'tcx ty::Const<'tcx>),
796 /// Array and slice patterns.
798 /// Constants that must not be matched structurally. They are treated as black
799 /// boxes for the purposes of exhaustiveness: we must not inspect them, and they
800 /// don't count towards making a match exhaustive.
802 /// Fake extra constructor for enums that aren't allowed to be matched exhaustively. Also used
803 /// for those types for which we cannot list constructors explicitly, like `f64` and `str`.
805 /// Wildcard pattern.
809 impl<'tcx> Constructor<'tcx> {
810 fn is_wildcard(&self) -> bool {
811 matches!(self, Wildcard)
814 fn as_int_range(&self) -> Option<&IntRange<'tcx>> {
816 IntRange(range) => Some(range),
821 fn as_slice(&self) -> Option<Slice> {
823 Slice(slice) => Some(*slice),
828 fn variant_index_for_adt(&self, adt: &'tcx ty::AdtDef) -> VariantIdx {
830 Variant(id) => adt.variant_index_with_id(id),
832 assert!(!adt.is_enum());
835 _ => bug!("bad constructor {:?} for adt {:?}", self, adt),
839 /// Some constructors (namely `Wildcard`, `IntRange` and `Slice`) actually stand for a set of actual
840 /// constructors (like variants, integers or fixed-sized slices). When specializing for these
841 /// constructors, we want to be specialising for the actual underlying constructors.
842 /// Naively, we would simply return the list of constructors they correspond to. We instead are
843 /// more clever: if there are constructors that we know will behave the same wrt the current
844 /// matrix, we keep them grouped. For example, all slices of a sufficiently large length
845 /// will either be all useful or all non-useful with a given matrix.
847 /// See the branches for details on how the splitting is done.
849 /// This function may discard some irrelevant constructors if this preserves behavior and
850 /// diagnostics. Eg. for the `_` case, we ignore the constructors already present in the
851 /// matrix, unless all of them are.
853 /// `hir_id` is `None` when we're evaluating the wildcard pattern. In that case we do not want
854 /// to lint for overlapping ranges.
855 fn split<'p>(&self, pcx: PatCtxt<'_, 'p, 'tcx>, hir_id: Option<HirId>) -> SmallVec<[Self; 1]> {
856 debug!("Constructor::split({:#?}, {:#?})", self, pcx.matrix);
859 Wildcard => Constructor::split_wildcard(pcx),
860 // Fast-track if the range is trivial. In particular, we don't do the overlapping
863 if ctor_range.treat_exhaustively(pcx.cx.tcx) && !ctor_range.is_singleton() =>
865 ctor_range.split(pcx, hir_id)
867 Slice(slice @ Slice { kind: VarLen(..), .. }) => slice.split(pcx),
868 // Any other constructor can be used unchanged.
869 _ => smallvec![self.clone()],
873 /// For wildcards, there are two groups of constructors: there are the constructors actually
874 /// present in the matrix (`head_ctors`), and the constructors not present (`missing_ctors`).
875 /// Two constructors that are not in the matrix will either both be caught (by a wildcard), or
876 /// both not be caught. Therefore we can keep the missing constructors grouped together.
877 fn split_wildcard<'p>(pcx: PatCtxt<'_, 'p, 'tcx>) -> SmallVec<[Self; 1]> {
878 // Missing constructors are those that are not matched by any non-wildcard patterns in the
879 // current column. We only fully construct them on-demand, because they're rarely used and
881 let missing_ctors = MissingConstructors::new(pcx);
882 if missing_ctors.is_empty(pcx) {
883 // All the constructors are present in the matrix, so we just go through them all.
884 // We must also split them first.
885 missing_ctors.all_ctors
887 // Some constructors are missing, thus we can specialize with the wildcard constructor,
888 // which will stand for those constructors that are missing, and behaves like any of
894 /// Returns whether `self` is covered by `other`, i.e. whether `self` is a subset of `other`.
895 /// For the simple cases, this is simply checking for equality. For the "grouped" constructors,
896 /// this checks for inclusion.
897 fn is_covered_by<'p>(&self, pcx: PatCtxt<'_, 'p, 'tcx>, other: &Self) -> bool {
898 // This must be kept in sync with `is_covered_by_any`.
899 match (self, other) {
900 // Wildcards cover anything
901 (_, Wildcard) => true,
902 // Wildcards are only covered by wildcards
903 (Wildcard, _) => false,
905 (Single, Single) => true,
906 (Variant(self_id), Variant(other_id)) => self_id == other_id,
908 (IntRange(self_range), IntRange(other_range)) => {
909 self_range.is_covered_by(pcx, other_range)
912 FloatRange(self_from, self_to, self_end),
913 FloatRange(other_from, other_to, other_end),
916 compare_const_vals(pcx.cx.tcx, self_to, other_to, pcx.cx.param_env, pcx.ty),
917 compare_const_vals(pcx.cx.tcx, self_from, other_from, pcx.cx.param_env, pcx.ty),
919 (Some(to), Some(from)) => {
920 (from == Ordering::Greater || from == Ordering::Equal)
921 && (to == Ordering::Less
922 || (other_end == self_end && to == Ordering::Equal))
927 (Str(self_val), Str(other_val)) => {
928 // FIXME: there's probably a more direct way of comparing for equality
929 match compare_const_vals(pcx.cx.tcx, self_val, other_val, pcx.cx.param_env, pcx.ty)
931 Some(comparison) => comparison == Ordering::Equal,
935 (Slice(self_slice), Slice(other_slice)) => self_slice.is_covered_by(*other_slice),
937 // We are trying to inspect an opaque constant. Thus we skip the row.
938 (Opaque, _) | (_, Opaque) => false,
939 // Only a wildcard pattern can match the special extra constructor.
940 (NonExhaustive, _) => false,
944 "trying to compare incompatible constructors {:?} and {:?}",
951 /// Faster version of `is_covered_by` when applied to many constructors. `used_ctors` is
952 /// assumed to be built from `matrix.head_ctors()` with wildcards filtered out, and `self` is
953 /// assumed to have been split from a wildcard.
954 fn is_covered_by_any<'p>(
956 pcx: PatCtxt<'_, 'p, 'tcx>,
957 used_ctors: &[Constructor<'tcx>],
959 if used_ctors.is_empty() {
963 // This must be kept in sync with `is_covered_by`.
965 // If `self` is `Single`, `used_ctors` cannot contain anything else than `Single`s.
966 Single => !used_ctors.is_empty(),
967 Variant(_) => used_ctors.iter().any(|c| c == self),
968 IntRange(range) => used_ctors
970 .filter_map(|c| c.as_int_range())
971 .any(|other| range.is_covered_by(pcx, other)),
972 Slice(slice) => used_ctors
974 .filter_map(|c| c.as_slice())
975 .any(|other| slice.is_covered_by(other)),
976 // This constructor is never covered by anything else
977 NonExhaustive => false,
978 Str(..) | FloatRange(..) | Opaque | Wildcard => {
979 bug!("found unexpected ctor in all_ctors: {:?}", self)
984 /// Apply a constructor to a list of patterns, yielding a new pattern. `pats`
985 /// must have as many elements as this constructor's arity.
987 /// This is roughly the inverse of `specialize_constructor`.
990 /// `self`: `Constructor::Single`
991 /// `ty`: `(u32, u32, u32)`
992 /// `pats`: `[10, 20, _]`
993 /// returns `(10, 20, _)`
995 /// `self`: `Constructor::Variant(Option::Some)`
996 /// `ty`: `Option<bool>`
997 /// `pats`: `[false]`
998 /// returns `Some(false)`
999 fn apply<'p>(&self, pcx: PatCtxt<'_, 'p, 'tcx>, fields: Fields<'p, 'tcx>) -> Pat<'tcx> {
1000 let mut subpatterns = fields.all_patterns();
1002 let pat = match self {
1003 Single | Variant(_) => match pcx.ty.kind() {
1004 ty::Adt(..) | ty::Tuple(..) => {
1005 let subpatterns = subpatterns
1007 .map(|(i, p)| FieldPat { field: Field::new(i), pattern: p })
1010 if let ty::Adt(adt, substs) = pcx.ty.kind() {
1015 variant_index: self.variant_index_for_adt(adt),
1019 PatKind::Leaf { subpatterns }
1022 PatKind::Leaf { subpatterns }
1025 // Note: given the expansion of `&str` patterns done in `expand_pattern`, we should
1026 // be careful to reconstruct the correct constant pattern here. However a string
1027 // literal pattern will never be reported as a non-exhaustiveness witness, so we
1028 // can ignore this issue.
1029 ty::Ref(..) => PatKind::Deref { subpattern: subpatterns.next().unwrap() },
1030 ty::Slice(_) | ty::Array(..) => bug!("bad slice pattern {:?} {:?}", self, pcx.ty),
1033 Slice(slice) => match slice.kind {
1035 PatKind::Slice { prefix: subpatterns.collect(), slice: None, suffix: vec![] }
1037 VarLen(prefix, _) => {
1038 let mut prefix: Vec<_> = subpatterns.by_ref().take(prefix as usize).collect();
1039 if slice.array_len.is_some() {
1040 // Improves diagnostics a bit: if the type is a known-size array, instead
1041 // of reporting `[x, _, .., _, y]`, we prefer to report `[x, .., y]`.
1042 // This is incorrect if the size is not known, since `[_, ..]` captures
1043 // arrays of lengths `>= 1` whereas `[..]` captures any length.
1044 while !prefix.is_empty() && prefix.last().unwrap().is_wildcard() {
1048 let suffix: Vec<_> = if slice.array_len.is_some() {
1050 subpatterns.skip_while(Pat::is_wildcard).collect()
1052 subpatterns.collect()
1054 let wild = Pat::wildcard_from_ty(pcx.ty);
1055 PatKind::Slice { prefix, slice: Some(wild), suffix }
1058 &Str(value) => PatKind::Constant { value },
1059 &FloatRange(lo, hi, end) => PatKind::Range(PatRange { lo, hi, end }),
1060 IntRange(range) => return range.to_pat(pcx.cx.tcx),
1061 NonExhaustive => PatKind::Wild,
1062 Opaque => bug!("we should not try to apply an opaque constructor"),
1064 "trying to apply a wildcard constructor; this should have been done in `apply_constructors`"
1068 Pat { ty: pcx.ty, span: DUMMY_SP, kind: Box::new(pat) }
1072 /// Some fields need to be explicitly hidden away in certain cases; see the comment above the
1073 /// `Fields` struct. This struct represents such a potentially-hidden field. When a field is hidden
1074 /// we still keep its type around.
1075 #[derive(Debug, Copy, Clone)]
1076 enum FilteredField<'p, 'tcx> {
1077 Kept(&'p Pat<'tcx>),
1081 impl<'p, 'tcx> FilteredField<'p, 'tcx> {
1082 fn kept(self) -> Option<&'p Pat<'tcx>> {
1084 FilteredField::Kept(p) => Some(p),
1085 FilteredField::Hidden(_) => None,
1089 fn to_pattern(self) -> Pat<'tcx> {
1091 FilteredField::Kept(p) => p.clone(),
1092 FilteredField::Hidden(ty) => Pat::wildcard_from_ty(ty),
1097 /// A value can be decomposed into a constructor applied to some fields. This struct represents
1098 /// those fields, generalized to allow patterns in each field. See also `Constructor`.
1100 /// If a private or `non_exhaustive` field is uninhabited, the code mustn't observe that it is
1101 /// uninhabited. For that, we filter these fields out of the matrix. This is subtle because we
1102 /// still need to have those fields back when going to/from a `Pat`. Most of this is handled
1103 /// automatically in `Fields`, but when constructing or deconstructing `Fields` you need to be
1104 /// careful. As a rule, when going to/from the matrix, use the filtered field list; when going
1105 /// to/from `Pat`, use the full field list.
1106 /// This filtering is uncommon in practice, because uninhabited fields are rarely used, so we avoid
1107 /// it when possible to preserve performance.
1108 #[derive(Debug, Clone)]
1109 enum Fields<'p, 'tcx> {
1110 /// Lists of patterns that don't contain any filtered fields.
1111 /// `Slice` and `Vec` behave the same; the difference is only to avoid allocating and
1112 /// triple-dereferences when possible. Frankly this is premature optimization, I (Nadrieril)
1113 /// have not measured if it really made a difference.
1114 Slice(&'p [Pat<'tcx>]),
1115 Vec(SmallVec<[&'p Pat<'tcx>; 2]>),
1116 /// Patterns where some of the fields need to be hidden. `kept_count` caches the number of
1117 /// non-hidden fields.
1119 fields: SmallVec<[FilteredField<'p, 'tcx>; 2]>,
1124 impl<'p, 'tcx> Fields<'p, 'tcx> {
1125 fn empty() -> Self {
1129 /// Construct a new `Fields` from the given pattern. Must not be used if the pattern is a field
1130 /// of a struct/tuple/variant.
1131 fn from_single_pattern(pat: &'p Pat<'tcx>) -> Self {
1132 Fields::Slice(std::slice::from_ref(pat))
1135 /// Convenience; internal use.
1136 fn wildcards_from_tys(
1137 cx: &MatchCheckCtxt<'p, 'tcx>,
1138 tys: impl IntoIterator<Item = Ty<'tcx>>,
1140 let wilds = tys.into_iter().map(Pat::wildcard_from_ty);
1141 let pats = cx.pattern_arena.alloc_from_iter(wilds);
1145 /// Creates a new list of wildcard fields for a given constructor.
1146 fn wildcards(pcx: PatCtxt<'_, 'p, 'tcx>, constructor: &Constructor<'tcx>) -> Self {
1149 let wildcard_from_ty = |ty| &*cx.pattern_arena.alloc(Pat::wildcard_from_ty(ty));
1151 let ret = match constructor {
1152 Single | Variant(_) => match ty.kind() {
1153 ty::Tuple(ref fs) => {
1154 Fields::wildcards_from_tys(cx, fs.into_iter().map(|ty| ty.expect_ty()))
1156 ty::Ref(_, rty, _) => Fields::from_single_pattern(wildcard_from_ty(rty)),
1157 ty::Adt(adt, substs) => {
1159 // Use T as the sub pattern type of Box<T>.
1160 Fields::from_single_pattern(wildcard_from_ty(substs.type_at(0)))
1162 let variant = &adt.variants[constructor.variant_index_for_adt(adt)];
1163 // Whether we must not match the fields of this variant exhaustively.
1164 let is_non_exhaustive =
1165 variant.is_field_list_non_exhaustive() && !adt.did.is_local();
1166 let field_tys = variant.fields.iter().map(|field| field.ty(cx.tcx, substs));
1167 // In the following cases, we don't need to filter out any fields. This is
1168 // the vast majority of real cases, since uninhabited fields are uncommon.
1169 let has_no_hidden_fields = (adt.is_enum() && !is_non_exhaustive)
1170 || !field_tys.clone().any(|ty| cx.is_uninhabited(ty));
1172 if has_no_hidden_fields {
1173 Fields::wildcards_from_tys(cx, field_tys)
1175 let mut kept_count = 0;
1176 let fields = variant
1180 let ty = field.ty(cx.tcx, substs);
1181 let is_visible = adt.is_enum()
1182 || field.vis.is_accessible_from(cx.module, cx.tcx);
1183 let is_uninhabited = cx.is_uninhabited(ty);
1185 // In the cases of either a `#[non_exhaustive]` field list
1186 // or a non-public field, we hide uninhabited fields in
1187 // order not to reveal the uninhabitedness of the whole
1189 if is_uninhabited && (!is_visible || is_non_exhaustive) {
1190 FilteredField::Hidden(ty)
1193 FilteredField::Kept(wildcard_from_ty(ty))
1197 Fields::Filtered { fields, kept_count }
1201 _ => bug!("Unexpected type for `Single` constructor: {:?}", ty),
1203 Slice(slice) => match *ty.kind() {
1204 ty::Slice(ty) | ty::Array(ty, _) => {
1205 let arity = slice.arity();
1206 Fields::wildcards_from_tys(cx, (0..arity).map(|_| ty))
1208 _ => bug!("bad slice pattern {:?} {:?}", constructor, ty),
1210 Str(..) | FloatRange(..) | IntRange(..) | NonExhaustive | Opaque | Wildcard => {
1214 debug!("Fields::wildcards({:?}, {:?}) = {:#?}", constructor, ty, ret);
1218 /// Returns the number of patterns from the viewpoint of match-checking, i.e. excluding hidden
1219 /// fields. This is what we want in most cases in this file, the only exception being
1220 /// conversion to/from `Pat`.
1221 fn len(&self) -> usize {
1223 Fields::Slice(pats) => pats.len(),
1224 Fields::Vec(pats) => pats.len(),
1225 Fields::Filtered { kept_count, .. } => *kept_count,
1229 /// Returns the complete list of patterns, including hidden fields.
1230 fn all_patterns(self) -> impl Iterator<Item = Pat<'tcx>> {
1231 let pats: SmallVec<[_; 2]> = match self {
1232 Fields::Slice(pats) => pats.iter().cloned().collect(),
1233 Fields::Vec(pats) => pats.into_iter().cloned().collect(),
1234 Fields::Filtered { fields, .. } => {
1235 // We don't skip any fields here.
1236 fields.into_iter().map(|p| p.to_pattern()).collect()
1242 /// Overrides some of the fields with the provided patterns. Exactly like
1243 /// `replace_fields_indexed`, except that it takes `FieldPat`s as input.
1244 fn replace_with_fieldpats(
1246 new_pats: impl IntoIterator<Item = &'p FieldPat<'tcx>>,
1248 self.replace_fields_indexed(
1249 new_pats.into_iter().map(|pat| (pat.field.index(), &pat.pattern)),
1253 /// Overrides some of the fields with the provided patterns. This is used when a pattern
1254 /// defines some fields but not all, for example `Foo { field1: Some(_), .. }`: here we start with a
1255 /// `Fields` that is just one wildcard per field of the `Foo` struct, and override the entry
1256 /// corresponding to `field1` with the pattern `Some(_)`. This is also used for slice patterns
1257 /// for the same reason.
1258 fn replace_fields_indexed(
1260 new_pats: impl IntoIterator<Item = (usize, &'p Pat<'tcx>)>,
1262 let mut fields = self.clone();
1263 if let Fields::Slice(pats) = fields {
1264 fields = Fields::Vec(pats.iter().collect());
1268 Fields::Vec(pats) => {
1269 for (i, pat) in new_pats {
1273 Fields::Filtered { fields, .. } => {
1274 for (i, pat) in new_pats {
1275 if let FilteredField::Kept(p) = &mut fields[i] {
1280 Fields::Slice(_) => unreachable!(),
1285 /// Replaces contained fields with the given filtered list of patterns, e.g. taken from the
1286 /// matrix. There must be `len()` patterns in `pats`.
1289 cx: &MatchCheckCtxt<'p, 'tcx>,
1290 pats: impl IntoIterator<Item = Pat<'tcx>>,
1292 let pats: &[_] = cx.pattern_arena.alloc_from_iter(pats);
1295 Fields::Filtered { fields, kept_count } => {
1296 let mut pats = pats.iter();
1297 let mut fields = fields.clone();
1298 for f in &mut fields {
1299 if let FilteredField::Kept(p) = f {
1300 // We take one input pattern for each `Kept` field, in order.
1301 *p = pats.next().unwrap();
1304 Fields::Filtered { fields, kept_count: *kept_count }
1306 _ => Fields::Slice(pats),
1310 /// Replaces contained fields with the arguments of the given pattern. Only use on a pattern
1311 /// that is compatible with the constructor used to build `self`.
1312 /// This is meant to be used on the result of `Fields::wildcards()`. The idea is that
1313 /// `wildcards` constructs a list of fields where all entries are wildcards, and the pattern
1314 /// provided to this function fills some of the fields with non-wildcards.
1315 /// In the following example `Fields::wildcards` would return `[_, _, _, _]`. If we call
1316 /// `replace_with_pattern_arguments` on it with the pattern, the result will be `[Some(0), _,
1319 /// let x: [Option<u8>; 4] = foo();
1321 /// [Some(0), ..] => {}
1324 /// This is guaranteed to preserve the number of patterns in `self`.
1325 fn replace_with_pattern_arguments(&self, pat: &'p Pat<'tcx>) -> Self {
1326 match pat.kind.as_ref() {
1327 PatKind::Deref { subpattern } => {
1328 assert_eq!(self.len(), 1);
1329 Fields::from_single_pattern(subpattern)
1331 PatKind::Leaf { subpatterns } | PatKind::Variant { subpatterns, .. } => {
1332 self.replace_with_fieldpats(subpatterns)
1334 PatKind::Array { prefix, suffix, .. } | PatKind::Slice { prefix, suffix, .. } => {
1335 // Number of subpatterns for the constructor
1336 let ctor_arity = self.len();
1338 // Replace the prefix and the suffix with the given patterns, leaving wildcards in
1339 // the middle if there was a subslice pattern `..`.
1340 let prefix = prefix.iter().enumerate();
1342 suffix.iter().enumerate().map(|(i, p)| (ctor_arity - suffix.len() + i, p));
1343 self.replace_fields_indexed(prefix.chain(suffix))
1349 fn push_on_patstack(self, stack: &[&'p Pat<'tcx>]) -> PatStack<'p, 'tcx> {
1350 let pats: SmallVec<_> = match self {
1351 Fields::Slice(pats) => pats.iter().chain(stack.iter().copied()).collect(),
1352 Fields::Vec(mut pats) => {
1353 pats.extend_from_slice(stack);
1356 Fields::Filtered { fields, .. } => {
1357 // We skip hidden fields here
1358 fields.into_iter().filter_map(|p| p.kept()).chain(stack.iter().copied()).collect()
1361 PatStack::from_vec(pats)
1365 #[derive(Clone, Debug)]
1366 crate enum Usefulness<'tcx> {
1367 /// Carries a list of unreachable subpatterns. Used only in the presence of or-patterns.
1369 /// Carries a list of witnesses of non-exhaustiveness.
1370 UsefulWithWitness(Vec<Witness<'tcx>>),
1374 impl<'tcx> Usefulness<'tcx> {
1375 fn new_useful(preference: WitnessPreference) -> Self {
1377 ConstructWitness => UsefulWithWitness(vec![Witness(vec![])]),
1378 LeaveOutWitness => Useful(vec![]),
1382 fn is_useful(&self) -> bool {
1389 fn apply_constructor<'p>(
1391 pcx: PatCtxt<'_, 'p, 'tcx>,
1392 ctor: &Constructor<'tcx>,
1393 ctor_wild_subpatterns: &Fields<'p, 'tcx>,
1397 UsefulWithWitness(witnesses) => {
1398 let new_witnesses = if ctor.is_wildcard() {
1399 let missing_ctors = MissingConstructors::new(pcx);
1400 let new_patterns = missing_ctors.report_patterns(pcx, is_top_level);
1403 .flat_map(|witness| {
1404 new_patterns.iter().map(move |pat| {
1405 let mut witness = witness.clone();
1406 witness.0.push(pat.clone());
1414 .map(|witness| witness.apply_constructor(pcx, &ctor, ctor_wild_subpatterns))
1417 UsefulWithWitness(new_witnesses)
1424 #[derive(Copy, Clone, Debug)]
1425 crate enum WitnessPreference {
1430 #[derive(Copy, Clone)]
1431 struct PatCtxt<'a, 'p, 'tcx> {
1432 cx: &'a MatchCheckCtxt<'p, 'tcx>,
1433 /// Current state of the matrix.
1434 matrix: &'a Matrix<'p, 'tcx>,
1435 /// Type of the current column under investigation.
1437 /// Span of the current pattern under investigation.
1441 /// A witness of non-exhaustiveness for error reporting, represented
1442 /// as a list of patterns (in reverse order of construction) with
1443 /// wildcards inside to represent elements that can take any inhabitant
1444 /// of the type as a value.
1446 /// A witness against a list of patterns should have the same types
1447 /// and length as the pattern matched against. Because Rust `match`
1448 /// is always against a single pattern, at the end the witness will
1449 /// have length 1, but in the middle of the algorithm, it can contain
1450 /// multiple patterns.
1452 /// For example, if we are constructing a witness for the match against
1455 /// struct Pair(Option<(u32, u32)>, bool);
1457 /// match (p: Pair) {
1458 /// Pair(None, _) => {}
1459 /// Pair(_, false) => {}
1463 /// We'll perform the following steps:
1464 /// 1. Start with an empty witness
1465 /// `Witness(vec![])`
1466 /// 2. Push a witness `Some(_)` against the `None`
1467 /// `Witness(vec![Some(_)])`
1468 /// 3. Push a witness `true` against the `false`
1469 /// `Witness(vec![Some(_), true])`
1470 /// 4. Apply the `Pair` constructor to the witnesses
1471 /// `Witness(vec![Pair(Some(_), true)])`
1473 /// The final `Pair(Some(_), true)` is then the resulting witness.
1474 #[derive(Clone, Debug)]
1475 crate struct Witness<'tcx>(Vec<Pat<'tcx>>);
1477 impl<'tcx> Witness<'tcx> {
1478 crate fn single_pattern(self) -> Pat<'tcx> {
1479 assert_eq!(self.0.len(), 1);
1480 self.0.into_iter().next().unwrap()
1483 /// Constructs a partial witness for a pattern given a list of
1484 /// patterns expanded by the specialization step.
1486 /// When a pattern P is discovered to be useful, this function is used bottom-up
1487 /// to reconstruct a complete witness, e.g., a pattern P' that covers a subset
1488 /// of values, V, where each value in that set is not covered by any previously
1489 /// used patterns and is covered by the pattern P'. Examples:
1491 /// left_ty: tuple of 3 elements
1492 /// pats: [10, 20, _] => (10, 20, _)
1494 /// left_ty: struct X { a: (bool, &'static str), b: usize}
1495 /// pats: [(false, "foo"), 42] => X { a: (false, "foo"), b: 42 }
1496 fn apply_constructor<'p>(
1498 pcx: PatCtxt<'_, 'p, 'tcx>,
1499 ctor: &Constructor<'tcx>,
1500 ctor_wild_subpatterns: &Fields<'p, 'tcx>,
1503 let len = self.0.len();
1504 let arity = ctor_wild_subpatterns.len();
1505 let pats = self.0.drain((len - arity)..).rev();
1506 let fields = ctor_wild_subpatterns.replace_fields(pcx.cx, pats);
1507 ctor.apply(pcx, fields)
1516 /// This determines the set of all possible constructors of a pattern matching
1517 /// values of type `left_ty`. For vectors, this would normally be an infinite set
1518 /// but is instead bounded by the maximum fixed length of slice patterns in
1519 /// the column of patterns being analyzed.
1521 /// We make sure to omit constructors that are statically impossible. E.g., for
1522 /// `Option<!>`, we do not include `Some(_)` in the returned list of constructors.
1523 /// Invariant: this returns an empty `Vec` if and only if the type is uninhabited (as determined by
1524 /// `cx.is_uninhabited()`).
1525 fn all_constructors<'p, 'tcx>(pcx: PatCtxt<'_, 'p, 'tcx>) -> Vec<Constructor<'tcx>> {
1526 debug!("all_constructors({:?})", pcx.ty);
1528 let make_range = |start, end| {
1530 // `unwrap()` is ok because we know the type is an integer.
1531 IntRange::from_range(cx.tcx, start, end, pcx.ty, &RangeEnd::Included, pcx.span)
1535 match pcx.ty.kind() {
1536 ty::Bool => vec![make_range(0, 1)],
1537 ty::Array(sub_ty, len) if len.try_eval_usize(cx.tcx, cx.param_env).is_some() => {
1538 let len = len.eval_usize(cx.tcx, cx.param_env);
1539 if len != 0 && cx.is_uninhabited(sub_ty) {
1542 vec![Slice(Slice::new(Some(len), VarLen(0, 0)))]
1545 // Treat arrays of a constant but unknown length like slices.
1546 ty::Array(sub_ty, _) | ty::Slice(sub_ty) => {
1547 let kind = if cx.is_uninhabited(sub_ty) { FixedLen(0) } else { VarLen(0, 0) };
1548 vec![Slice(Slice::new(None, kind))]
1550 ty::Adt(def, substs) if def.is_enum() => {
1551 // If the enum is declared as `#[non_exhaustive]`, we treat it as if it had an
1552 // additional "unknown" constructor.
1553 // There is no point in enumerating all possible variants, because the user can't
1554 // actually match against them all themselves. So we always return only the fictitious
1556 // E.g., in an example like:
1559 // let err: io::ErrorKind = ...;
1561 // io::ErrorKind::NotFound => {},
1565 // we don't want to show every possible IO error, but instead have only `_` as the
1567 let is_declared_nonexhaustive = cx.is_foreign_non_exhaustive_enum(pcx.ty);
1569 // If `exhaustive_patterns` is disabled and our scrutinee is an empty enum, we treat it
1570 // as though it had an "unknown" constructor to avoid exposing its emptyness. Note that
1571 // an empty match will still be considered exhaustive because that case is handled
1572 // separately in `check_match`.
1573 let is_secretly_empty =
1574 def.variants.is_empty() && !cx.tcx.features().exhaustive_patterns;
1576 if is_secretly_empty || is_declared_nonexhaustive {
1578 } else if cx.tcx.features().exhaustive_patterns {
1579 // If `exhaustive_patterns` is enabled, we exclude variants known to be
1584 !v.uninhabited_from(cx.tcx, substs, def.adt_kind(), cx.param_env)
1585 .contains(cx.tcx, cx.module)
1587 .map(|v| Variant(v.def_id))
1590 def.variants.iter().map(|v| Variant(v.def_id)).collect()
1595 // The valid Unicode Scalar Value ranges.
1596 make_range('\u{0000}' as u128, '\u{D7FF}' as u128),
1597 make_range('\u{E000}' as u128, '\u{10FFFF}' as u128),
1600 ty::Int(_) | ty::Uint(_)
1601 if pcx.ty.is_ptr_sized_integral()
1602 && !cx.tcx.features().precise_pointer_size_matching =>
1604 // `usize`/`isize` are not allowed to be matched exhaustively unless the
1605 // `precise_pointer_size_matching` feature is enabled. So we treat those types like
1606 // `#[non_exhaustive]` enums by returning a special unmatcheable constructor.
1610 let bits = Integer::from_attr(&cx.tcx, SignedInt(ity)).size().bits() as u128;
1611 let min = 1u128 << (bits - 1);
1613 vec![make_range(min, max)]
1616 let size = Integer::from_attr(&cx.tcx, UnsignedInt(uty)).size();
1617 let max = truncate(u128::MAX, size);
1618 vec![make_range(0, max)]
1620 _ if cx.is_uninhabited(pcx.ty) => vec![],
1621 ty::Adt(..) | ty::Tuple(..) | ty::Ref(..) => vec![Single],
1622 // This type is one for which we cannot list constructors, like `str` or `f64`.
1623 _ => vec![NonExhaustive],
1627 /// An inclusive interval, used for precise integer exhaustiveness checking.
1628 /// `IntRange`s always store a contiguous range. This means that values are
1629 /// encoded such that `0` encodes the minimum value for the integer,
1630 /// regardless of the signedness.
1631 /// For example, the pattern `-128..=127i8` is encoded as `0..=255`.
1632 /// This makes comparisons and arithmetic on interval endpoints much more
1633 /// straightforward. See `signed_bias` for details.
1635 /// `IntRange` is never used to encode an empty range or a "range" that wraps
1636 /// around the (offset) space: i.e., `range.lo <= range.hi`.
1637 #[derive(Clone, Debug)]
1638 struct IntRange<'tcx> {
1639 range: RangeInclusive<u128>,
1644 impl<'tcx> IntRange<'tcx> {
1646 fn is_integral(ty: Ty<'_>) -> bool {
1648 ty::Char | ty::Int(_) | ty::Uint(_) | ty::Bool => true,
1653 fn is_singleton(&self) -> bool {
1654 self.range.start() == self.range.end()
1657 fn boundaries(&self) -> (u128, u128) {
1658 (*self.range.start(), *self.range.end())
1661 /// Don't treat `usize`/`isize` exhaustively unless the `precise_pointer_size_matching` feature
1663 fn treat_exhaustively(&self, tcx: TyCtxt<'tcx>) -> bool {
1664 !self.ty.is_ptr_sized_integral() || tcx.features().precise_pointer_size_matching
1668 fn integral_size_and_signed_bias(tcx: TyCtxt<'tcx>, ty: Ty<'_>) -> Option<(Size, u128)> {
1670 ty::Bool => Some((Size::from_bytes(1), 0)),
1671 ty::Char => Some((Size::from_bytes(4), 0)),
1673 let size = Integer::from_attr(&tcx, SignedInt(ity)).size();
1674 Some((size, 1u128 << (size.bits() as u128 - 1)))
1676 ty::Uint(uty) => Some((Integer::from_attr(&tcx, UnsignedInt(uty)).size(), 0)),
1684 param_env: ty::ParamEnv<'tcx>,
1685 value: &Const<'tcx>,
1687 ) -> Option<IntRange<'tcx>> {
1688 if let Some((target_size, bias)) = Self::integral_size_and_signed_bias(tcx, value.ty) {
1691 if let ty::ConstKind::Value(ConstValue::Scalar(scalar)) = value.val {
1692 // For this specific pattern we can skip a lot of effort and go
1693 // straight to the result, after doing a bit of checking. (We
1694 // could remove this branch and just fall through, which
1695 // is more general but much slower.)
1696 if let Ok(bits) = scalar.to_bits_or_ptr(target_size, &tcx) {
1700 // This is a more general form of the previous case.
1701 value.try_eval_bits(tcx, param_env, ty)
1703 let val = val ^ bias;
1704 Some(IntRange { range: val..=val, ty, span })
1718 ) -> Option<IntRange<'tcx>> {
1719 if Self::is_integral(ty) {
1720 // Perform a shift if the underlying types are signed,
1721 // which makes the interval arithmetic simpler.
1722 let bias = IntRange::signed_bias(tcx, ty);
1723 let (lo, hi) = (lo ^ bias, hi ^ bias);
1724 let offset = (*end == RangeEnd::Excluded) as u128;
1725 if lo > hi || (lo == hi && *end == RangeEnd::Excluded) {
1726 // This should have been caught earlier by E0030.
1727 bug!("malformed range pattern: {}..={}", lo, (hi - offset));
1729 Some(IntRange { range: lo..=(hi - offset), ty, span })
1735 // The return value of `signed_bias` should be XORed with an endpoint to encode/decode it.
1736 fn signed_bias(tcx: TyCtxt<'tcx>, ty: Ty<'tcx>) -> u128 {
1739 let bits = Integer::from_attr(&tcx, SignedInt(ity)).size().bits() as u128;
1746 fn is_subrange(&self, other: &Self) -> bool {
1747 other.range.start() <= self.range.start() && self.range.end() <= other.range.end()
1750 fn intersection(&self, tcx: TyCtxt<'tcx>, other: &Self) -> Option<Self> {
1752 let (lo, hi) = self.boundaries();
1753 let (other_lo, other_hi) = other.boundaries();
1754 if self.treat_exhaustively(tcx) {
1755 if lo <= other_hi && other_lo <= hi {
1756 let span = other.span;
1757 Some(IntRange { range: max(lo, other_lo)..=min(hi, other_hi), ty, span })
1762 // If the range should not be treated exhaustively, fallback to checking for inclusion.
1763 if self.is_subrange(other) { Some(self.clone()) } else { None }
1767 fn suspicious_intersection(&self, other: &Self) -> bool {
1768 // `false` in the following cases:
1769 // 1 ---- // 1 ---------- // 1 ---- // 1 ----
1770 // 2 ---------- // 2 ---- // 2 ---- // 2 ----
1772 // The following are currently `false`, but could be `true` in the future (#64007):
1773 // 1 --------- // 1 ---------
1774 // 2 ---------- // 2 ----------
1776 // `true` in the following cases:
1777 // 1 ------- // 1 -------
1778 // 2 -------- // 2 -------
1779 let (lo, hi) = self.boundaries();
1780 let (other_lo, other_hi) = other.boundaries();
1781 lo == other_hi || hi == other_lo
1784 fn to_pat(&self, tcx: TyCtxt<'tcx>) -> Pat<'tcx> {
1785 let (lo, hi) = self.boundaries();
1787 let bias = IntRange::signed_bias(tcx, self.ty);
1788 let (lo, hi) = (lo ^ bias, hi ^ bias);
1790 let ty = ty::ParamEnv::empty().and(self.ty);
1791 let lo_const = ty::Const::from_bits(tcx, lo, ty);
1792 let hi_const = ty::Const::from_bits(tcx, hi, ty);
1794 let kind = if lo == hi {
1795 PatKind::Constant { value: lo_const }
1797 PatKind::Range(PatRange { lo: lo_const, hi: hi_const, end: RangeEnd::Included })
1800 // This is a brand new pattern, so we don't reuse `self.span`.
1801 Pat { ty: self.ty, span: DUMMY_SP, kind: Box::new(kind) }
1804 /// For exhaustive integer matching, some constructors are grouped within other constructors
1805 /// (namely integer typed values are grouped within ranges). However, when specialising these
1806 /// constructors, we want to be specialising for the underlying constructors (the integers), not
1807 /// the groups (the ranges). Thus we need to split the groups up. Splitting them up naïvely would
1808 /// mean creating a separate constructor for every single value in the range, which is clearly
1809 /// impractical. However, observe that for some ranges of integers, the specialisation will be
1810 /// identical across all values in that range (i.e., there are equivalence classes of ranges of
1811 /// constructors based on their `U(S(c, P), S(c, p))` outcome). These classes are grouped by
1812 /// the patterns that apply to them (in the matrix `P`). We can split the range whenever the
1813 /// patterns that apply to that range (specifically: the patterns that *intersect* with that range)
1815 /// Our solution, therefore, is to split the range constructor into subranges at every single point
1816 /// the group of intersecting patterns changes (using the method described below).
1817 /// And voilà! We're testing precisely those ranges that we need to, without any exhaustive matching
1818 /// on actual integers. The nice thing about this is that the number of subranges is linear in the
1819 /// number of rows in the matrix (i.e., the number of cases in the `match` statement), so we don't
1820 /// need to be worried about matching over gargantuan ranges.
1822 /// Essentially, given the first column of a matrix representing ranges, looking like the following:
1824 /// |------| |----------| |-------| ||
1825 /// |-------| |-------| |----| ||
1828 /// We split the ranges up into equivalence classes so the ranges are no longer overlapping:
1830 /// |--|--|||-||||--||---|||-------| |-|||| ||
1832 /// The logic for determining how to split the ranges is fairly straightforward: we calculate
1833 /// boundaries for each interval range, sort them, then create constructors for each new interval
1834 /// between every pair of boundary points. (This essentially sums up to performing the intuitive
1835 /// merging operation depicted above.)
1838 pcx: PatCtxt<'_, 'p, 'tcx>,
1839 hir_id: Option<HirId>,
1840 ) -> SmallVec<[Constructor<'tcx>; 1]> {
1843 /// Represents a border between 2 integers. Because the intervals spanning borders
1844 /// must be able to cover every integer, we need to be able to represent
1845 /// 2^128 + 1 such borders.
1846 #[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Debug)]
1852 // A function for extracting the borders of an integer interval.
1853 fn range_borders(r: IntRange<'_>) -> impl Iterator<Item = Border> {
1854 let (lo, hi) = r.range.into_inner();
1855 let from = Border::JustBefore(lo);
1856 let to = match hi.checked_add(1) {
1857 Some(m) => Border::JustBefore(m),
1858 None => Border::AfterMax,
1860 vec![from, to].into_iter()
1863 // Collect the span and range of all the intersecting ranges to lint on likely
1864 // incorrect range patterns. (#63987)
1865 let mut overlaps = vec![];
1866 let row_len = pcx.matrix.patterns.get(0).map(|r| r.len()).unwrap_or(0);
1867 // `borders` is the set of borders between equivalence classes: each equivalence
1868 // class lies between 2 borders.
1869 let row_borders = pcx
1872 .filter_map(|ctor| ctor.as_int_range())
1873 .filter_map(|range| {
1874 let intersection = self.intersection(pcx.cx.tcx, &range);
1875 let should_lint = self.suspicious_intersection(&range);
1876 if let (Some(range), 1, true) = (&intersection, row_len, should_lint) {
1877 // FIXME: for now, only check for overlapping ranges on simple range
1878 // patterns. Otherwise with the current logic the following is detected
1880 // match (10u8, true) {
1881 // (0 ..= 125, false) => {}
1882 // (126 ..= 255, false) => {}
1883 // (0 ..= 255, true) => {}
1885 overlaps.push(range.clone());
1889 .flat_map(range_borders);
1890 let self_borders = range_borders(self.clone());
1891 let mut borders: Vec<_> = row_borders.chain(self_borders).collect();
1892 borders.sort_unstable();
1894 self.lint_overlapping_patterns(pcx.cx.tcx, hir_id, ty, overlaps);
1896 // We're going to iterate through every adjacent pair of borders, making sure that
1897 // each represents an interval of nonnegative length, and convert each such
1898 // interval into a constructor.
1901 .filter_map(|&pair| match pair {
1902 [Border::JustBefore(n), Border::JustBefore(m)] => {
1909 [Border::JustBefore(n), Border::AfterMax] => Some(n..=u128::MAX),
1910 [Border::AfterMax, _] => None,
1912 .map(|range| IntRange { range, ty, span: pcx.span })
1917 fn lint_overlapping_patterns(
1920 hir_id: Option<HirId>,
1922 overlaps: Vec<IntRange<'tcx>>,
1924 if let (true, Some(hir_id)) = (!overlaps.is_empty(), hir_id) {
1925 tcx.struct_span_lint_hir(
1926 lint::builtin::OVERLAPPING_PATTERNS,
1930 let mut err = lint.build("multiple patterns covering the same range");
1931 err.span_label(self.span, "overlapping patterns");
1932 for int_range in overlaps {
1933 // Use the real type for user display of the ranges:
1937 "this range overlaps on `{}`",
1938 IntRange { range: int_range.range, ty, span: DUMMY_SP }.to_pat(tcx),
1948 /// See `Constructor::is_covered_by`
1949 fn is_covered_by<'p>(&self, pcx: PatCtxt<'_, 'p, 'tcx>, other: &Self) -> bool {
1950 if self.intersection(pcx.cx.tcx, other).is_some() {
1951 // Constructor splitting should ensure that all intersections we encounter are actually
1953 assert!(self.is_subrange(other));
1961 /// Ignore spans when comparing, they don't carry semantic information as they are only for lints.
1962 impl<'tcx> std::cmp::PartialEq for IntRange<'tcx> {
1963 fn eq(&self, other: &Self) -> bool {
1964 self.range == other.range && self.ty == other.ty
1968 // A struct to compute a set of constructors equivalent to `all_ctors \ used_ctors`.
1970 struct MissingConstructors<'tcx> {
1971 all_ctors: SmallVec<[Constructor<'tcx>; 1]>,
1972 used_ctors: Vec<Constructor<'tcx>>,
1975 impl<'tcx> MissingConstructors<'tcx> {
1976 fn new<'p>(pcx: PatCtxt<'_, 'p, 'tcx>) -> Self {
1977 let used_ctors: Vec<Constructor<'_>> =
1978 pcx.matrix.head_ctors(pcx.cx).cloned().filter(|c| !c.is_wildcard()).collect();
1979 // Since `all_ctors` never contains wildcards, this won't recurse further.
1981 all_constructors(pcx).into_iter().flat_map(|ctor| ctor.split(pcx, None)).collect();
1983 MissingConstructors { all_ctors, used_ctors }
1986 fn is_empty<'p>(&self, pcx: PatCtxt<'_, 'p, 'tcx>) -> bool {
1987 self.iter(pcx).next().is_none()
1990 /// Iterate over all_ctors \ used_ctors
1993 pcx: PatCtxt<'a, 'p, 'tcx>,
1994 ) -> impl Iterator<Item = &'a Constructor<'tcx>> + Captures<'p> {
1995 self.all_ctors.iter().filter(move |ctor| !ctor.is_covered_by_any(pcx, &self.used_ctors))
1998 /// List the patterns corresponding to the missing constructors. In some cases, instead of
1999 /// listing all constructors of a given type, we prefer to simply report a wildcard.
2000 fn report_patterns<'p>(
2002 pcx: PatCtxt<'_, 'p, 'tcx>,
2004 ) -> SmallVec<[Pat<'tcx>; 1]> {
2005 // There are 2 ways we can report a witness here.
2006 // Commonly, we can report all the "free"
2007 // constructors as witnesses, e.g., if we have:
2010 // enum Direction { N, S, E, W }
2011 // let Direction::N = ...;
2014 // we can report 3 witnesses: `S`, `E`, and `W`.
2016 // However, there is a case where we don't want
2017 // to do this and instead report a single `_` witness:
2018 // if the user didn't actually specify a constructor
2019 // in this arm, e.g., in
2022 // let x: (Direction, Direction, bool) = ...;
2023 // let (_, _, false) = x;
2026 // we don't want to show all 16 possible witnesses
2027 // `(<direction-1>, <direction-2>, true)` - we are
2028 // satisfied with `(_, _, true)`. In this case,
2029 // `used_ctors` is empty.
2030 // The exception is: if we are at the top-level, for example in an empty match, we
2031 // sometimes prefer reporting the list of constructors instead of just `_`.
2032 let report_when_all_missing = is_top_level && !IntRange::is_integral(pcx.ty);
2033 if self.used_ctors.is_empty() && !report_when_all_missing {
2034 // All constructors are unused. Report only a wildcard
2035 // rather than each individual constructor.
2036 smallvec![Pat::wildcard_from_ty(pcx.ty)]
2038 // Construct for each missing constructor a "wild" version of this
2039 // constructor, that matches everything that can be built with
2040 // it. For example, if `ctor` is a `Constructor::Variant` for
2041 // `Option::Some`, we get the pattern `Some(_)`.
2043 .map(|missing_ctor| {
2044 let fields = Fields::wildcards(pcx, &missing_ctor);
2045 missing_ctor.apply(pcx, fields)
2052 /// Algorithm from http://moscova.inria.fr/~maranget/papers/warn/index.html.
2053 /// The algorithm from the paper has been modified to correctly handle empty
2054 /// types. The changes are:
2055 /// (0) We don't exit early if the pattern matrix has zero rows. We just
2056 /// continue to recurse over columns.
2057 /// (1) all_constructors will only return constructors that are statically
2058 /// possible. E.g., it will only return `Ok` for `Result<T, !>`.
2060 /// This finds whether a (row) vector `v` of patterns is 'useful' in relation
2061 /// to a set of such vectors `m` - this is defined as there being a set of
2062 /// inputs that will match `v` but not any of the sets in `m`.
2064 /// All the patterns at each column of the `matrix ++ v` matrix must have the same type.
2066 /// This is used both for reachability checking (if a pattern isn't useful in
2067 /// relation to preceding patterns, it is not reachable) and exhaustiveness
2068 /// checking (if a wildcard pattern is useful in relation to a matrix, the
2069 /// matrix isn't exhaustive).
2071 /// `is_under_guard` is used to inform if the pattern has a guard. If it
2072 /// has one it must not be inserted into the matrix. This shouldn't be
2073 /// relied on for soundness.
2074 crate fn is_useful<'p, 'tcx>(
2075 cx: &MatchCheckCtxt<'p, 'tcx>,
2076 matrix: &Matrix<'p, 'tcx>,
2077 v: &PatStack<'p, 'tcx>,
2078 witness_preference: WitnessPreference,
2080 is_under_guard: bool,
2082 ) -> Usefulness<'tcx> {
2083 let Matrix { patterns: rows, .. } = matrix;
2084 debug!("is_useful({:#?}, {:#?})", matrix, v);
2086 // The base case. We are pattern-matching on () and the return value is
2087 // based on whether our matrix has a row or not.
2088 // NOTE: This could potentially be optimized by checking rows.is_empty()
2089 // first and then, if v is non-empty, the return value is based on whether
2090 // the type of the tuple we're checking is inhabited or not.
2092 return if rows.is_empty() {
2093 Usefulness::new_useful(witness_preference)
2099 assert!(rows.iter().all(|r| r.len() == v.len()));
2101 // If the first pattern is an or-pattern, expand it.
2102 if let Some(vs) = v.expand_or_pat() {
2103 // We need to push the already-seen patterns into the matrix in order to detect redundant
2104 // branches like `Some(_) | Some(0)`. We also keep track of the unreachable subpatterns.
2105 let mut matrix = matrix.clone();
2106 // `Vec` of all the unreachable branches of the current or-pattern.
2107 let mut unreachable_branches = Vec::new();
2108 // Subpatterns that are unreachable from all branches. E.g. in the following case, the last
2109 // `true` is unreachable only from one branch, so it is overall reachable.
2112 // match (true, true) {
2113 // (true, true) => {}
2114 // (false | true, false | true) => {}
2117 let mut unreachable_subpats = FxHashSet::default();
2118 // Whether any branch at all is useful.
2119 let mut any_is_useful = false;
2122 let res = is_useful(cx, &matrix, &v, witness_preference, hir_id, is_under_guard, false);
2126 any_is_useful = true;
2127 // Initialize with the first set of unreachable subpatterns encountered.
2128 unreachable_subpats = pats.into_iter().collect();
2130 // Keep the patterns unreachable from both this and previous branches.
2131 unreachable_subpats =
2132 pats.into_iter().filter(|p| unreachable_subpats.contains(p)).collect();
2135 NotUseful => unreachable_branches.push(v.head().span),
2136 UsefulWithWitness(_) => {
2137 bug!("Encountered or-pat in `v` during exhaustiveness checking")
2140 // If pattern has a guard don't add it to the matrix
2141 if !is_under_guard {
2146 // Collect all the unreachable patterns.
2147 unreachable_branches.extend(unreachable_subpats);
2148 return Useful(unreachable_branches);
2154 // FIXME(Nadrieril): Hack to work around type normalization issues (see #72476).
2155 let ty = matrix.heads().next().map(|r| r.ty).unwrap_or(v.head().ty);
2156 let pcx = PatCtxt { cx, matrix, ty, span: v.head().span };
2158 debug!("is_useful_expand_first_col: ty={:#?}, expanding {:#?}", pcx.ty, v.head());
2162 .split(pcx, Some(hir_id))
2165 // We cache the result of `Fields::wildcards` because it is used a lot.
2166 let ctor_wild_subpatterns = Fields::wildcards(pcx, &ctor);
2167 let matrix = pcx.matrix.specialize_constructor(pcx, &ctor, &ctor_wild_subpatterns);
2168 let v = v.pop_head_constructor(&ctor_wild_subpatterns);
2170 is_useful(pcx.cx, &matrix, &v, witness_preference, hir_id, is_under_guard, false);
2171 usefulness.apply_constructor(pcx, &ctor, &ctor_wild_subpatterns, is_top_level)
2173 .find(|result| result.is_useful())
2174 .unwrap_or(NotUseful);
2175 debug!("is_useful::returns({:#?}, {:#?}) = {:?}", matrix, v, ret);
2179 /// Determines the constructor that the given pattern can be specialized to.
2180 /// Returns `None` in case of a catch-all, which can't be specialized.
2181 fn pat_constructor<'p, 'tcx>(
2182 cx: &MatchCheckCtxt<'p, 'tcx>,
2184 ) -> Constructor<'tcx> {
2185 match pat.kind.as_ref() {
2186 PatKind::AscribeUserType { .. } => bug!(), // Handled by `expand_pattern`
2187 PatKind::Binding { .. } | PatKind::Wild => Wildcard,
2188 PatKind::Leaf { .. } | PatKind::Deref { .. } => Single,
2189 &PatKind::Variant { adt_def, variant_index, .. } => {
2190 Variant(adt_def.variants[variant_index].def_id)
2192 PatKind::Constant { value } => {
2193 if let Some(int_range) = IntRange::from_const(cx.tcx, cx.param_env, value, pat.span) {
2196 match pat.ty.kind() {
2197 ty::Float(_) => FloatRange(value, value, RangeEnd::Included),
2198 // In `expand_pattern`, we convert string literals to `&CONST` patterns with
2199 // `CONST` a pattern of type `str`. In truth this contains a constant of type
2201 ty::Str => Str(value),
2202 // All constants that can be structurally matched have already been expanded
2203 // into the corresponding `Pat`s by `const_to_pat`. Constants that remain are
2209 &PatKind::Range(PatRange { lo, hi, end }) => {
2211 if let Some(int_range) = IntRange::from_range(
2213 lo.eval_bits(cx.tcx, cx.param_env, lo.ty),
2214 hi.eval_bits(cx.tcx, cx.param_env, hi.ty),
2221 FloatRange(lo, hi, end)
2224 PatKind::Array { prefix, slice, suffix } | PatKind::Slice { prefix, slice, suffix } => {
2225 let array_len = match pat.ty.kind() {
2226 ty::Array(_, length) => Some(length.eval_usize(cx.tcx, cx.param_env)),
2227 ty::Slice(_) => None,
2228 _ => span_bug!(pat.span, "bad ty {:?} for slice pattern", pat.ty),
2230 let prefix = prefix.len() as u64;
2231 let suffix = suffix.len() as u64;
2233 if slice.is_some() { VarLen(prefix, suffix) } else { FixedLen(prefix + suffix) };
2234 Slice(Slice::new(array_len, kind))
2236 PatKind::Or { .. } => bug!("Or-pattern should have been expanded earlier on."),