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:
82 //! // x: (Option<bool>, Result<()>)
84 //! (Some(true), _) => {}
85 //! (None, Err(())) => {}
86 //! (None, Err(_)) => {}
89 //! Here, the matrix `P` starts as:
91 //! [(Some(true), _)],
92 //! [(None, Err(()))],
95 //! We can tell it's not exhaustive, because `U(P, _)` is true (we're not covering
96 //! `[(Some(false), _)]`, for instance). In addition, row 3 is not useful, because
97 //! all the values it covers are already covered by row 2.
99 //! A list of patterns can be thought of as a stack, because we are mainly interested in the top of
100 //! the stack at any given point, and we can pop or apply constructors to get new pattern-stacks.
101 //! To match the paper, the top of the stack is at the beginning / on the left.
103 //! There are two important operations on pattern-stacks necessary to understand the algorithm:
105 //! 1. We can pop a given constructor off the top of a stack. This operation is called
106 //! `specialize`, and is denoted `S(c, p)` where `c` is a constructor (like `Some` or
107 //! `None`) and `p` a pattern-stack.
108 //! If the pattern on top of the stack can cover `c`, this removes the constructor and
109 //! pushes its arguments onto the stack. It also expands OR-patterns into distinct patterns.
110 //! Otherwise the pattern-stack is discarded.
111 //! This essentially filters those pattern-stacks whose top covers the constructor `c` and
112 //! discards the others.
114 //! For example, the first pattern above initially gives a stack `[(Some(true), _)]`. If we
115 //! pop the tuple constructor, we are left with `[Some(true), _]`, and if we then pop the
116 //! `Some` constructor we get `[true, _]`. If we had popped `None` instead, we would get
119 //! This returns zero or more new pattern-stacks, as follows. We look at the pattern `p_1`
120 //! on top of the stack, and we have four cases:
121 //! 1.1. `p_1 = c(r_1, .., r_a)`, i.e. the top of the stack has constructor `c`. We
122 //! push onto the stack the arguments of this constructor, and return the result:
123 //! r_1, .., r_a, p_2, .., p_n
124 //! 1.2. `p_1 = c'(r_1, .., r_a')` where `c ≠ c'`. We discard the current stack and
126 //! 1.3. `p_1 = _`. We push onto the stack as many wildcards as the constructor `c` has
127 //! arguments (its arity), and return the resulting stack:
128 //! _, .., _, p_2, .., p_n
129 //! 1.4. `p_1 = r_1 | r_2`. We expand the OR-pattern and then recurse on each resulting
131 //! S(c, (r_1, p_2, .., p_n))
132 //! S(c, (r_2, p_2, .., p_n))
134 //! 2. We can pop a wildcard off the top of the stack. This is called `D(p)`, where `p` is
136 //! This is used when we know there are missing constructor cases, but there might be
137 //! existing wildcard patterns, so to check the usefulness of the matrix, we have to check
138 //! all its *other* components.
140 //! It is computed as follows. We look at the pattern `p_1` on top of the stack,
141 //! and we have three cases:
142 //! 2.1. `p_1 = c(r_1, .., r_a)`. We discard the current stack and return nothing.
143 //! 2.2. `p_1 = _`. We return the rest of the stack:
145 //! 2.3. `p_1 = r_1 | r_2`. We expand the OR-pattern and then recurse on each resulting
147 //! D((r_1, p_2, .., p_n))
148 //! D((r_2, p_2, .., p_n))
150 //! Note that the OR-patterns are not always used directly in Rust, but are used to derive the
151 //! exhaustive integer matching rules, so they're written here for posterity.
153 //! Both those operations extend straightforwardly to a list or pattern-stacks, i.e. a matrix, by
154 //! working row-by-row. Popping a constructor ends up keeping only the matrix rows that start with
155 //! the given constructor, and popping a wildcard keeps those rows that start with a wildcard.
158 //! The algorithm for computing `U`
159 //! -------------------------------
160 //! The algorithm is inductive (on the number of columns: i.e., components of tuple patterns).
161 //! That means we're going to check the components from left-to-right, so the algorithm
162 //! operates principally on the first component of the matrix and new pattern-stack `p`.
163 //! This algorithm is realised in the `is_useful` function.
165 //! Base case. (`n = 0`, i.e., an empty tuple pattern)
166 //! - If `P` already contains an empty pattern (i.e., if the number of patterns `m > 0`),
167 //! then `U(P, p)` is false.
168 //! - Otherwise, `P` must be empty, so `U(P, p)` is true.
170 //! Inductive step. (`n > 0`, i.e., whether there's at least one column
171 //! [which may then be expanded into further columns later])
172 //! We're going to match on the top of the new pattern-stack, `p_1`.
173 //! - If `p_1 == c(r_1, .., r_a)`, i.e. we have a constructor pattern.
174 //! Then, the usefulness of `p_1` can be reduced to whether it is useful when
175 //! we ignore all the patterns in the first column of `P` that involve other constructors.
176 //! This is where `S(c, P)` comes in:
177 //! `U(P, p) := U(S(c, P), S(c, p))`
178 //! This special case is handled in `is_useful_specialized`.
180 //! For example, if `P` is:
185 //! and `p` is [Some(false), 0], then we don't care about row 2 since we know `p` only
186 //! matches values that row 2 doesn't. For row 1 however, we need to dig into the
187 //! arguments of `Some` to know whether some new value is covered. So we compute
188 //! `U([[true, _]], [false, 0])`.
190 //! - If `p_1 == _`, then we look at the list of constructors that appear in the first
191 //! component of the rows of `P`:
192 //! + If there are some constructors that aren't present, then we might think that the
193 //! wildcard `_` is useful, since it covers those constructors that weren't covered
195 //! That's almost correct, but only works if there were no wildcards in those first
196 //! components. So we need to check that `p` is useful with respect to the rows that
197 //! start with a wildcard, if there are any. This is where `D` comes in:
198 //! `U(P, p) := U(D(P), D(p))`
200 //! For example, if `P` is:
203 //! [None, false, 1],
205 //! and `p` is [_, false, _], the `Some` constructor doesn't appear in `P`. So if we
206 //! only had row 2, we'd know that `p` is useful. However row 1 starts with a
207 //! wildcard, so we need to check whether `U([[true, _]], [false, 1])`.
209 //! + Otherwise, all possible constructors (for the relevant type) are present. In this
210 //! case we must check whether the wildcard pattern covers any unmatched value. For
211 //! that, we can think of the `_` pattern as a big OR-pattern that covers all
212 //! possible constructors. For `Option`, that would mean `_ = None | Some(_)` for
213 //! example. The wildcard pattern is useful in this case if it is useful when
214 //! specialized to one of the possible constructors. So we compute:
215 //! `U(P, p) := ∃(k ϵ constructors) U(S(k, P), S(k, p))`
217 //! For example, if `P` is:
222 //! and `p` is [_, false], both `None` and `Some` constructors appear in the first
223 //! components of `P`. We will therefore try popping both constructors in turn: we
224 //! compute `U([[true, _]], [_, false])` for the `Some` constructor, and `U([[false]],
225 //! [false])` for the `None` constructor. The first case returns true, so we know that
226 //! `p` is useful for `P`. Indeed, it matches `[Some(false), _]` that wasn't matched
229 //! - If `p_1 == r_1 | r_2`, then the usefulness depends on each `r_i` separately:
230 //! `U(P, p) := U(P, (r_1, p_2, .., p_n))
231 //! || U(P, (r_2, p_2, .., p_n))`
233 //! Modifications to the algorithm
234 //! ------------------------------
235 //! The algorithm in the paper doesn't cover some of the special cases that arise in Rust, for
236 //! example uninhabited types and variable-length slice patterns. These are drawn attention to
237 //! throughout the code below. I'll make a quick note here about how exhaustive integer matching is
238 //! accounted for, though.
240 //! Exhaustive integer matching
241 //! ---------------------------
242 //! An integer type can be thought of as a (huge) sum type: 1 | 2 | 3 | ...
243 //! So to support exhaustive integer matching, we can make use of the logic in the paper for
244 //! OR-patterns. However, we obviously can't just treat ranges x..=y as individual sums, because
245 //! they are likely gigantic. So we instead treat ranges as constructors of the integers. This means
246 //! that we have a constructor *of* constructors (the integers themselves). We then need to work
247 //! through all the inductive step rules above, deriving how the ranges would be treated as
248 //! OR-patterns, and making sure that they're treated in the same way even when they're ranges.
249 //! There are really only four special cases here:
250 //! - When we match on a constructor that's actually a range, we have to treat it as if we would
252 //! + It turns out that we can simply extend the case for single-value patterns in
253 //! `specialize` to either be *equal* to a value constructor, or *contained within* a range
255 //! + When the pattern itself is a range, you just want to tell whether any of the values in
256 //! the pattern range coincide with values in the constructor range, which is precisely
258 //! Since when encountering a range pattern for a value constructor, we also use inclusion, it
259 //! means that whenever the constructor is a value/range and the pattern is also a value/range,
260 //! we can simply use intersection to test usefulness.
261 //! - When we're testing for usefulness of a pattern and the pattern's first component is a
263 //! + If all the constructors appear in the matrix, we have a slight complication. By default,
264 //! the behaviour (i.e., a disjunction over specialised matrices for each constructor) is
265 //! invalid, because we want a disjunction over every *integer* in each range, not just a
266 //! disjunction over every range. This is a bit more tricky to deal with: essentially we need
267 //! to form equivalence classes of subranges of the constructor range for which the behaviour
268 //! of the matrix `P` and new pattern `p` are the same. This is described in more
269 //! detail in `split_grouped_constructors`.
270 //! + If some constructors are missing from the matrix, it turns out we don't need to do
271 //! anything special (because we know none of the integers are actually wildcards: i.e., we
272 //! can't span wildcards using ranges).
273 use self::Constructor::*;
274 use self::SliceKind::*;
275 use self::Usefulness::*;
276 use self::WitnessPreference::*;
278 use rustc_data_structures::captures::Captures;
279 use rustc_data_structures::fx::{FxHashMap, FxHashSet};
280 use rustc_index::vec::Idx;
282 use super::{compare_const_vals, PatternFoldable, PatternFolder};
283 use super::{FieldPat, Pat, PatKind, PatRange};
285 use rustc_arena::TypedArena;
286 use rustc_attr::{SignedInt, UnsignedInt};
287 use rustc_errors::ErrorReported;
288 use rustc_hir::def_id::DefId;
289 use rustc_hir::{HirId, RangeEnd};
290 use rustc_middle::mir::interpret::{truncate, AllocId, ConstValue, Pointer, Scalar};
291 use rustc_middle::mir::Field;
292 use rustc_middle::ty::layout::IntegerExt;
293 use rustc_middle::ty::{self, Const, Ty, TyCtxt};
294 use rustc_session::lint;
295 use rustc_span::{Span, DUMMY_SP};
296 use rustc_target::abi::{Integer, Size, VariantIdx};
298 use smallvec::{smallvec, SmallVec};
299 use std::borrow::Cow;
300 use std::cmp::{self, max, min, Ordering};
301 use std::convert::TryInto;
303 use std::iter::{FromIterator, IntoIterator};
304 use std::ops::RangeInclusive;
306 crate fn expand_pattern<'a, 'tcx>(cx: &MatchCheckCtxt<'a, 'tcx>, pat: Pat<'tcx>) -> Pat<'tcx> {
307 LiteralExpander { tcx: cx.tcx, param_env: cx.param_env }.fold_pattern(&pat)
310 struct LiteralExpander<'tcx> {
312 param_env: ty::ParamEnv<'tcx>,
315 impl<'tcx> LiteralExpander<'tcx> {
316 /// Derefs `val` and potentially unsizes the value if `crty` is an array and `rty` a slice.
318 /// `crty` and `rty` can differ because you can use array constants in the presence of slice
319 /// patterns. So the pattern may end up being a slice, but the constant is an array. We convert
320 /// the array to a slice in that case.
321 fn fold_const_value_deref(
323 val: ConstValue<'tcx>,
324 // the pattern's pointee type
326 // the constant's pointee type
328 ) -> ConstValue<'tcx> {
329 debug!("fold_const_value_deref {:?} {:?} {:?}", val, rty, crty);
330 match (val, &crty.kind(), &rty.kind()) {
331 // the easy case, deref a reference
332 (ConstValue::Scalar(p), x, y) if x == y => {
335 let alloc = self.tcx.global_alloc(p.alloc_id).unwrap_memory();
336 ConstValue::ByRef { alloc, offset: p.offset }
338 Scalar::Raw { .. } => {
339 let layout = self.tcx.layout_of(self.param_env.and(rty)).unwrap();
341 // Deref of a reference to a ZST is a nop.
342 ConstValue::Scalar(Scalar::zst())
344 // FIXME(oli-obk): this is reachable for `const FOO: &&&u32 = &&&42;`
345 bug!("cannot deref {:#?}, {} -> {}", val, crty, rty);
350 // unsize array to slice if pattern is array but match value or other patterns are slice
351 (ConstValue::Scalar(Scalar::Ptr(p)), ty::Array(t, n), ty::Slice(u)) => {
354 data: self.tcx.global_alloc(p.alloc_id).unwrap_memory(),
355 start: p.offset.bytes().try_into().unwrap(),
356 end: n.eval_usize(self.tcx, ty::ParamEnv::empty()).try_into().unwrap(),
359 // fat pointers stay the same
360 (ConstValue::Slice { .. }, _, _)
361 | (_, ty::Slice(_), ty::Slice(_))
362 | (_, ty::Str, ty::Str) => val,
363 // FIXME(oli-obk): this is reachable for `const FOO: &&&u32 = &&&42;` being used
364 _ => bug!("cannot deref {:#?}, {} -> {}", val, crty, rty),
369 impl<'tcx> PatternFolder<'tcx> for LiteralExpander<'tcx> {
370 fn fold_pattern(&mut self, pat: &Pat<'tcx>) -> Pat<'tcx> {
371 debug!("fold_pattern {:?} {:?} {:?}", pat, pat.ty.kind(), pat.kind);
372 match (pat.ty.kind(), &*pat.kind) {
373 (&ty::Ref(_, rty, _), &PatKind::Constant { value: Const { val, ty: const_ty } })
374 if const_ty.is_ref() =>
377 if let ty::Ref(_, crty, _) = const_ty.kind() { crty } else { unreachable!() };
378 if let ty::ConstKind::Value(val) = val {
382 kind: box PatKind::Deref {
386 kind: box PatKind::Constant {
387 value: Const::from_value(
389 self.fold_const_value_deref(*val, rty, crty),
397 bug!("cannot deref {:#?}, {} -> {}", val, crty, rty)
401 (_, &PatKind::Binding { subpattern: Some(ref s), .. }) => s.fold_with(self),
402 (_, &PatKind::AscribeUserType { subpattern: ref s, .. }) => s.fold_with(self),
403 _ => pat.super_fold_with(self),
408 impl<'tcx> Pat<'tcx> {
409 pub(super) fn is_wildcard(&self) -> bool {
411 PatKind::Binding { subpattern: None, .. } | PatKind::Wild => true,
417 /// A row of a matrix. Rows of len 1 are very common, which is why `SmallVec[_; 2]`
419 #[derive(Debug, Clone)]
420 crate struct PatStack<'p, 'tcx>(SmallVec<[&'p Pat<'tcx>; 2]>);
422 impl<'p, 'tcx> PatStack<'p, 'tcx> {
423 crate fn from_pattern(pat: &'p Pat<'tcx>) -> Self {
424 PatStack(smallvec![pat])
427 fn from_vec(vec: SmallVec<[&'p Pat<'tcx>; 2]>) -> Self {
431 fn from_slice(s: &[&'p Pat<'tcx>]) -> Self {
432 PatStack(SmallVec::from_slice(s))
435 fn is_empty(&self) -> bool {
439 fn len(&self) -> usize {
443 fn head(&self) -> &'p Pat<'tcx> {
447 fn to_tail(&self) -> Self {
448 PatStack::from_slice(&self.0[1..])
451 fn iter(&self) -> impl Iterator<Item = &Pat<'tcx>> {
452 self.0.iter().copied()
455 // If the first pattern is an or-pattern, expand this pattern. Otherwise, return `None`.
456 fn expand_or_pat(&self) -> Option<Vec<Self>> {
459 } else if let PatKind::Or { pats } = &*self.head().kind {
463 let mut new_patstack = PatStack::from_pattern(pat);
464 new_patstack.0.extend_from_slice(&self.0[1..]);
474 /// This computes `D(self)`. See top of the file for explanations.
475 fn specialize_wildcard(&self) -> Option<Self> {
476 if self.head().is_wildcard() { Some(self.to_tail()) } else { None }
479 /// This computes `S(constructor, self)`. See top of the file for explanations.
480 fn specialize_constructor(
482 cx: &mut MatchCheckCtxt<'p, 'tcx>,
483 constructor: &Constructor<'tcx>,
484 ctor_wild_subpatterns: &Fields<'p, 'tcx>,
485 ) -> Option<PatStack<'p, 'tcx>> {
487 specialize_one_pattern(cx, self.head(), constructor, ctor_wild_subpatterns)?;
488 Some(new_fields.push_on_patstack(&self.0[1..]))
492 impl<'p, 'tcx> Default for PatStack<'p, 'tcx> {
493 fn default() -> Self {
494 PatStack(smallvec![])
498 impl<'p, 'tcx> FromIterator<&'p Pat<'tcx>> for PatStack<'p, 'tcx> {
499 fn from_iter<T>(iter: T) -> Self
501 T: IntoIterator<Item = &'p Pat<'tcx>>,
503 PatStack(iter.into_iter().collect())
507 /// Depending on the match patterns, the specialization process might be able to use a fast path.
508 /// Tracks whether we can use the fast path and the lookup table needed in those cases.
509 #[derive(Clone, Debug)]
510 enum SpecializationCache {
511 /// Patterns consist of only enum variants.
512 /// Variant patterns does not intersect with each other (in contrast to range patterns),
513 /// so it is possible to precompute the result of `Matrix::specialize_constructor` at a
514 /// lower computational complexity.
515 /// `lookup` is responsible for holding the precomputed result of
516 /// `Matrix::specialize_constructor`, while `wilds` is used for two purposes: the first one is
517 /// the precomputed result of `Matrix::specialize_wildcard`, and the second is to be used as a
518 /// fallback for `Matrix::specialize_constructor` when it tries to apply a constructor that
519 /// has not been seen in the `Matrix`. See `update_cache` for further explanations.
520 Variants { lookup: FxHashMap<DefId, SmallVec<[usize; 1]>>, wilds: SmallVec<[usize; 1]> },
521 /// Does not belong to the cases above, use the slow path.
527 crate struct Matrix<'p, 'tcx> {
528 patterns: Vec<PatStack<'p, 'tcx>>,
529 cache: SpecializationCache,
532 impl<'p, 'tcx> Matrix<'p, 'tcx> {
533 crate fn empty() -> Self {
534 // Use `SpecializationCache::Incompatible` as a placeholder; we will initialize it on the
535 // first call to `push`. See the first half of `update_cache`.
536 Matrix { patterns: vec![], cache: SpecializationCache::Incompatible }
539 /// Pushes a new row to the matrix. If the row starts with an or-pattern, this expands it.
540 crate fn push(&mut self, row: PatStack<'p, 'tcx>) {
541 if let Some(rows) = row.expand_or_pat() {
543 // We recursively expand the or-patterns of the new rows.
544 // This is necessary as we might have `0 | (1 | 2)` or e.g., `x @ 0 | x @ (1 | 2)`.
548 self.patterns.push(row);
549 self.update_cache(self.patterns.len() - 1);
553 fn update_cache(&mut self, idx: usize) {
554 let row = &self.patterns[idx];
555 // We don't know which kind of cache could be used until we see the first row; therefore an
556 // empty `Matrix` is initialized with `SpecializationCache::Empty`, then the cache is
557 // assigned the appropriate variant below on the first call to `push`.
558 if self.patterns.is_empty() {
559 self.cache = if row.is_empty() {
560 SpecializationCache::Incompatible
562 match *row.head().kind {
563 PatKind::Variant { .. } => SpecializationCache::Variants {
564 lookup: FxHashMap::default(),
565 wilds: SmallVec::new(),
567 // Note: If the first pattern is a wildcard, then all patterns after that is not
568 // useful. The check is simple enough so we treat it as the same as unsupported
570 _ => SpecializationCache::Incompatible,
575 match &mut self.cache {
576 SpecializationCache::Variants { ref mut lookup, ref mut wilds } => {
577 let head = row.head();
579 _ if head.is_wildcard() => {
580 // Per rule 1.3 in the top-level comments, a wildcard pattern is included in
581 // the result of `specialize_constructor` for *any* `Constructor`.
582 // We push the wildcard pattern to the precomputed result for constructors
583 // that we have seen before; results for constructors we have not yet seen
584 // defaults to `wilds`, which is updated right below.
585 for (_, v) in lookup.iter_mut() {
588 // Per rule 2.1 and 2.2 in the top-level comments, only wildcard patterns
589 // are included in the result of `specialize_wildcard`.
590 // What we do here is to track the wildcards we have seen; so in addition to
591 // acting as the precomputed result of `specialize_wildcard`, `wilds` also
592 // serves as the default value of `specialize_constructor` for constructors
593 // that are not in `lookup`.
596 PatKind::Variant { adt_def, variant_index, .. } => {
597 // Handle the cases of rule 1.1 and 1.2 in the top-level comments.
598 // A variant pattern can only be included in the results of
599 // `specialize_constructor` for a particular constructor, therefore we are
600 // using a HashMap to track that.
602 .entry(adt_def.variants[variant_index].def_id)
603 // Default to `wilds` for absent keys. See above for an explanation.
604 .or_insert_with(|| wilds.clone())
608 self.cache = SpecializationCache::Incompatible;
612 SpecializationCache::Incompatible => {}
616 /// Iterate over the first component of each row
617 fn heads<'a>(&'a self) -> impl Iterator<Item = &'a Pat<'tcx>> + Captures<'p> {
618 self.patterns.iter().map(|r| r.head())
621 /// This computes `D(self)`. See top of the file for explanations.
622 fn specialize_wildcard(&self) -> Self {
624 SpecializationCache::Variants { wilds, .. } => {
625 wilds.iter().filter_map(|&i| self.patterns[i].specialize_wildcard()).collect()
627 SpecializationCache::Incompatible => {
628 self.patterns.iter().filter_map(|r| r.specialize_wildcard()).collect()
633 /// This computes `S(constructor, self)`. See top of the file for explanations.
634 fn specialize_constructor(
636 cx: &mut MatchCheckCtxt<'p, 'tcx>,
637 constructor: &Constructor<'tcx>,
638 ctor_wild_subpatterns: &Fields<'p, 'tcx>,
639 ) -> Matrix<'p, 'tcx> {
641 SpecializationCache::Variants { lookup, wilds } => {
642 if let Constructor::Variant(id) = constructor {
645 // Default to `wilds` for absent keys. See `update_cache` for an explanation.
649 self.patterns[i].specialize_constructor(
652 ctor_wild_subpatterns,
660 SpecializationCache::Incompatible => self
663 .filter_map(|r| r.specialize_constructor(cx, constructor, ctor_wild_subpatterns))
669 /// Pretty-printer for matrices of patterns, example:
672 /// +++++++++++++++++++++++++++++
674 /// +++++++++++++++++++++++++++++
675 /// + true + [First] +
676 /// +++++++++++++++++++++++++++++
677 /// + true + [Second(true)] +
678 /// +++++++++++++++++++++++++++++
680 /// +++++++++++++++++++++++++++++
681 /// + _ + [_, _, tail @ ..] +
682 /// +++++++++++++++++++++++++++++
683 impl<'p, 'tcx> fmt::Debug for Matrix<'p, 'tcx> {
684 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
687 let Matrix { patterns: m, .. } = self;
688 let pretty_printed_matrix: Vec<Vec<String>> =
689 m.iter().map(|row| row.iter().map(|pat| format!("{:?}", pat)).collect()).collect();
691 let column_count = m.iter().map(|row| row.len()).max().unwrap_or(0);
692 assert!(m.iter().all(|row| row.len() == column_count));
693 let column_widths: Vec<usize> = (0..column_count)
694 .map(|col| pretty_printed_matrix.iter().map(|row| row[col].len()).max().unwrap_or(0))
697 let total_width = column_widths.iter().cloned().sum::<usize>() + column_count * 3 + 1;
698 let br = "+".repeat(total_width);
699 write!(f, "{}\n", br)?;
700 for row in pretty_printed_matrix {
702 for (column, pat_str) in row.into_iter().enumerate() {
704 write!(f, "{:1$}", pat_str, column_widths[column])?;
708 write!(f, "{}\n", br)?;
714 impl<'p, 'tcx> FromIterator<PatStack<'p, 'tcx>> for Matrix<'p, 'tcx> {
715 fn from_iter<T>(iter: T) -> Self
717 T: IntoIterator<Item = PatStack<'p, 'tcx>>,
719 let mut matrix = Matrix::empty();
721 // Using `push` ensures we correctly expand or-patterns.
728 crate struct MatchCheckCtxt<'a, 'tcx> {
729 crate tcx: TyCtxt<'tcx>,
730 /// The module in which the match occurs. This is necessary for
731 /// checking inhabited-ness of types because whether a type is (visibly)
732 /// inhabited can depend on whether it was defined in the current module or
733 /// not. E.g., `struct Foo { _private: ! }` cannot be seen to be empty
734 /// outside it's module and should not be matchable with an empty match
737 crate param_env: ty::ParamEnv<'tcx>,
738 crate pattern_arena: &'a TypedArena<Pat<'tcx>>,
741 impl<'a, 'tcx> MatchCheckCtxt<'a, 'tcx> {
742 fn is_uninhabited(&self, ty: Ty<'tcx>) -> bool {
743 if self.tcx.features().exhaustive_patterns {
744 self.tcx.is_ty_uninhabited_from(self.module, ty, self.param_env)
750 /// Returns whether the given type is an enum from another crate declared `#[non_exhaustive]`.
751 crate fn is_foreign_non_exhaustive_enum(&self, ty: Ty<'tcx>) -> bool {
753 ty::Adt(def, ..) => {
754 def.is_enum() && def.is_variant_list_non_exhaustive() && !def.did.is_local()
761 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
763 /// Patterns of length `n` (`[x, y]`).
765 /// Patterns using the `..` notation (`[x, .., y]`).
766 /// Captures any array constructor of `length >= i + j`.
767 /// In the case where `array_len` is `Some(_)`,
768 /// this indicates that we only care about the first `i` and the last `j` values of the array,
769 /// and everything in between is a wildcard `_`.
774 fn arity(self) -> u64 {
776 FixedLen(length) => length,
777 VarLen(prefix, suffix) => prefix + suffix,
781 /// Whether this pattern includes patterns of length `other_len`.
782 fn covers_length(self, other_len: u64) -> bool {
784 FixedLen(len) => len == other_len,
785 VarLen(prefix, suffix) => prefix + suffix <= other_len,
789 /// Returns a collection of slices that spans the values covered by `self`, subtracted by the
790 /// values covered by `other`: i.e., `self \ other` (in set notation).
791 fn subtract(self, other: Self) -> SmallVec<[Self; 1]> {
792 // Remember, `VarLen(i, j)` covers the union of `FixedLen` from `i + j` to infinity.
793 // Naming: we remove the "neg" constructors from the "pos" ones.
795 FixedLen(pos_len) => {
796 if other.covers_length(pos_len) {
802 VarLen(pos_prefix, pos_suffix) => {
803 let pos_len = pos_prefix + pos_suffix;
805 FixedLen(neg_len) => {
806 if neg_len < pos_len {
811 // We know that `neg_len + 1 >= pos_len >= pos_suffix`.
812 .chain(Some(VarLen(neg_len + 1 - pos_suffix, pos_suffix)))
816 VarLen(neg_prefix, neg_suffix) => {
817 let neg_len = neg_prefix + neg_suffix;
818 if neg_len <= pos_len {
821 (pos_len..neg_len).map(FixedLen).collect()
830 /// A constructor for array and slice patterns.
831 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
833 /// `None` if the matched value is a slice, `Some(n)` if it is an array of size `n`.
834 array_len: Option<u64>,
835 /// The kind of pattern it is: fixed-length `[x, y]` or variable length `[x, .., y]`.
840 /// Returns what patterns this constructor covers: either fixed-length patterns or
841 /// variable-length patterns.
842 fn pattern_kind(self) -> SliceKind {
844 Slice { array_len: Some(len), kind: VarLen(prefix, suffix) }
845 if prefix + suffix == len =>
853 /// Returns what values this constructor covers: either values of only one given length, or
854 /// values of length above a given length.
855 /// This is different from `pattern_kind()` because in some cases the pattern only takes into
856 /// account a subset of the entries of the array, but still only captures values of a given
858 fn value_kind(self) -> SliceKind {
860 Slice { array_len: Some(len), kind: VarLen(_, _) } => FixedLen(len),
865 fn arity(self) -> u64 {
866 self.pattern_kind().arity()
870 /// A value can be decomposed into a constructor applied to some fields. This struct represents
871 /// the constructor. See also `Fields`.
873 /// `pat_constructor` retrieves the constructor corresponding to a pattern.
874 /// `specialize_one_pattern` returns the list of fields corresponding to a pattern, given a
875 /// constructor. `Constructor::apply` reconstructs the pattern from a pair of `Constructor` and
877 #[derive(Clone, Debug, PartialEq)]
878 enum Constructor<'tcx> {
879 /// The constructor for patterns that have a single constructor, like tuples, struct patterns
880 /// and fixed-length arrays.
885 ConstantValue(&'tcx ty::Const<'tcx>),
886 /// Ranges of integer literal values (`2`, `2..=5` or `2..5`).
887 IntRange(IntRange<'tcx>),
888 /// Ranges of floating-point literal values (`2.0..=5.2`).
889 FloatRange(&'tcx ty::Const<'tcx>, &'tcx ty::Const<'tcx>, RangeEnd),
890 /// Array and slice patterns.
892 /// Fake extra constructor for enums that aren't allowed to be matched exhaustively.
896 impl<'tcx> Constructor<'tcx> {
897 fn is_slice(&self) -> bool {
904 fn variant_index_for_adt<'a>(
906 cx: &MatchCheckCtxt<'a, 'tcx>,
907 adt: &'tcx ty::AdtDef,
910 Variant(id) => adt.variant_index_with_id(id),
912 assert!(!adt.is_enum());
915 ConstantValue(c) => cx
917 .destructure_const(cx.param_env.and(c))
919 .expect("destructed const of adt without variant id"),
920 _ => bug!("bad constructor {:?} for adt {:?}", self, adt),
924 // Returns the set of constructors covered by `self` but not by
925 // anything in `other_ctors`.
926 fn subtract_ctors(&self, other_ctors: &Vec<Constructor<'tcx>>) -> Vec<Constructor<'tcx>> {
927 if other_ctors.is_empty() {
928 return vec![self.clone()];
932 // Those constructors can only match themselves.
933 Single | Variant(_) | ConstantValue(..) | FloatRange(..) => {
934 if other_ctors.iter().any(|c| c == self) { vec![] } else { vec![self.clone()] }
937 let mut other_slices = other_ctors
939 .filter_map(|c: &Constructor<'_>| match c {
940 Slice(slice) => Some(*slice),
941 // FIXME(oli-obk): implement `deref` for `ConstValue`
942 ConstantValue(..) => None,
943 _ => bug!("bad slice pattern constructor {:?}", c),
945 .map(Slice::value_kind);
947 match slice.value_kind() {
948 FixedLen(self_len) => {
949 if other_slices.any(|other_slice| other_slice.covers_length(self_len)) {
955 kind @ VarLen(..) => {
956 let mut remaining_slices = vec![kind];
958 // For each used slice, subtract from the current set of slices.
959 for other_slice in other_slices {
960 remaining_slices = remaining_slices
962 .flat_map(|remaining_slice| remaining_slice.subtract(other_slice))
965 // If the constructors that have been considered so far already cover
966 // the entire range of `self`, no need to look at more constructors.
967 if remaining_slices.is_empty() {
974 .map(|kind| Slice { array_len: slice.array_len, kind })
980 IntRange(self_range) => {
981 let mut remaining_ranges = vec![self_range.clone()];
982 for other_ctor in other_ctors {
983 if let IntRange(other_range) = other_ctor {
984 if other_range == self_range {
985 // If the `self` range appears directly in a `match` arm, we can
986 // eliminate it straight away.
987 remaining_ranges = vec![];
989 // Otherwise explicitly compute the remaining ranges.
990 remaining_ranges = other_range.subtract_from(remaining_ranges);
993 // If the ranges that have been considered so far already cover the entire
994 // range of values, we can return early.
995 if remaining_ranges.is_empty() {
1001 // Convert the ranges back into constructors.
1002 remaining_ranges.into_iter().map(IntRange).collect()
1004 // This constructor is never covered by anything else
1005 NonExhaustive => vec![NonExhaustive],
1009 /// Apply a constructor to a list of patterns, yielding a new pattern. `pats`
1010 /// must have as many elements as this constructor's arity.
1012 /// This is roughly the inverse of `specialize_one_pattern`.
1015 /// `self`: `Constructor::Single`
1016 /// `ty`: `(u32, u32, u32)`
1017 /// `pats`: `[10, 20, _]`
1018 /// returns `(10, 20, _)`
1020 /// `self`: `Constructor::Variant(Option::Some)`
1021 /// `ty`: `Option<bool>`
1022 /// `pats`: `[false]`
1023 /// returns `Some(false)`
1026 cx: &MatchCheckCtxt<'p, 'tcx>,
1028 fields: Fields<'p, 'tcx>,
1030 let mut subpatterns = fields.all_patterns();
1032 let pat = match self {
1033 Single | Variant(_) => match ty.kind() {
1034 ty::Adt(..) | ty::Tuple(..) => {
1035 let subpatterns = subpatterns
1037 .map(|(i, p)| FieldPat { field: Field::new(i), pattern: p })
1040 if let ty::Adt(adt, substs) = ty.kind() {
1045 variant_index: self.variant_index_for_adt(cx, adt),
1049 PatKind::Leaf { subpatterns }
1052 PatKind::Leaf { subpatterns }
1055 ty::Ref(..) => PatKind::Deref { subpattern: subpatterns.next().unwrap() },
1056 ty::Slice(_) | ty::Array(..) => bug!("bad slice pattern {:?} {:?}", self, ty),
1059 Slice(slice) => match slice.pattern_kind() {
1061 PatKind::Slice { prefix: subpatterns.collect(), slice: None, suffix: vec![] }
1063 VarLen(prefix, _) => {
1064 let mut prefix: Vec<_> = subpatterns.by_ref().take(prefix as usize).collect();
1065 if slice.array_len.is_some() {
1066 // Improves diagnostics a bit: if the type is a known-size array, instead
1067 // of reporting `[x, _, .., _, y]`, we prefer to report `[x, .., y]`.
1068 // This is incorrect if the size is not known, since `[_, ..]` captures
1069 // arrays of lengths `>= 1` whereas `[..]` captures any length.
1070 while !prefix.is_empty() && prefix.last().unwrap().is_wildcard() {
1074 let suffix: Vec<_> = if slice.array_len.is_some() {
1076 subpatterns.skip_while(Pat::is_wildcard).collect()
1078 subpatterns.collect()
1080 let wild = Pat::wildcard_from_ty(ty);
1081 PatKind::Slice { prefix, slice: Some(wild), suffix }
1084 &ConstantValue(value) => PatKind::Constant { value },
1085 &FloatRange(lo, hi, end) => PatKind::Range(PatRange { lo, hi, end }),
1086 IntRange(range) => return range.to_pat(cx.tcx),
1087 NonExhaustive => PatKind::Wild,
1090 Pat { ty, span: DUMMY_SP, kind: Box::new(pat) }
1093 /// Like `apply`, but where all the subpatterns are wildcards `_`.
1094 fn apply_wildcards<'a>(&self, cx: &MatchCheckCtxt<'a, 'tcx>, ty: Ty<'tcx>) -> Pat<'tcx> {
1095 self.apply(cx, ty, Fields::wildcards(cx, self, ty))
1099 /// Some fields need to be explicitly hidden away in certain cases; see the comment above the
1100 /// `Fields` struct. This struct represents such a potentially-hidden field. When a field is hidden
1101 /// we still keep its type around.
1102 #[derive(Debug, Copy, Clone)]
1103 enum FilteredField<'p, 'tcx> {
1104 Kept(&'p Pat<'tcx>),
1108 impl<'p, 'tcx> FilteredField<'p, 'tcx> {
1109 fn kept(self) -> Option<&'p Pat<'tcx>> {
1111 FilteredField::Kept(p) => Some(p),
1112 FilteredField::Hidden(_) => None,
1116 fn to_pattern(self) -> Pat<'tcx> {
1118 FilteredField::Kept(p) => p.clone(),
1119 FilteredField::Hidden(ty) => Pat::wildcard_from_ty(ty),
1124 /// A value can be decomposed into a constructor applied to some fields. This struct represents
1125 /// those fields, generalized to allow patterns in each field. See also `Constructor`.
1127 /// If a private or `non_exhaustive` field is uninhabited, the code mustn't observe that it is
1128 /// uninhabited. For that, we filter these fields out of the matrix. This is subtle because we
1129 /// still need to have those fields back when going to/from a `Pat`. Most of this is handled
1130 /// automatically in `Fields`, but when constructing or deconstructing `Fields` you need to be
1131 /// careful. As a rule, when going to/from the matrix, use the filtered field list; when going
1132 /// to/from `Pat`, use the full field list.
1133 /// This filtering is uncommon in practice, because uninhabited fields are rarely used, so we avoid
1134 /// it when possible to preserve performance.
1135 #[derive(Debug, Clone)]
1136 enum Fields<'p, 'tcx> {
1137 /// Lists of patterns that don't contain any filtered fields.
1138 /// `Slice` and `Vec` behave the same; the difference is only to avoid allocating and
1139 /// triple-dereferences when possible. Frankly this is premature optimization, I (Nadrieril)
1140 /// have not measured if it really made a difference.
1141 Slice(&'p [Pat<'tcx>]),
1142 Vec(SmallVec<[&'p Pat<'tcx>; 2]>),
1143 /// Patterns where some of the fields need to be hidden. `kept_count` caches the number of
1144 /// non-hidden fields.
1146 fields: SmallVec<[FilteredField<'p, 'tcx>; 2]>,
1151 impl<'p, 'tcx> Fields<'p, 'tcx> {
1152 fn empty() -> Self {
1156 /// Construct a new `Fields` from the given pattern. Must not be used if the pattern is a field
1157 /// of a struct/tuple/variant.
1158 fn from_single_pattern(pat: &'p Pat<'tcx>) -> Self {
1159 Fields::Slice(std::slice::from_ref(pat))
1162 /// Construct a new `Fields` from the given patterns. You must be sure those patterns can't
1163 /// contain fields that need to be filtered out. When in doubt, prefer `replace_fields`.
1164 fn from_slice_unfiltered(pats: &'p [Pat<'tcx>]) -> Self {
1168 /// Convenience; internal use.
1169 fn wildcards_from_tys(
1170 cx: &MatchCheckCtxt<'p, 'tcx>,
1171 tys: impl IntoIterator<Item = Ty<'tcx>>,
1173 let wilds = tys.into_iter().map(Pat::wildcard_from_ty);
1174 let pats = cx.pattern_arena.alloc_from_iter(wilds);
1178 /// Creates a new list of wildcard fields for a given constructor.
1180 cx: &MatchCheckCtxt<'p, 'tcx>,
1181 constructor: &Constructor<'tcx>,
1184 let wildcard_from_ty = |ty| &*cx.pattern_arena.alloc(Pat::wildcard_from_ty(ty));
1186 let ret = match constructor {
1187 Single | Variant(_) => match ty.kind() {
1188 ty::Tuple(ref fs) => {
1189 Fields::wildcards_from_tys(cx, fs.into_iter().map(|ty| ty.expect_ty()))
1191 ty::Ref(_, rty, _) => Fields::from_single_pattern(wildcard_from_ty(rty)),
1192 ty::Adt(adt, substs) => {
1194 // Use T as the sub pattern type of Box<T>.
1195 Fields::from_single_pattern(wildcard_from_ty(substs.type_at(0)))
1197 let variant = &adt.variants[constructor.variant_index_for_adt(cx, adt)];
1198 // Whether we must not match the fields of this variant exhaustively.
1199 let is_non_exhaustive =
1200 variant.is_field_list_non_exhaustive() && !adt.did.is_local();
1201 let field_tys = variant.fields.iter().map(|field| field.ty(cx.tcx, substs));
1202 // In the following cases, we don't need to filter out any fields. This is
1203 // the vast majority of real cases, since uninhabited fields are uncommon.
1204 let has_no_hidden_fields = (adt.is_enum() && !is_non_exhaustive)
1205 || !field_tys.clone().any(|ty| cx.is_uninhabited(ty));
1207 if has_no_hidden_fields {
1208 Fields::wildcards_from_tys(cx, field_tys)
1210 let mut kept_count = 0;
1211 let fields = variant
1215 let ty = field.ty(cx.tcx, substs);
1216 let is_visible = adt.is_enum()
1217 || field.vis.is_accessible_from(cx.module, cx.tcx);
1218 let is_uninhabited = cx.is_uninhabited(ty);
1220 // In the cases of either a `#[non_exhaustive]` field list
1221 // or a non-public field, we hide uninhabited fields in
1222 // order not to reveal the uninhabitedness of the whole
1224 if is_uninhabited && (!is_visible || is_non_exhaustive) {
1225 FilteredField::Hidden(ty)
1228 FilteredField::Kept(wildcard_from_ty(ty))
1232 Fields::Filtered { fields, kept_count }
1236 _ => Fields::empty(),
1238 Slice(slice) => match *ty.kind() {
1239 ty::Slice(ty) | ty::Array(ty, _) => {
1240 let arity = slice.arity();
1241 Fields::wildcards_from_tys(cx, (0..arity).map(|_| ty))
1243 _ => bug!("bad slice pattern {:?} {:?}", constructor, ty),
1245 ConstantValue(..) | FloatRange(..) | IntRange(..) | NonExhaustive => Fields::empty(),
1247 debug!("Fields::wildcards({:?}, {:?}) = {:#?}", constructor, ty, ret);
1251 /// Returns the number of patterns from the viewpoint of match-checking, i.e. excluding hidden
1252 /// fields. This is what we want in most cases in this file, the only exception being
1253 /// conversion to/from `Pat`.
1254 fn len(&self) -> usize {
1256 Fields::Slice(pats) => pats.len(),
1257 Fields::Vec(pats) => pats.len(),
1258 Fields::Filtered { kept_count, .. } => *kept_count,
1262 /// Returns the complete list of patterns, including hidden fields.
1263 fn all_patterns(self) -> impl Iterator<Item = Pat<'tcx>> {
1264 let pats: SmallVec<[_; 2]> = match self {
1265 Fields::Slice(pats) => pats.iter().cloned().collect(),
1266 Fields::Vec(pats) => pats.into_iter().cloned().collect(),
1267 Fields::Filtered { fields, .. } => {
1268 // We don't skip any fields here.
1269 fields.into_iter().map(|p| p.to_pattern()).collect()
1275 /// Overrides some of the fields with the provided patterns. Exactly like
1276 /// `replace_fields_indexed`, except that it takes `FieldPat`s as input.
1277 fn replace_with_fieldpats(
1279 new_pats: impl IntoIterator<Item = &'p FieldPat<'tcx>>,
1281 self.replace_fields_indexed(
1282 new_pats.into_iter().map(|pat| (pat.field.index(), &pat.pattern)),
1286 /// Overrides some of the fields with the provided patterns. This is used when a pattern
1287 /// defines some fields but not all, for example `Foo { field1: Some(_), .. }`: here we start with a
1288 /// `Fields` that is just one wildcard per field of the `Foo` struct, and override the entry
1289 /// corresponding to `field1` with the pattern `Some(_)`. This is also used for slice patterns
1290 /// for the same reason.
1291 fn replace_fields_indexed(
1293 new_pats: impl IntoIterator<Item = (usize, &'p Pat<'tcx>)>,
1295 let mut fields = self.clone();
1296 if let Fields::Slice(pats) = fields {
1297 fields = Fields::Vec(pats.iter().collect());
1301 Fields::Vec(pats) => {
1302 for (i, pat) in new_pats {
1306 Fields::Filtered { fields, .. } => {
1307 for (i, pat) in new_pats {
1308 if let FilteredField::Kept(p) = &mut fields[i] {
1313 Fields::Slice(_) => unreachable!(),
1318 /// Replaces contained fields with the given filtered list of patterns, e.g. taken from the
1319 /// matrix. There must be `len()` patterns in `pats`.
1322 cx: &MatchCheckCtxt<'p, 'tcx>,
1323 pats: impl IntoIterator<Item = Pat<'tcx>>,
1325 let pats: &[_] = cx.pattern_arena.alloc_from_iter(pats);
1328 Fields::Filtered { fields, kept_count } => {
1329 let mut pats = pats.iter();
1330 let mut fields = fields.clone();
1331 for f in &mut fields {
1332 if let FilteredField::Kept(p) = f {
1333 // We take one input pattern for each `Kept` field, in order.
1334 *p = pats.next().unwrap();
1337 Fields::Filtered { fields, kept_count: *kept_count }
1339 _ => Fields::Slice(pats),
1343 fn push_on_patstack(self, stack: &[&'p Pat<'tcx>]) -> PatStack<'p, 'tcx> {
1344 let pats: SmallVec<_> = match self {
1345 Fields::Slice(pats) => pats.iter().chain(stack.iter().copied()).collect(),
1346 Fields::Vec(mut pats) => {
1347 pats.extend_from_slice(stack);
1350 Fields::Filtered { fields, .. } => {
1351 // We skip hidden fields here
1352 fields.into_iter().filter_map(|p| p.kept()).chain(stack.iter().copied()).collect()
1355 PatStack::from_vec(pats)
1359 #[derive(Clone, Debug)]
1360 crate enum Usefulness<'tcx> {
1361 /// Carries a list of unreachable subpatterns. Used only in the presence of or-patterns.
1363 /// Carries a list of witnesses of non-exhaustiveness.
1364 UsefulWithWitness(Vec<Witness<'tcx>>),
1368 impl<'tcx> Usefulness<'tcx> {
1369 fn new_useful(preference: WitnessPreference) -> Self {
1371 ConstructWitness => UsefulWithWitness(vec![Witness(vec![])]),
1372 LeaveOutWitness => Useful(vec![]),
1376 fn is_useful(&self) -> bool {
1383 fn apply_constructor<'p>(
1385 cx: &MatchCheckCtxt<'p, 'tcx>,
1386 ctor: &Constructor<'tcx>,
1388 ctor_wild_subpatterns: &Fields<'p, 'tcx>,
1391 UsefulWithWitness(witnesses) => UsefulWithWitness(
1394 .map(|witness| witness.apply_constructor(cx, &ctor, ty, ctor_wild_subpatterns))
1401 fn apply_wildcard(self, ty: Ty<'tcx>) -> Self {
1403 UsefulWithWitness(witnesses) => {
1404 let wild = Pat::wildcard_from_ty(ty);
1408 .map(|mut witness| {
1409 witness.0.push(wild.clone());
1419 fn apply_missing_ctors(
1421 cx: &MatchCheckCtxt<'_, 'tcx>,
1423 missing_ctors: &MissingConstructors<'tcx>,
1426 UsefulWithWitness(witnesses) => {
1427 let new_patterns: Vec<_> =
1428 missing_ctors.iter().map(|ctor| ctor.apply_wildcards(cx, ty)).collect();
1429 // Add the new patterns to each witness
1433 .flat_map(|witness| {
1434 new_patterns.iter().map(move |pat| {
1435 let mut witness = witness.clone();
1436 witness.0.push(pat.clone());
1448 #[derive(Copy, Clone, Debug)]
1449 crate enum WitnessPreference {
1454 #[derive(Copy, Clone, Debug)]
1455 struct PatCtxt<'tcx> {
1460 /// A witness of non-exhaustiveness for error reporting, represented
1461 /// as a list of patterns (in reverse order of construction) with
1462 /// wildcards inside to represent elements that can take any inhabitant
1463 /// of the type as a value.
1465 /// A witness against a list of patterns should have the same types
1466 /// and length as the pattern matched against. Because Rust `match`
1467 /// is always against a single pattern, at the end the witness will
1468 /// have length 1, but in the middle of the algorithm, it can contain
1469 /// multiple patterns.
1471 /// For example, if we are constructing a witness for the match against
1473 /// struct Pair(Option<(u32, u32)>, bool);
1475 /// match (p: Pair) {
1476 /// Pair(None, _) => {}
1477 /// Pair(_, false) => {}
1481 /// We'll perform the following steps:
1482 /// 1. Start with an empty witness
1483 /// `Witness(vec![])`
1484 /// 2. Push a witness `Some(_)` against the `None`
1485 /// `Witness(vec![Some(_)])`
1486 /// 3. Push a witness `true` against the `false`
1487 /// `Witness(vec![Some(_), true])`
1488 /// 4. Apply the `Pair` constructor to the witnesses
1489 /// `Witness(vec![Pair(Some(_), true)])`
1491 /// The final `Pair(Some(_), true)` is then the resulting witness.
1492 #[derive(Clone, Debug)]
1493 crate struct Witness<'tcx>(Vec<Pat<'tcx>>);
1495 impl<'tcx> Witness<'tcx> {
1496 crate fn single_pattern(self) -> Pat<'tcx> {
1497 assert_eq!(self.0.len(), 1);
1498 self.0.into_iter().next().unwrap()
1501 /// Constructs a partial witness for a pattern given a list of
1502 /// patterns expanded by the specialization step.
1504 /// When a pattern P is discovered to be useful, this function is used bottom-up
1505 /// to reconstruct a complete witness, e.g., a pattern P' that covers a subset
1506 /// of values, V, where each value in that set is not covered by any previously
1507 /// used patterns and is covered by the pattern P'. Examples:
1509 /// left_ty: tuple of 3 elements
1510 /// pats: [10, 20, _] => (10, 20, _)
1512 /// left_ty: struct X { a: (bool, &'static str), b: usize}
1513 /// pats: [(false, "foo"), 42] => X { a: (false, "foo"), b: 42 }
1514 fn apply_constructor<'p>(
1516 cx: &MatchCheckCtxt<'p, 'tcx>,
1517 ctor: &Constructor<'tcx>,
1519 ctor_wild_subpatterns: &Fields<'p, 'tcx>,
1522 let len = self.0.len();
1523 let arity = ctor_wild_subpatterns.len();
1524 let pats = self.0.drain((len - arity)..).rev();
1525 let fields = ctor_wild_subpatterns.replace_fields(cx, pats);
1526 ctor.apply(cx, ty, fields)
1535 /// This determines the set of all possible constructors of a pattern matching
1536 /// values of type `left_ty`. For vectors, this would normally be an infinite set
1537 /// but is instead bounded by the maximum fixed length of slice patterns in
1538 /// the column of patterns being analyzed.
1540 /// We make sure to omit constructors that are statically impossible. E.g., for
1541 /// `Option<!>`, we do not include `Some(_)` in the returned list of constructors.
1542 /// Invariant: this returns an empty `Vec` if and only if the type is uninhabited (as determined by
1543 /// `cx.is_uninhabited()`).
1544 fn all_constructors<'a, 'tcx>(
1545 cx: &mut MatchCheckCtxt<'a, 'tcx>,
1547 ) -> Vec<Constructor<'tcx>> {
1548 debug!("all_constructors({:?})", pcx.ty);
1549 let make_range = |start, end| {
1551 // `unwrap()` is ok because we know the type is an integer.
1552 IntRange::from_range(cx.tcx, start, end, pcx.ty, &RangeEnd::Included, pcx.span)
1556 match *pcx.ty.kind() {
1558 [true, false].iter().map(|&b| ConstantValue(ty::Const::from_bool(cx.tcx, b))).collect()
1560 ty::Array(ref sub_ty, len) if len.try_eval_usize(cx.tcx, cx.param_env).is_some() => {
1561 let len = len.eval_usize(cx.tcx, cx.param_env);
1562 if len != 0 && cx.is_uninhabited(sub_ty) {
1565 vec![Slice(Slice { array_len: Some(len), kind: VarLen(0, 0) })]
1568 // Treat arrays of a constant but unknown length like slices.
1569 ty::Array(ref sub_ty, _) | ty::Slice(ref sub_ty) => {
1570 let kind = if cx.is_uninhabited(sub_ty) { FixedLen(0) } else { VarLen(0, 0) };
1571 vec![Slice(Slice { array_len: None, kind })]
1573 ty::Adt(def, substs) if def.is_enum() => {
1574 let ctors: Vec<_> = if cx.tcx.features().exhaustive_patterns {
1575 // If `exhaustive_patterns` is enabled, we exclude variants known to be
1580 !v.uninhabited_from(cx.tcx, substs, def.adt_kind(), cx.param_env)
1581 .contains(cx.tcx, cx.module)
1583 .map(|v| Variant(v.def_id))
1586 def.variants.iter().map(|v| Variant(v.def_id)).collect()
1589 // If the enum is declared as `#[non_exhaustive]`, we treat it as if it had an
1590 // additional "unknown" constructor.
1591 // There is no point in enumerating all possible variants, because the user can't
1592 // actually match against them all themselves. So we always return only the fictitious
1594 // E.g., in an example like:
1596 // let err: io::ErrorKind = ...;
1598 // io::ErrorKind::NotFound => {},
1601 // we don't want to show every possible IO error, but instead have only `_` as the
1603 let is_declared_nonexhaustive = cx.is_foreign_non_exhaustive_enum(pcx.ty);
1605 // If `exhaustive_patterns` is disabled and our scrutinee is an empty enum, we treat it
1606 // as though it had an "unknown" constructor to avoid exposing its emptyness. Note that
1607 // an empty match will still be considered exhaustive because that case is handled
1608 // separately in `check_match`.
1609 let is_secretly_empty =
1610 def.variants.is_empty() && !cx.tcx.features().exhaustive_patterns;
1612 if is_secretly_empty || is_declared_nonexhaustive { vec![NonExhaustive] } else { ctors }
1616 // The valid Unicode Scalar Value ranges.
1617 make_range('\u{0000}' as u128, '\u{D7FF}' as u128),
1618 make_range('\u{E000}' as u128, '\u{10FFFF}' as u128),
1621 ty::Int(_) | ty::Uint(_)
1622 if pcx.ty.is_ptr_sized_integral()
1623 && !cx.tcx.features().precise_pointer_size_matching =>
1625 // `usize`/`isize` are not allowed to be matched exhaustively unless the
1626 // `precise_pointer_size_matching` feature is enabled. So we treat those types like
1627 // `#[non_exhaustive]` enums by returning a special unmatcheable constructor.
1631 let bits = Integer::from_attr(&cx.tcx, SignedInt(ity)).size().bits() as u128;
1632 let min = 1u128 << (bits - 1);
1634 vec![make_range(min, max)]
1637 let size = Integer::from_attr(&cx.tcx, UnsignedInt(uty)).size();
1638 let max = truncate(u128::MAX, size);
1639 vec![make_range(0, max)]
1642 if cx.is_uninhabited(pcx.ty) {
1651 /// An inclusive interval, used for precise integer exhaustiveness checking.
1652 /// `IntRange`s always store a contiguous range. This means that values are
1653 /// encoded such that `0` encodes the minimum value for the integer,
1654 /// regardless of the signedness.
1655 /// For example, the pattern `-128..=127i8` is encoded as `0..=255`.
1656 /// This makes comparisons and arithmetic on interval endpoints much more
1657 /// straightforward. See `signed_bias` for details.
1659 /// `IntRange` is never used to encode an empty range or a "range" that wraps
1660 /// around the (offset) space: i.e., `range.lo <= range.hi`.
1661 #[derive(Clone, Debug)]
1662 struct IntRange<'tcx> {
1663 range: RangeInclusive<u128>,
1668 impl<'tcx> IntRange<'tcx> {
1670 fn is_integral(ty: Ty<'_>) -> bool {
1672 ty::Char | ty::Int(_) | ty::Uint(_) => true,
1677 fn is_singleton(&self) -> bool {
1678 self.range.start() == self.range.end()
1681 fn boundaries(&self) -> (u128, u128) {
1682 (*self.range.start(), *self.range.end())
1685 /// Don't treat `usize`/`isize` exhaustively unless the `precise_pointer_size_matching` feature
1687 fn treat_exhaustively(&self, tcx: TyCtxt<'tcx>) -> bool {
1688 !self.ty.is_ptr_sized_integral() || tcx.features().precise_pointer_size_matching
1692 fn integral_size_and_signed_bias(tcx: TyCtxt<'tcx>, ty: Ty<'_>) -> Option<(Size, u128)> {
1694 ty::Char => Some((Size::from_bytes(4), 0)),
1696 let size = Integer::from_attr(&tcx, SignedInt(ity)).size();
1697 Some((size, 1u128 << (size.bits() as u128 - 1)))
1699 ty::Uint(uty) => Some((Integer::from_attr(&tcx, UnsignedInt(uty)).size(), 0)),
1707 param_env: ty::ParamEnv<'tcx>,
1708 value: &Const<'tcx>,
1710 ) -> Option<IntRange<'tcx>> {
1711 if let Some((target_size, bias)) = Self::integral_size_and_signed_bias(tcx, value.ty) {
1714 if let ty::ConstKind::Value(ConstValue::Scalar(scalar)) = value.val {
1715 // For this specific pattern we can skip a lot of effort and go
1716 // straight to the result, after doing a bit of checking. (We
1717 // could remove this branch and just fall through, which
1718 // is more general but much slower.)
1719 if let Ok(bits) = scalar.to_bits_or_ptr(target_size, &tcx) {
1723 // This is a more general form of the previous case.
1724 value.try_eval_bits(tcx, param_env, ty)
1726 let val = val ^ bias;
1727 Some(IntRange { range: val..=val, ty, span })
1741 ) -> Option<IntRange<'tcx>> {
1742 if Self::is_integral(ty) {
1743 // Perform a shift if the underlying types are signed,
1744 // which makes the interval arithmetic simpler.
1745 let bias = IntRange::signed_bias(tcx, ty);
1746 let (lo, hi) = (lo ^ bias, hi ^ bias);
1747 let offset = (*end == RangeEnd::Excluded) as u128;
1748 if lo > hi || (lo == hi && *end == RangeEnd::Excluded) {
1749 // This should have been caught earlier by E0030.
1750 bug!("malformed range pattern: {}..={}", lo, (hi - offset));
1752 Some(IntRange { range: lo..=(hi - offset), ty, span })
1760 param_env: ty::ParamEnv<'tcx>,
1762 ) -> Option<IntRange<'tcx>> {
1763 match pat_constructor(tcx, param_env, pat)? {
1764 IntRange(range) => Some(range),
1769 // The return value of `signed_bias` should be XORed with an endpoint to encode/decode it.
1770 fn signed_bias(tcx: TyCtxt<'tcx>, ty: Ty<'tcx>) -> u128 {
1773 let bits = Integer::from_attr(&tcx, SignedInt(ity)).size().bits() as u128;
1780 /// Returns a collection of ranges that spans the values covered by `ranges`, subtracted
1781 /// by the values covered by `self`: i.e., `ranges \ self` (in set notation).
1782 fn subtract_from(&self, ranges: Vec<IntRange<'tcx>>) -> Vec<IntRange<'tcx>> {
1783 let mut remaining_ranges = vec![];
1785 let span = self.span;
1786 let (lo, hi) = self.boundaries();
1787 for subrange in ranges {
1788 let (subrange_lo, subrange_hi) = subrange.range.into_inner();
1789 if lo > subrange_hi || subrange_lo > hi {
1790 // The pattern doesn't intersect with the subrange at all,
1791 // so the subrange remains untouched.
1792 remaining_ranges.push(IntRange { range: subrange_lo..=subrange_hi, ty, span });
1794 if lo > subrange_lo {
1795 // The pattern intersects an upper section of the
1796 // subrange, so a lower section will remain.
1797 remaining_ranges.push(IntRange { range: subrange_lo..=(lo - 1), ty, span });
1799 if hi < subrange_hi {
1800 // The pattern intersects a lower section of the
1801 // subrange, so an upper section will remain.
1802 remaining_ranges.push(IntRange { range: (hi + 1)..=subrange_hi, ty, span });
1809 fn is_subrange(&self, other: &Self) -> bool {
1810 other.range.start() <= self.range.start() && self.range.end() <= other.range.end()
1813 fn intersection(&self, tcx: TyCtxt<'tcx>, other: &Self) -> Option<Self> {
1815 let (lo, hi) = self.boundaries();
1816 let (other_lo, other_hi) = other.boundaries();
1817 if self.treat_exhaustively(tcx) {
1818 if lo <= other_hi && other_lo <= hi {
1819 let span = other.span;
1820 Some(IntRange { range: max(lo, other_lo)..=min(hi, other_hi), ty, span })
1825 // If the range should not be treated exhaustively, fallback to checking for inclusion.
1826 if self.is_subrange(other) { Some(self.clone()) } else { None }
1830 fn suspicious_intersection(&self, other: &Self) -> bool {
1831 // `false` in the following cases:
1832 // 1 ---- // 1 ---------- // 1 ---- // 1 ----
1833 // 2 ---------- // 2 ---- // 2 ---- // 2 ----
1835 // The following are currently `false`, but could be `true` in the future (#64007):
1836 // 1 --------- // 1 ---------
1837 // 2 ---------- // 2 ----------
1839 // `true` in the following cases:
1840 // 1 ------- // 1 -------
1841 // 2 -------- // 2 -------
1842 let (lo, hi) = self.boundaries();
1843 let (other_lo, other_hi) = other.boundaries();
1844 lo == other_hi || hi == other_lo
1847 fn to_pat(&self, tcx: TyCtxt<'tcx>) -> Pat<'tcx> {
1848 let (lo, hi) = self.boundaries();
1850 let bias = IntRange::signed_bias(tcx, self.ty);
1851 let (lo, hi) = (lo ^ bias, hi ^ bias);
1853 let ty = ty::ParamEnv::empty().and(self.ty);
1854 let lo_const = ty::Const::from_bits(tcx, lo, ty);
1855 let hi_const = ty::Const::from_bits(tcx, hi, ty);
1857 let kind = if lo == hi {
1858 PatKind::Constant { value: lo_const }
1860 PatKind::Range(PatRange { lo: lo_const, hi: hi_const, end: RangeEnd::Included })
1863 // This is a brand new pattern, so we don't reuse `self.span`.
1864 Pat { ty: self.ty, span: DUMMY_SP, kind: Box::new(kind) }
1868 /// Ignore spans when comparing, they don't carry semantic information as they are only for lints.
1869 impl<'tcx> std::cmp::PartialEq for IntRange<'tcx> {
1870 fn eq(&self, other: &Self) -> bool {
1871 self.range == other.range && self.ty == other.ty
1875 // A struct to compute a set of constructors equivalent to `all_ctors \ used_ctors`.
1876 struct MissingConstructors<'tcx> {
1877 all_ctors: Vec<Constructor<'tcx>>,
1878 used_ctors: Vec<Constructor<'tcx>>,
1881 impl<'tcx> MissingConstructors<'tcx> {
1882 fn new(all_ctors: Vec<Constructor<'tcx>>, used_ctors: Vec<Constructor<'tcx>>) -> Self {
1883 MissingConstructors { all_ctors, used_ctors }
1886 fn into_inner(self) -> (Vec<Constructor<'tcx>>, Vec<Constructor<'tcx>>) {
1887 (self.all_ctors, self.used_ctors)
1890 fn is_empty(&self) -> bool {
1891 self.iter().next().is_none()
1893 /// Whether this contains all the constructors for the given type or only a
1895 fn all_ctors_are_missing(&self) -> bool {
1896 self.used_ctors.is_empty()
1899 /// Iterate over all_ctors \ used_ctors
1900 fn iter<'a>(&'a self) -> impl Iterator<Item = Constructor<'tcx>> + Captures<'a> {
1901 self.all_ctors.iter().flat_map(move |req_ctor| req_ctor.subtract_ctors(&self.used_ctors))
1905 impl<'tcx> fmt::Debug for MissingConstructors<'tcx> {
1906 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1907 let ctors: Vec<_> = self.iter().collect();
1908 write!(f, "{:?}", ctors)
1912 /// Algorithm from http://moscova.inria.fr/~maranget/papers/warn/index.html.
1913 /// The algorithm from the paper has been modified to correctly handle empty
1914 /// types. The changes are:
1915 /// (0) We don't exit early if the pattern matrix has zero rows. We just
1916 /// continue to recurse over columns.
1917 /// (1) all_constructors will only return constructors that are statically
1918 /// possible. E.g., it will only return `Ok` for `Result<T, !>`.
1920 /// This finds whether a (row) vector `v` of patterns is 'useful' in relation
1921 /// to a set of such vectors `m` - this is defined as there being a set of
1922 /// inputs that will match `v` but not any of the sets in `m`.
1924 /// All the patterns at each column of the `matrix ++ v` matrix must have the same type.
1926 /// This is used both for reachability checking (if a pattern isn't useful in
1927 /// relation to preceding patterns, it is not reachable) and exhaustiveness
1928 /// checking (if a wildcard pattern is useful in relation to a matrix, the
1929 /// matrix isn't exhaustive).
1931 /// `is_under_guard` is used to inform if the pattern has a guard. If it
1932 /// has one it must not be inserted into the matrix. This shouldn't be
1933 /// relied on for soundness.
1934 crate fn is_useful<'p, 'tcx>(
1935 cx: &mut MatchCheckCtxt<'p, 'tcx>,
1936 matrix: &Matrix<'p, 'tcx>,
1937 v: &PatStack<'p, 'tcx>,
1938 witness_preference: WitnessPreference,
1940 is_under_guard: bool,
1942 ) -> Usefulness<'tcx> {
1943 let Matrix { patterns: rows, .. } = matrix;
1944 debug!("is_useful({:#?}, {:#?})", matrix, v);
1946 // The base case. We are pattern-matching on () and the return value is
1947 // based on whether our matrix has a row or not.
1948 // NOTE: This could potentially be optimized by checking rows.is_empty()
1949 // first and then, if v is non-empty, the return value is based on whether
1950 // the type of the tuple we're checking is inhabited or not.
1952 return if rows.is_empty() {
1953 Usefulness::new_useful(witness_preference)
1959 assert!(rows.iter().all(|r| r.len() == v.len()));
1961 // If the first pattern is an or-pattern, expand it.
1962 if let Some(vs) = v.expand_or_pat() {
1963 // We need to push the already-seen patterns into the matrix in order to detect redundant
1964 // branches like `Some(_) | Some(0)`. We also keep track of the unreachable subpatterns.
1965 let mut matrix = matrix.clone();
1966 // `Vec` of all the unreachable branches of the current or-pattern.
1967 let mut unreachable_branches = Vec::new();
1968 // Subpatterns that are unreachable from all branches. E.g. in the following case, the last
1969 // `true` is unreachable only from one branch, so it is overall reachable.
1971 // match (true, true) {
1972 // (true, true) => {}
1973 // (false | true, false | true) => {}
1976 let mut unreachable_subpats = FxHashSet::default();
1977 // Whether any branch at all is useful.
1978 let mut any_is_useful = false;
1981 let res = is_useful(cx, &matrix, &v, witness_preference, hir_id, is_under_guard, false);
1985 any_is_useful = true;
1986 // Initialize with the first set of unreachable subpatterns encountered.
1987 unreachable_subpats = pats.into_iter().collect();
1989 // Keep the patterns unreachable from both this and previous branches.
1990 unreachable_subpats =
1991 pats.into_iter().filter(|p| unreachable_subpats.contains(p)).collect();
1994 NotUseful => unreachable_branches.push(v.head().span),
1995 UsefulWithWitness(_) => {
1996 bug!("Encountered or-pat in `v` during exhaustiveness checking")
1999 // If pattern has a guard don't add it to the matrix
2000 if !is_under_guard {
2005 // Collect all the unreachable patterns.
2006 unreachable_branches.extend(unreachable_subpats);
2007 return Useful(unreachable_branches);
2013 // FIXME(Nadrieril): Hack to work around type normalization issues (see #72476).
2014 let ty = matrix.heads().next().map(|r| r.ty).unwrap_or(v.head().ty);
2015 let pcx = PatCtxt { ty, span: v.head().span };
2017 debug!("is_useful_expand_first_col: pcx={:#?}, expanding {:#?}", pcx, v.head());
2019 let ret = if let Some(constructor) = pat_constructor(cx.tcx, cx.param_env, v.head()) {
2020 debug!("is_useful - expanding constructor: {:#?}", constructor);
2021 split_grouped_constructors(
2032 is_useful_specialized(
2043 .find(|result| result.is_useful())
2044 .unwrap_or(NotUseful)
2046 debug!("is_useful - expanding wildcard");
2048 let used_ctors: Vec<Constructor<'_>> =
2049 matrix.heads().filter_map(|p| pat_constructor(cx.tcx, cx.param_env, p)).collect();
2050 debug!("is_useful_used_ctors = {:#?}", used_ctors);
2051 // `all_ctors` are all the constructors for the given type, which
2052 // should all be represented (or caught with the wild pattern `_`).
2053 let all_ctors = all_constructors(cx, pcx);
2054 debug!("is_useful_all_ctors = {:#?}", all_ctors);
2056 // `missing_ctors` is the set of constructors from the same type as the
2057 // first column of `matrix` that are matched only by wildcard patterns
2058 // from the first column.
2060 // Therefore, if there is some pattern that is unmatched by `matrix`,
2061 // it will still be unmatched if the first constructor is replaced by
2062 // any of the constructors in `missing_ctors`
2064 // Missing constructors are those that are not matched by any non-wildcard patterns in the
2065 // current column. We only fully construct them on-demand, because they're rarely used and
2067 let missing_ctors = MissingConstructors::new(all_ctors, used_ctors);
2069 debug!("is_useful_missing_ctors.empty()={:#?}", missing_ctors.is_empty(),);
2071 if missing_ctors.is_empty() {
2072 let (all_ctors, _) = missing_ctors.into_inner();
2073 split_grouped_constructors(cx.tcx, cx.param_env, pcx, all_ctors, matrix, DUMMY_SP, None)
2076 is_useful_specialized(
2087 .find(|result| result.is_useful())
2088 .unwrap_or(NotUseful)
2090 let matrix = matrix.specialize_wildcard();
2091 let v = v.to_tail();
2093 is_useful(cx, &matrix, &v, witness_preference, hir_id, is_under_guard, false);
2095 // In this case, there's at least one "free"
2096 // constructor that is only matched against by
2097 // wildcard patterns.
2099 // There are 2 ways we can report a witness here.
2100 // Commonly, we can report all the "free"
2101 // constructors as witnesses, e.g., if we have:
2104 // enum Direction { N, S, E, W }
2105 // let Direction::N = ...;
2108 // we can report 3 witnesses: `S`, `E`, and `W`.
2110 // However, there is a case where we don't want
2111 // to do this and instead report a single `_` witness:
2112 // if the user didn't actually specify a constructor
2113 // in this arm, e.g., in
2115 // let x: (Direction, Direction, bool) = ...;
2116 // let (_, _, false) = x;
2118 // we don't want to show all 16 possible witnesses
2119 // `(<direction-1>, <direction-2>, true)` - we are
2120 // satisfied with `(_, _, true)`. In this case,
2121 // `used_ctors` is empty.
2122 // The exception is: if we are at the top-level, for example in an empty match, we
2123 // sometimes prefer reporting the list of constructors instead of just `_`.
2124 let report_ctors_rather_than_wildcard = is_top_level && !IntRange::is_integral(pcx.ty);
2125 if missing_ctors.all_ctors_are_missing() && !report_ctors_rather_than_wildcard {
2126 // All constructors are unused. Add a wild pattern
2127 // rather than each individual constructor.
2128 usefulness.apply_wildcard(pcx.ty)
2130 // Construct for each missing constructor a "wild" version of this
2131 // constructor, that matches everything that can be built with
2132 // it. For example, if `ctor` is a `Constructor::Variant` for
2133 // `Option::Some`, we get the pattern `Some(_)`.
2134 usefulness.apply_missing_ctors(cx, pcx.ty, &missing_ctors)
2138 debug!("is_useful::returns({:#?}, {:#?}) = {:?}", matrix, v, ret);
2142 /// A shorthand for the `U(S(c, P), S(c, q))` operation from the paper. I.e., `is_useful` applied
2143 /// to the specialised version of both the pattern matrix `P` and the new pattern `q`.
2144 fn is_useful_specialized<'p, 'tcx>(
2145 cx: &mut MatchCheckCtxt<'p, 'tcx>,
2146 matrix: &Matrix<'p, 'tcx>,
2147 v: &PatStack<'p, 'tcx>,
2148 ctor: Constructor<'tcx>,
2150 witness_preference: WitnessPreference,
2152 is_under_guard: bool,
2153 ) -> Usefulness<'tcx> {
2154 debug!("is_useful_specialized({:#?}, {:#?}, {:?})", v, ctor, ty);
2156 // We cache the result of `Fields::wildcards` because it is used a lot.
2157 let ctor_wild_subpatterns = Fields::wildcards(cx, &ctor, ty);
2158 let matrix = matrix.specialize_constructor(cx, &ctor, &ctor_wild_subpatterns);
2159 v.specialize_constructor(cx, &ctor, &ctor_wild_subpatterns)
2160 .map(|v| is_useful(cx, &matrix, &v, witness_preference, hir_id, is_under_guard, false))
2161 .map(|u| u.apply_constructor(cx, &ctor, ty, &ctor_wild_subpatterns))
2162 .unwrap_or(NotUseful)
2165 /// Determines the constructor that the given pattern can be specialized to.
2166 /// Returns `None` in case of a catch-all, which can't be specialized.
2167 fn pat_constructor<'tcx>(
2169 param_env: ty::ParamEnv<'tcx>,
2171 ) -> Option<Constructor<'tcx>> {
2173 PatKind::AscribeUserType { .. } => bug!(), // Handled by `expand_pattern`
2174 PatKind::Binding { .. } | PatKind::Wild => None,
2175 PatKind::Leaf { .. } | PatKind::Deref { .. } => Some(Single),
2176 PatKind::Variant { adt_def, variant_index, .. } => {
2177 Some(Variant(adt_def.variants[variant_index].def_id))
2179 PatKind::Constant { value } => {
2180 if let Some(int_range) = IntRange::from_const(tcx, param_env, value, pat.span) {
2181 Some(IntRange(int_range))
2183 match (value.val, &value.ty.kind()) {
2184 (_, ty::Array(_, n)) => {
2185 let len = n.eval_usize(tcx, param_env);
2186 Some(Slice(Slice { array_len: Some(len), kind: FixedLen(len) }))
2188 (ty::ConstKind::Value(ConstValue::Slice { start, end, .. }), ty::Slice(_)) => {
2189 let len = (end - start) as u64;
2190 Some(Slice(Slice { array_len: None, kind: FixedLen(len) }))
2192 // FIXME(oli-obk): implement `deref` for `ConstValue`
2193 // (ty::ConstKind::Value(ConstValue::ByRef { .. }), ty::Slice(_)) => { ... }
2194 _ => Some(ConstantValue(value)),
2198 PatKind::Range(PatRange { lo, hi, end }) => {
2200 if let Some(int_range) = IntRange::from_range(
2202 lo.eval_bits(tcx, param_env, lo.ty),
2203 hi.eval_bits(tcx, param_env, hi.ty),
2208 Some(IntRange(int_range))
2210 Some(FloatRange(lo, hi, end))
2213 PatKind::Array { ref prefix, ref slice, ref suffix }
2214 | PatKind::Slice { ref prefix, ref slice, ref suffix } => {
2215 let array_len = match pat.ty.kind() {
2216 ty::Array(_, length) => Some(length.eval_usize(tcx, param_env)),
2217 ty::Slice(_) => None,
2218 _ => span_bug!(pat.span, "bad ty {:?} for slice pattern", pat.ty),
2220 let prefix = prefix.len() as u64;
2221 let suffix = suffix.len() as u64;
2223 if slice.is_some() { VarLen(prefix, suffix) } else { FixedLen(prefix + suffix) };
2224 Some(Slice(Slice { array_len, kind }))
2226 PatKind::Or { .. } => bug!("Or-pattern should have been expanded earlier on."),
2230 // checks whether a constant is equal to a user-written slice pattern. Only supports byte slices,
2231 // meaning all other types will compare unequal and thus equal patterns often do not cause the
2232 // second pattern to lint about unreachable match arms.
2233 fn slice_pat_covered_by_const<'tcx>(
2236 const_val: &'tcx ty::Const<'tcx>,
2237 prefix: &[Pat<'tcx>],
2238 slice: &Option<Pat<'tcx>>,
2239 suffix: &[Pat<'tcx>],
2240 param_env: ty::ParamEnv<'tcx>,
2241 ) -> Result<bool, ErrorReported> {
2242 let const_val_val = if let ty::ConstKind::Value(val) = const_val.val {
2246 "slice_pat_covered_by_const: {:#?}, {:#?}, {:#?}, {:#?}",
2254 let data: &[u8] = match (const_val_val, &const_val.ty.kind()) {
2255 (ConstValue::ByRef { offset, alloc, .. }, ty::Array(t, n)) => {
2256 assert_eq!(*t, tcx.types.u8);
2257 let n = n.eval_usize(tcx, param_env);
2258 let ptr = Pointer::new(AllocId(0), offset);
2259 alloc.get_bytes(&tcx, ptr, Size::from_bytes(n)).unwrap()
2261 (ConstValue::Slice { data, start, end }, ty::Slice(t)) => {
2262 assert_eq!(*t, tcx.types.u8);
2263 let ptr = Pointer::new(AllocId(0), Size::from_bytes(start));
2264 data.get_bytes(&tcx, ptr, Size::from_bytes(end - start)).unwrap()
2266 // FIXME(oli-obk): create a way to extract fat pointers from ByRef
2267 (_, ty::Slice(_)) => return Ok(false),
2269 "slice_pat_covered_by_const: {:#?}, {:#?}, {:#?}, {:#?}",
2277 let pat_len = prefix.len() + suffix.len();
2278 if data.len() < pat_len || (slice.is_none() && data.len() > pat_len) {
2282 for (ch, pat) in data[..prefix.len()]
2285 .chain(data[data.len() - suffix.len()..].iter().zip(suffix))
2287 if let box PatKind::Constant { value } = pat.kind {
2288 let b = value.eval_bits(tcx, param_env, pat.ty);
2289 assert_eq!(b as u8 as u128, b);
2299 /// For exhaustive integer matching, some constructors are grouped within other constructors
2300 /// (namely integer typed values are grouped within ranges). However, when specialising these
2301 /// constructors, we want to be specialising for the underlying constructors (the integers), not
2302 /// the groups (the ranges). Thus we need to split the groups up. Splitting them up naïvely would
2303 /// mean creating a separate constructor for every single value in the range, which is clearly
2304 /// impractical. However, observe that for some ranges of integers, the specialisation will be
2305 /// identical across all values in that range (i.e., there are equivalence classes of ranges of
2306 /// constructors based on their `is_useful_specialized` outcome). These classes are grouped by
2307 /// the patterns that apply to them (in the matrix `P`). We can split the range whenever the
2308 /// patterns that apply to that range (specifically: the patterns that *intersect* with that range)
2310 /// Our solution, therefore, is to split the range constructor into subranges at every single point
2311 /// the group of intersecting patterns changes (using the method described below).
2312 /// And voilà! We're testing precisely those ranges that we need to, without any exhaustive matching
2313 /// on actual integers. The nice thing about this is that the number of subranges is linear in the
2314 /// number of rows in the matrix (i.e., the number of cases in the `match` statement), so we don't
2315 /// need to be worried about matching over gargantuan ranges.
2317 /// Essentially, given the first column of a matrix representing ranges, looking like the following:
2319 /// |------| |----------| |-------| ||
2320 /// |-------| |-------| |----| ||
2323 /// We split the ranges up into equivalence classes so the ranges are no longer overlapping:
2325 /// |--|--|||-||||--||---|||-------| |-|||| ||
2327 /// The logic for determining how to split the ranges is fairly straightforward: we calculate
2328 /// boundaries for each interval range, sort them, then create constructors for each new interval
2329 /// between every pair of boundary points. (This essentially sums up to performing the intuitive
2330 /// merging operation depicted above.)
2332 /// `hir_id` is `None` when we're evaluating the wildcard pattern, do not lint for overlapping in
2333 /// ranges that case.
2335 /// This also splits variable-length slices into fixed-length slices.
2336 fn split_grouped_constructors<'p, 'tcx>(
2338 param_env: ty::ParamEnv<'tcx>,
2340 ctors: Vec<Constructor<'tcx>>,
2341 matrix: &Matrix<'p, 'tcx>,
2343 hir_id: Option<HirId>,
2344 ) -> Vec<Constructor<'tcx>> {
2346 let mut split_ctors = Vec::with_capacity(ctors.len());
2347 debug!("split_grouped_constructors({:#?}, {:#?})", matrix, ctors);
2349 for ctor in ctors.into_iter() {
2351 IntRange(ctor_range) if ctor_range.treat_exhaustively(tcx) => {
2352 // Fast-track if the range is trivial. In particular, don't do the overlapping
2354 if ctor_range.is_singleton() {
2355 split_ctors.push(IntRange(ctor_range));
2359 /// Represents a border between 2 integers. Because the intervals spanning borders
2360 /// must be able to cover every integer, we need to be able to represent
2361 /// 2^128 + 1 such borders.
2362 #[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Debug)]
2368 // A function for extracting the borders of an integer interval.
2369 fn range_borders(r: IntRange<'_>) -> impl Iterator<Item = Border> {
2370 let (lo, hi) = r.range.into_inner();
2371 let from = Border::JustBefore(lo);
2372 let to = match hi.checked_add(1) {
2373 Some(m) => Border::JustBefore(m),
2374 None => Border::AfterMax,
2376 vec![from, to].into_iter()
2379 // Collect the span and range of all the intersecting ranges to lint on likely
2380 // incorrect range patterns. (#63987)
2381 let mut overlaps = vec![];
2382 // `borders` is the set of borders between equivalence classes: each equivalence
2383 // class lies between 2 borders.
2384 let row_borders = matrix
2388 IntRange::from_pat(tcx, param_env, row.head()).map(|r| (r, row.len()))
2390 .flat_map(|(range, row_len)| {
2391 let intersection = ctor_range.intersection(tcx, &range);
2392 let should_lint = ctor_range.suspicious_intersection(&range);
2393 if let (Some(range), 1, true) = (&intersection, row_len, should_lint) {
2394 // FIXME: for now, only check for overlapping ranges on simple range
2395 // patterns. Otherwise with the current logic the following is detected
2397 // match (10u8, true) {
2398 // (0 ..= 125, false) => {}
2399 // (126 ..= 255, false) => {}
2400 // (0 ..= 255, true) => {}
2402 overlaps.push(range.clone());
2406 .flat_map(range_borders);
2407 let ctor_borders = range_borders(ctor_range.clone());
2408 let mut borders: Vec<_> = row_borders.chain(ctor_borders).collect();
2409 borders.sort_unstable();
2411 lint_overlapping_patterns(tcx, hir_id, ctor_range, ty, overlaps);
2413 // We're going to iterate through every adjacent pair of borders, making sure that
2414 // each represents an interval of nonnegative length, and convert each such
2415 // interval into a constructor.
2419 .filter_map(|window| match (window[0], window[1]) {
2420 (Border::JustBefore(n), Border::JustBefore(m)) => {
2422 Some(IntRange { range: n..=(m - 1), ty, span })
2427 (Border::JustBefore(n), Border::AfterMax) => {
2428 Some(IntRange { range: n..=u128::MAX, ty, span })
2430 (Border::AfterMax, _) => None,
2435 Slice(Slice { array_len, kind: VarLen(self_prefix, self_suffix) }) => {
2436 // The exhaustiveness-checking paper does not include any details on
2437 // checking variable-length slice patterns. However, they are matched
2438 // by an infinite collection of fixed-length array patterns.
2440 // Checking the infinite set directly would take an infinite amount
2441 // of time. However, it turns out that for each finite set of
2442 // patterns `P`, all sufficiently large array lengths are equivalent:
2444 // Each slice `s` with a "sufficiently-large" length `l ≥ L` that applies
2445 // to exactly the subset `Pₜ` of `P` can be transformed to a slice
2446 // `sₘ` for each sufficiently-large length `m` that applies to exactly
2447 // the same subset of `P`.
2449 // Because of that, each witness for reachability-checking from one
2450 // of the sufficiently-large lengths can be transformed to an
2451 // equally-valid witness from any other length, so we only have
2452 // to check slice lengths from the "minimal sufficiently-large length"
2455 // Note that the fact that there is a *single* `sₘ` for each `m`
2456 // not depending on the specific pattern in `P` is important: if
2457 // you look at the pair of patterns
2460 // Then any slice of length ≥1 that matches one of these two
2461 // patterns can be trivially turned to a slice of any
2462 // other length ≥1 that matches them and vice-versa - for
2463 // but the slice from length 2 `[false, true]` that matches neither
2464 // of these patterns can't be turned to a slice from length 1 that
2465 // matches neither of these patterns, so we have to consider
2466 // slices from length 2 there.
2468 // Now, to see that that length exists and find it, observe that slice
2469 // patterns are either "fixed-length" patterns (`[_, _, _]`) or
2470 // "variable-length" patterns (`[_, .., _]`).
2472 // For fixed-length patterns, all slices with lengths *longer* than
2473 // the pattern's length have the same outcome (of not matching), so
2474 // as long as `L` is greater than the pattern's length we can pick
2475 // any `sₘ` from that length and get the same result.
2477 // For variable-length patterns, the situation is more complicated,
2478 // because as seen above the precise value of `sₘ` matters.
2480 // However, for each variable-length pattern `p` with a prefix of length
2481 // `plₚ` and suffix of length `slₚ`, only the first `plₚ` and the last
2482 // `slₚ` elements are examined.
2484 // Therefore, as long as `L` is positive (to avoid concerns about empty
2485 // types), all elements after the maximum prefix length and before
2486 // the maximum suffix length are not examined by any variable-length
2487 // pattern, and therefore can be added/removed without affecting
2488 // them - creating equivalent patterns from any sufficiently-large
2491 // Of course, if fixed-length patterns exist, we must be sure
2492 // that our length is large enough to miss them all, so
2493 // we can pick `L = max(max(FIXED_LEN)+1, max(PREFIX_LEN) + max(SUFFIX_LEN))`
2495 // for example, with the above pair of patterns, all elements
2496 // but the first and last can be added/removed, so any
2497 // witness of length ≥2 (say, `[false, false, true]`) can be
2498 // turned to a witness from any other length ≥2.
2500 let mut max_prefix_len = self_prefix;
2501 let mut max_suffix_len = self_suffix;
2502 let mut max_fixed_len = 0;
2505 matrix.heads().filter_map(|pat| pat_constructor(tcx, param_env, pat));
2506 for ctor in head_ctors {
2507 if let Slice(slice) = ctor {
2508 match slice.pattern_kind() {
2510 max_fixed_len = cmp::max(max_fixed_len, len);
2512 VarLen(prefix, suffix) => {
2513 max_prefix_len = cmp::max(max_prefix_len, prefix);
2514 max_suffix_len = cmp::max(max_suffix_len, suffix);
2520 // For diagnostics, we keep the prefix and suffix lengths separate, so in the case
2521 // where `max_fixed_len + 1` is the largest, we adapt `max_prefix_len` accordingly,
2522 // so that `L = max_prefix_len + max_suffix_len`.
2523 if max_fixed_len + 1 >= max_prefix_len + max_suffix_len {
2524 // The subtraction can't overflow thanks to the above check.
2525 // The new `max_prefix_len` is also guaranteed to be larger than its previous
2527 max_prefix_len = max_fixed_len + 1 - max_suffix_len;
2532 let kind = if max_prefix_len + max_suffix_len < len {
2533 VarLen(max_prefix_len, max_suffix_len)
2537 split_ctors.push(Slice(Slice { array_len, kind }));
2540 // `ctor` originally covered the range `(self_prefix +
2541 // self_suffix..infinity)`. We now split it into two: lengths smaller than
2542 // `max_prefix_len + max_suffix_len` are treated independently as
2543 // fixed-lengths slices, and lengths above are captured by a final VarLen
2546 (self_prefix + self_suffix..max_prefix_len + max_suffix_len)
2547 .map(|len| Slice(Slice { array_len, kind: FixedLen(len) })),
2549 split_ctors.push(Slice(Slice {
2551 kind: VarLen(max_prefix_len, max_suffix_len),
2556 // Any other constructor can be used unchanged.
2557 _ => split_ctors.push(ctor),
2561 debug!("split_grouped_constructors(..)={:#?}", split_ctors);
2565 fn lint_overlapping_patterns<'tcx>(
2567 hir_id: Option<HirId>,
2568 ctor_range: IntRange<'tcx>,
2570 overlaps: Vec<IntRange<'tcx>>,
2572 if let (true, Some(hir_id)) = (!overlaps.is_empty(), hir_id) {
2573 tcx.struct_span_lint_hir(
2574 lint::builtin::OVERLAPPING_PATTERNS,
2578 let mut err = lint.build("multiple patterns covering the same range");
2579 err.span_label(ctor_range.span, "overlapping patterns");
2580 for int_range in overlaps {
2581 // Use the real type for user display of the ranges:
2585 "this range overlaps on `{}`",
2586 IntRange { range: int_range.range, ty, span: DUMMY_SP }.to_pat(tcx),
2596 fn constructor_covered_by_range<'tcx>(
2598 param_env: ty::ParamEnv<'tcx>,
2599 ctor: &Constructor<'tcx>,
2602 if let Single = ctor {
2606 let (pat_from, pat_to, pat_end, ty) = match *pat.kind {
2607 PatKind::Constant { value } => (value, value, RangeEnd::Included, value.ty),
2608 PatKind::Range(PatRange { lo, hi, end }) => (lo, hi, end, lo.ty),
2609 _ => bug!("`constructor_covered_by_range` called with {:?}", pat),
2611 let (ctor_from, ctor_to, ctor_end) = match *ctor {
2612 ConstantValue(value) => (value, value, RangeEnd::Included),
2613 FloatRange(from, to, ctor_end) => (from, to, ctor_end),
2614 _ => bug!("`constructor_covered_by_range` called with {:?}", ctor),
2616 trace!("constructor_covered_by_range {:#?}, {:#?}, {:#?}, {}", ctor, pat_from, pat_to, ty);
2618 let to = compare_const_vals(tcx, ctor_to, pat_to, param_env, ty)?;
2619 let from = compare_const_vals(tcx, ctor_from, pat_from, param_env, ty)?;
2620 let intersects = (from == Ordering::Greater || from == Ordering::Equal)
2621 && (to == Ordering::Less || (pat_end == ctor_end && to == Ordering::Equal));
2622 if intersects { Some(()) } else { None }
2625 /// This is the main specialization step. It expands the pattern
2626 /// into `arity` patterns based on the constructor. For most patterns, the step is trivial,
2627 /// for instance tuple patterns are flattened and box patterns expand into their inner pattern.
2628 /// Returns `None` if the pattern does not have the given constructor.
2630 /// OTOH, slice patterns with a subslice pattern (tail @ ..) can be expanded into multiple
2631 /// different patterns.
2632 /// Structure patterns with a partial wild pattern (Foo { a: 42, .. }) have their missing
2633 /// fields filled with wild patterns.
2635 /// This is roughly the inverse of `Constructor::apply`.
2636 fn specialize_one_pattern<'p, 'tcx>(
2637 cx: &mut MatchCheckCtxt<'p, 'tcx>,
2639 constructor: &Constructor<'tcx>,
2640 ctor_wild_subpatterns: &Fields<'p, 'tcx>,
2641 ) -> Option<Fields<'p, 'tcx>> {
2642 if let NonExhaustive = constructor {
2643 // Only a wildcard pattern can match the special extra constructor
2644 if !pat.is_wildcard() {
2647 return Some(Fields::empty());
2650 let result = match *pat.kind {
2651 PatKind::AscribeUserType { .. } => bug!(), // Handled by `expand_pattern`
2653 PatKind::Binding { .. } | PatKind::Wild => Some(ctor_wild_subpatterns.clone()),
2655 PatKind::Variant { adt_def, variant_index, ref subpatterns, .. } => {
2656 let variant = &adt_def.variants[variant_index];
2657 if constructor != &Variant(variant.def_id) {
2660 Some(ctor_wild_subpatterns.replace_with_fieldpats(subpatterns))
2663 PatKind::Leaf { ref subpatterns } => {
2664 Some(ctor_wild_subpatterns.replace_with_fieldpats(subpatterns))
2667 PatKind::Deref { ref subpattern } => Some(Fields::from_single_pattern(subpattern)),
2669 PatKind::Constant { value } if constructor.is_slice() => {
2670 // We extract an `Option` for the pointer because slices of zero
2671 // elements don't necessarily point to memory, they are usually
2672 // just integers. The only time they should be pointing to memory
2673 // is when they are subslices of nonzero slices.
2674 let (alloc, offset, n, ty) = match value.ty.kind() {
2675 ty::Array(t, n) => {
2676 let n = n.eval_usize(cx.tcx, cx.param_env);
2677 // Shortcut for `n == 0` where no matter what `alloc` and `offset` we produce,
2678 // the result would be exactly what we early return here.
2680 if ctor_wild_subpatterns.len() as u64 != n {
2683 return Some(Fields::empty());
2686 ty::ConstKind::Value(ConstValue::ByRef { offset, alloc, .. }) => {
2687 (Cow::Borrowed(alloc), offset, n, t)
2689 _ => span_bug!(pat.span, "array pattern is {:?}", value,),
2694 ty::ConstKind::Value(ConstValue::Slice { data, start, end }) => {
2695 let offset = Size::from_bytes(start);
2696 let n = (end - start) as u64;
2697 (Cow::Borrowed(data), offset, n, t)
2699 ty::ConstKind::Value(ConstValue::ByRef { .. }) => {
2700 // FIXME(oli-obk): implement `deref` for `ConstValue`
2705 "slice pattern constant must be scalar pair but is {:?}",
2712 "unexpected const-val {:?} with ctor {:?}",
2717 if ctor_wild_subpatterns.len() as u64 != n {
2721 // Convert a constant slice/array pattern to a list of patterns.
2722 let layout = cx.tcx.layout_of(cx.param_env.and(ty)).ok()?;
2723 let ptr = Pointer::new(AllocId(0), offset);
2724 let pats = cx.pattern_arena.alloc_from_iter((0..n).filter_map(|i| {
2725 let ptr = ptr.offset(layout.size * i, &cx.tcx).ok()?;
2726 let scalar = alloc.read_scalar(&cx.tcx, ptr, layout.size).ok()?;
2727 let scalar = scalar.check_init().ok()?;
2728 let value = ty::Const::from_scalar(cx.tcx, scalar, ty);
2729 let pattern = Pat { ty, span: pat.span, kind: box PatKind::Constant { value } };
2732 // Ensure none of the dereferences failed.
2733 if pats.len() as u64 != n {
2736 Some(Fields::from_slice_unfiltered(pats))
2739 PatKind::Constant { .. } | PatKind::Range { .. } => {
2740 // If the constructor is a:
2741 // - Single value: add a row if the pattern contains the constructor.
2742 // - Range: add a row if the constructor intersects the pattern.
2743 if let IntRange(ctor) = constructor {
2744 let pat = IntRange::from_pat(cx.tcx, cx.param_env, pat)?;
2745 ctor.intersection(cx.tcx, &pat)?;
2746 // Constructor splitting should ensure that all intersections we encounter
2747 // are actually inclusions.
2748 assert!(ctor.is_subrange(&pat));
2750 // Fallback for non-ranges and ranges that involve
2751 // floating-point numbers, which are not conveniently handled
2752 // by `IntRange`. For these cases, the constructor may not be a
2753 // range so intersection actually devolves into being covered
2755 constructor_covered_by_range(cx.tcx, cx.param_env, constructor, pat)?;
2757 Some(Fields::empty())
2760 PatKind::Array { ref prefix, ref slice, ref suffix }
2761 | PatKind::Slice { ref prefix, ref slice, ref suffix } => match *constructor {
2763 // Number of subpatterns for this pattern
2764 let pat_len = prefix.len() + suffix.len();
2765 // Number of subpatterns for this constructor
2766 let arity = ctor_wild_subpatterns.len();
2768 if (slice.is_none() && arity != pat_len) || pat_len > arity {
2772 // Replace the prefix and the suffix with the given patterns, leaving wildcards in
2773 // the middle if there was a subslice pattern `..`.
2774 let prefix = prefix.iter().enumerate();
2775 let suffix = suffix.iter().enumerate().map(|(i, p)| (arity - suffix.len() + i, p));
2776 Some(ctor_wild_subpatterns.replace_fields_indexed(prefix.chain(suffix)))
2778 ConstantValue(cv) => {
2779 match slice_pat_covered_by_const(
2788 Ok(true) => Some(Fields::empty()),
2790 Err(ErrorReported) => None,
2793 _ => span_bug!(pat.span, "unexpected ctor {:?} for slice pat", constructor),
2796 PatKind::Or { .. } => bug!("Or-pattern should have been expanded earlier on."),
2799 "specialize({:#?}, {:#?}, {:#?}) = {:#?}",
2800 pat, constructor, ctor_wild_subpatterns, result