diff options
Diffstat (limited to 'crates/hir_ty/src/diagnostics/match_check')
3 files changed, 2151 insertions, 0 deletions
diff --git a/crates/hir_ty/src/diagnostics/match_check/deconstruct_pat.rs b/crates/hir_ty/src/diagnostics/match_check/deconstruct_pat.rs new file mode 100644 index 000000000..1f4219b42 --- /dev/null +++ b/crates/hir_ty/src/diagnostics/match_check/deconstruct_pat.rs | |||
@@ -0,0 +1,907 @@ | |||
1 | //! [`super::usefulness`] explains most of what is happening in this file. As explained there, | ||
2 | //! values and patterns are made from constructors applied to fields. This file defines a | ||
3 | //! `Constructor` enum, a `Fields` struct, and various operations to manipulate them and convert | ||
4 | //! them from/to patterns. | ||
5 | //! | ||
6 | //! There's one idea that is not detailed in [`super::usefulness`] because the details are not | ||
7 | //! needed there: _constructor splitting_. | ||
8 | //! | ||
9 | //! # Constructor splitting | ||
10 | //! | ||
11 | //! The idea is as follows: given a constructor `c` and a matrix, we want to specialize in turn | ||
12 | //! with all the value constructors that are covered by `c`, and compute usefulness for each. | ||
13 | //! Instead of listing all those constructors (which is intractable), we group those value | ||
14 | //! constructors together as much as possible. Example: | ||
15 | //! | ||
16 | //! ``` | ||
17 | //! match (0, false) { | ||
18 | //! (0 ..=100, true) => {} // `p_1` | ||
19 | //! (50..=150, false) => {} // `p_2` | ||
20 | //! (0 ..=200, _) => {} // `q` | ||
21 | //! } | ||
22 | //! ``` | ||
23 | //! | ||
24 | //! The naive approach would try all numbers in the range `0..=200`. But we can be a lot more | ||
25 | //! clever: `0` and `1` for example will match the exact same rows, and return equivalent | ||
26 | //! witnesses. In fact all of `0..50` would. We can thus restrict our exploration to 4 | ||
27 | //! constructors: `0..50`, `50..=100`, `101..=150` and `151..=200`. That is enough and infinitely | ||
28 | //! more tractable. | ||
29 | //! | ||
30 | //! We capture this idea in a function `split(p_1 ... p_n, c)` which returns a list of constructors | ||
31 | //! `c'` covered by `c`. Given such a `c'`, we require that all value ctors `c''` covered by `c'` | ||
32 | //! return an equivalent set of witnesses after specializing and computing usefulness. | ||
33 | //! In the example above, witnesses for specializing by `c''` covered by `0..50` will only differ | ||
34 | //! in their first element. | ||
35 | //! | ||
36 | //! We usually also ask that the `c'` together cover all of the original `c`. However we allow | ||
37 | //! skipping some constructors as long as it doesn't change whether the resulting list of witnesses | ||
38 | //! is empty of not. We use this in the wildcard `_` case. | ||
39 | //! | ||
40 | //! Splitting is implemented in the [`Constructor::split`] function. We don't do splitting for | ||
41 | //! or-patterns; instead we just try the alternatives one-by-one. For details on splitting | ||
42 | //! wildcards, see [`SplitWildcard`]; for integer ranges, see [`SplitIntRange`]; for slices, see | ||
43 | //! [`SplitVarLenSlice`]. | ||
44 | |||
45 | use std::{ | ||
46 | cmp::{max, min}, | ||
47 | iter::once, | ||
48 | ops::RangeInclusive, | ||
49 | }; | ||
50 | |||
51 | use hir_def::{EnumVariantId, HasModule, LocalFieldId, VariantId}; | ||
52 | use smallvec::{smallvec, SmallVec}; | ||
53 | |||
54 | use crate::{AdtId, Interner, Scalar, Ty, TyExt, TyKind}; | ||
55 | |||
56 | use super::{ | ||
57 | usefulness::{MatchCheckCtx, PatCtxt}, | ||
58 | FieldPat, Pat, PatId, PatKind, | ||
59 | }; | ||
60 | |||
61 | use self::Constructor::*; | ||
62 | |||
63 | /// [Constructor] uses this in umimplemented variants. | ||
64 | /// It allows porting match expressions from upstream algorithm without losing semantics. | ||
65 | #[derive(Copy, Clone, Debug, PartialEq, Eq)] | ||
66 | pub(super) enum Void {} | ||
67 | |||
68 | /// An inclusive interval, used for precise integer exhaustiveness checking. | ||
69 | /// `IntRange`s always store a contiguous range. This means that values are | ||
70 | /// encoded such that `0` encodes the minimum value for the integer, | ||
71 | /// regardless of the signedness. | ||
72 | /// For example, the pattern `-128..=127i8` is encoded as `0..=255`. | ||
73 | /// This makes comparisons and arithmetic on interval endpoints much more | ||
74 | /// straightforward. See `signed_bias` for details. | ||
75 | /// | ||
76 | /// `IntRange` is never used to encode an empty range or a "range" that wraps | ||
77 | /// around the (offset) space: i.e., `range.lo <= range.hi`. | ||
78 | #[derive(Clone, Debug, PartialEq, Eq)] | ||
79 | pub(super) struct IntRange { | ||
80 | range: RangeInclusive<u128>, | ||
81 | } | ||
82 | |||
83 | impl IntRange { | ||
84 | #[inline] | ||
85 | fn is_integral(ty: &Ty) -> bool { | ||
86 | match ty.kind(&Interner) { | ||
87 | TyKind::Scalar(Scalar::Char) | ||
88 | | TyKind::Scalar(Scalar::Int(_)) | ||
89 | | TyKind::Scalar(Scalar::Uint(_)) | ||
90 | | TyKind::Scalar(Scalar::Bool) => true, | ||
91 | _ => false, | ||
92 | } | ||
93 | } | ||
94 | |||
95 | fn is_singleton(&self) -> bool { | ||
96 | self.range.start() == self.range.end() | ||
97 | } | ||
98 | |||
99 | fn boundaries(&self) -> (u128, u128) { | ||
100 | (*self.range.start(), *self.range.end()) | ||
101 | } | ||
102 | |||
103 | #[inline] | ||
104 | fn from_bool(value: bool) -> IntRange { | ||
105 | let val = value as u128; | ||
106 | IntRange { range: val..=val } | ||
107 | } | ||
108 | |||
109 | #[inline] | ||
110 | fn from_range(lo: u128, hi: u128, scalar_ty: Scalar) -> IntRange { | ||
111 | if let Scalar::Bool = scalar_ty { | ||
112 | IntRange { range: lo..=hi } | ||
113 | } else { | ||
114 | unimplemented!() | ||
115 | } | ||
116 | } | ||
117 | |||
118 | fn is_subrange(&self, other: &Self) -> bool { | ||
119 | other.range.start() <= self.range.start() && self.range.end() <= other.range.end() | ||
120 | } | ||
121 | |||
122 | fn intersection(&self, other: &Self) -> Option<Self> { | ||
123 | let (lo, hi) = self.boundaries(); | ||
124 | let (other_lo, other_hi) = other.boundaries(); | ||
125 | if lo <= other_hi && other_lo <= hi { | ||
126 | Some(IntRange { range: max(lo, other_lo)..=min(hi, other_hi) }) | ||
127 | } else { | ||
128 | None | ||
129 | } | ||
130 | } | ||
131 | |||
132 | /// See `Constructor::is_covered_by` | ||
133 | fn is_covered_by(&self, other: &Self) -> bool { | ||
134 | if self.intersection(other).is_some() { | ||
135 | // Constructor splitting should ensure that all intersections we encounter are actually | ||
136 | // inclusions. | ||
137 | assert!(self.is_subrange(other)); | ||
138 | true | ||
139 | } else { | ||
140 | false | ||
141 | } | ||
142 | } | ||
143 | } | ||
144 | |||
145 | /// Represents a border between 2 integers. Because the intervals spanning borders must be able to | ||
146 | /// cover every integer, we need to be able to represent 2^128 + 1 such borders. | ||
147 | #[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord)] | ||
148 | enum IntBorder { | ||
149 | JustBefore(u128), | ||
150 | AfterMax, | ||
151 | } | ||
152 | |||
153 | /// A range of integers that is partitioned into disjoint subranges. This does constructor | ||
154 | /// splitting for integer ranges as explained at the top of the file. | ||
155 | /// | ||
156 | /// This is fed multiple ranges, and returns an output that covers the input, but is split so that | ||
157 | /// the only intersections between an output range and a seen range are inclusions. No output range | ||
158 | /// straddles the boundary of one of the inputs. | ||
159 | /// | ||
160 | /// The following input: | ||
161 | /// ``` | ||
162 | /// |-------------------------| // `self` | ||
163 | /// |------| |----------| |----| | ||
164 | /// |-------| |-------| | ||
165 | /// ``` | ||
166 | /// would be iterated over as follows: | ||
167 | /// ``` | ||
168 | /// ||---|--||-|---|---|---|--| | ||
169 | /// ``` | ||
170 | #[derive(Debug, Clone)] | ||
171 | struct SplitIntRange { | ||
172 | /// The range we are splitting | ||
173 | range: IntRange, | ||
174 | /// The borders of ranges we have seen. They are all contained within `range`. This is kept | ||
175 | /// sorted. | ||
176 | borders: Vec<IntBorder>, | ||
177 | } | ||
178 | |||
179 | impl SplitIntRange { | ||
180 | fn new(range: IntRange) -> Self { | ||
181 | SplitIntRange { range, borders: Vec::new() } | ||
182 | } | ||
183 | |||
184 | /// Internal use | ||
185 | fn to_borders(r: IntRange) -> [IntBorder; 2] { | ||
186 | use IntBorder::*; | ||
187 | let (lo, hi) = r.boundaries(); | ||
188 | let lo = JustBefore(lo); | ||
189 | let hi = match hi.checked_add(1) { | ||
190 | Some(m) => JustBefore(m), | ||
191 | None => AfterMax, | ||
192 | }; | ||
193 | [lo, hi] | ||
194 | } | ||
195 | |||
196 | /// Add ranges relative to which we split. | ||
197 | fn split(&mut self, ranges: impl Iterator<Item = IntRange>) { | ||
198 | let this_range = &self.range; | ||
199 | let included_ranges = ranges.filter_map(|r| this_range.intersection(&r)); | ||
200 | let included_borders = included_ranges.flat_map(|r| { | ||
201 | let borders = Self::to_borders(r); | ||
202 | once(borders[0]).chain(once(borders[1])) | ||
203 | }); | ||
204 | self.borders.extend(included_borders); | ||
205 | self.borders.sort_unstable(); | ||
206 | } | ||
207 | |||
208 | /// Iterate over the contained ranges. | ||
209 | fn iter(&self) -> impl Iterator<Item = IntRange> + '_ { | ||
210 | use IntBorder::*; | ||
211 | |||
212 | let self_range = Self::to_borders(self.range.clone()); | ||
213 | // Start with the start of the range. | ||
214 | let mut prev_border = self_range[0]; | ||
215 | self.borders | ||
216 | .iter() | ||
217 | .copied() | ||
218 | // End with the end of the range. | ||
219 | .chain(once(self_range[1])) | ||
220 | // List pairs of adjacent borders. | ||
221 | .map(move |border| { | ||
222 | let ret = (prev_border, border); | ||
223 | prev_border = border; | ||
224 | ret | ||
225 | }) | ||
226 | // Skip duplicates. | ||
227 | .filter(|(prev_border, border)| prev_border != border) | ||
228 | // Finally, convert to ranges. | ||
229 | .map(|(prev_border, border)| { | ||
230 | let range = match (prev_border, border) { | ||
231 | (JustBefore(n), JustBefore(m)) if n < m => n..=(m - 1), | ||
232 | (JustBefore(n), AfterMax) => n..=u128::MAX, | ||
233 | _ => unreachable!(), // Ruled out by the sorting and filtering we did | ||
234 | }; | ||
235 | IntRange { range } | ||
236 | }) | ||
237 | } | ||
238 | } | ||
239 | |||
240 | /// A constructor for array and slice patterns. | ||
241 | #[derive(Copy, Clone, Debug, PartialEq, Eq)] | ||
242 | pub(super) struct Slice { | ||
243 | _unimplemented: Void, | ||
244 | } | ||
245 | |||
246 | impl Slice { | ||
247 | /// See `Constructor::is_covered_by` | ||
248 | fn is_covered_by(self, _other: Self) -> bool { | ||
249 | unimplemented!() // never called as Slice contains Void | ||
250 | } | ||
251 | } | ||
252 | |||
253 | /// A value can be decomposed into a constructor applied to some fields. This struct represents | ||
254 | /// the constructor. See also `Fields`. | ||
255 | /// | ||
256 | /// `pat_constructor` retrieves the constructor corresponding to a pattern. | ||
257 | /// `specialize_constructor` returns the list of fields corresponding to a pattern, given a | ||
258 | /// constructor. `Constructor::apply` reconstructs the pattern from a pair of `Constructor` and | ||
259 | /// `Fields`. | ||
260 | #[allow(dead_code)] | ||
261 | #[derive(Clone, Debug, PartialEq)] | ||
262 | pub(super) enum Constructor { | ||
263 | /// The constructor for patterns that have a single constructor, like tuples, struct patterns | ||
264 | /// and fixed-length arrays. | ||
265 | Single, | ||
266 | /// Enum variants. | ||
267 | Variant(EnumVariantId), | ||
268 | /// Ranges of integer literal values (`2`, `2..=5` or `2..5`). | ||
269 | IntRange(IntRange), | ||
270 | /// Ranges of floating-point literal values (`2.0..=5.2`). | ||
271 | FloatRange(Void), | ||
272 | /// String literals. Strings are not quite the same as `&[u8]` so we treat them separately. | ||
273 | Str(Void), | ||
274 | /// Array and slice patterns. | ||
275 | Slice(Slice), | ||
276 | /// Constants that must not be matched structurally. They are treated as black | ||
277 | /// boxes for the purposes of exhaustiveness: we must not inspect them, and they | ||
278 | /// don't count towards making a match exhaustive. | ||
279 | Opaque, | ||
280 | /// Fake extra constructor for enums that aren't allowed to be matched exhaustively. Also used | ||
281 | /// for those types for which we cannot list constructors explicitly, like `f64` and `str`. | ||
282 | NonExhaustive, | ||
283 | /// Stands for constructors that are not seen in the matrix, as explained in the documentation | ||
284 | /// for [`SplitWildcard`]. | ||
285 | Missing, | ||
286 | /// Wildcard pattern. | ||
287 | Wildcard, | ||
288 | } | ||
289 | |||
290 | impl Constructor { | ||
291 | pub(super) fn is_wildcard(&self) -> bool { | ||
292 | matches!(self, Wildcard) | ||
293 | } | ||
294 | |||
295 | fn as_int_range(&self) -> Option<&IntRange> { | ||
296 | match self { | ||
297 | IntRange(range) => Some(range), | ||
298 | _ => None, | ||
299 | } | ||
300 | } | ||
301 | |||
302 | fn as_slice(&self) -> Option<Slice> { | ||
303 | match self { | ||
304 | Slice(slice) => Some(*slice), | ||
305 | _ => None, | ||
306 | } | ||
307 | } | ||
308 | |||
309 | fn variant_id_for_adt(&self, adt: hir_def::AdtId) -> VariantId { | ||
310 | match *self { | ||
311 | Variant(id) => id.into(), | ||
312 | Single => { | ||
313 | assert!(!matches!(adt, hir_def::AdtId::EnumId(_))); | ||
314 | match adt { | ||
315 | hir_def::AdtId::EnumId(_) => unreachable!(), | ||
316 | hir_def::AdtId::StructId(id) => id.into(), | ||
317 | hir_def::AdtId::UnionId(id) => id.into(), | ||
318 | } | ||
319 | } | ||
320 | _ => panic!("bad constructor {:?} for adt {:?}", self, adt), | ||
321 | } | ||
322 | } | ||
323 | |||
324 | /// Determines the constructor that the given pattern can be specialized to. | ||
325 | pub(super) fn from_pat(cx: &MatchCheckCtx<'_>, pat: PatId) -> Self { | ||
326 | match cx.pattern_arena.borrow()[pat].kind.as_ref() { | ||
327 | PatKind::Binding { .. } | PatKind::Wild => Wildcard, | ||
328 | PatKind::Leaf { .. } | PatKind::Deref { .. } => Single, | ||
329 | &PatKind::Variant { enum_variant, .. } => Variant(enum_variant), | ||
330 | &PatKind::LiteralBool { value } => IntRange(IntRange::from_bool(value)), | ||
331 | PatKind::Or { .. } => cx.bug("Or-pattern should have been expanded earlier on."), | ||
332 | } | ||
333 | } | ||
334 | |||
335 | /// Some constructors (namely `Wildcard`, `IntRange` and `Slice`) actually stand for a set of actual | ||
336 | /// constructors (like variants, integers or fixed-sized slices). When specializing for these | ||
337 | /// constructors, we want to be specialising for the actual underlying constructors. | ||
338 | /// Naively, we would simply return the list of constructors they correspond to. We instead are | ||
339 | /// more clever: if there are constructors that we know will behave the same wrt the current | ||
340 | /// matrix, we keep them grouped. For example, all slices of a sufficiently large length | ||
341 | /// will either be all useful or all non-useful with a given matrix. | ||
342 | /// | ||
343 | /// See the branches for details on how the splitting is done. | ||
344 | /// | ||
345 | /// This function may discard some irrelevant constructors if this preserves behavior and | ||
346 | /// diagnostics. Eg. for the `_` case, we ignore the constructors already present in the | ||
347 | /// matrix, unless all of them are. | ||
348 | pub(super) fn split<'a>( | ||
349 | &self, | ||
350 | pcx: PatCtxt<'_>, | ||
351 | ctors: impl Iterator<Item = &'a Constructor> + Clone, | ||
352 | ) -> SmallVec<[Self; 1]> { | ||
353 | match self { | ||
354 | Wildcard => { | ||
355 | let mut split_wildcard = SplitWildcard::new(pcx); | ||
356 | split_wildcard.split(pcx, ctors); | ||
357 | split_wildcard.into_ctors(pcx) | ||
358 | } | ||
359 | // Fast-track if the range is trivial. In particular, we don't do the overlapping | ||
360 | // ranges check. | ||
361 | IntRange(ctor_range) if !ctor_range.is_singleton() => { | ||
362 | let mut split_range = SplitIntRange::new(ctor_range.clone()); | ||
363 | let int_ranges = ctors.filter_map(|ctor| ctor.as_int_range()); | ||
364 | split_range.split(int_ranges.cloned()); | ||
365 | split_range.iter().map(IntRange).collect() | ||
366 | } | ||
367 | Slice(_) => unimplemented!(), | ||
368 | // Any other constructor can be used unchanged. | ||
369 | _ => smallvec![self.clone()], | ||
370 | } | ||
371 | } | ||
372 | |||
373 | /// Returns whether `self` is covered by `other`, i.e. whether `self` is a subset of `other`. | ||
374 | /// For the simple cases, this is simply checking for equality. For the "grouped" constructors, | ||
375 | /// this checks for inclusion. | ||
376 | // We inline because this has a single call site in `Matrix::specialize_constructor`. | ||
377 | #[inline] | ||
378 | pub(super) fn is_covered_by(&self, pcx: PatCtxt<'_>, other: &Self) -> bool { | ||
379 | // This must be kept in sync with `is_covered_by_any`. | ||
380 | match (self, other) { | ||
381 | // Wildcards cover anything | ||
382 | (_, Wildcard) => true, | ||
383 | // The missing ctors are not covered by anything in the matrix except wildcards. | ||
384 | (Missing, _) | (Wildcard, _) => false, | ||
385 | |||
386 | (Single, Single) => true, | ||
387 | (Variant(self_id), Variant(other_id)) => self_id == other_id, | ||
388 | |||
389 | (IntRange(self_range), IntRange(other_range)) => self_range.is_covered_by(other_range), | ||
390 | (FloatRange(..), FloatRange(..)) => { | ||
391 | unimplemented!() | ||
392 | } | ||
393 | (Str(..), Str(..)) => { | ||
394 | unimplemented!() | ||
395 | } | ||
396 | (Slice(self_slice), Slice(other_slice)) => self_slice.is_covered_by(*other_slice), | ||
397 | |||
398 | // We are trying to inspect an opaque constant. Thus we skip the row. | ||
399 | (Opaque, _) | (_, Opaque) => false, | ||
400 | // Only a wildcard pattern can match the special extra constructor. | ||
401 | (NonExhaustive, _) => false, | ||
402 | |||
403 | _ => pcx.cx.bug(&format!( | ||
404 | "trying to compare incompatible constructors {:?} and {:?}", | ||
405 | self, other | ||
406 | )), | ||
407 | } | ||
408 | } | ||
409 | |||
410 | /// Faster version of `is_covered_by` when applied to many constructors. `used_ctors` is | ||
411 | /// assumed to be built from `matrix.head_ctors()` with wildcards filtered out, and `self` is | ||
412 | /// assumed to have been split from a wildcard. | ||
413 | fn is_covered_by_any(&self, pcx: PatCtxt<'_>, used_ctors: &[Constructor]) -> bool { | ||
414 | if used_ctors.is_empty() { | ||
415 | return false; | ||
416 | } | ||
417 | |||
418 | // This must be kept in sync with `is_covered_by`. | ||
419 | match self { | ||
420 | // If `self` is `Single`, `used_ctors` cannot contain anything else than `Single`s. | ||
421 | Single => !used_ctors.is_empty(), | ||
422 | Variant(_) => used_ctors.iter().any(|c| c == self), | ||
423 | IntRange(range) => used_ctors | ||
424 | .iter() | ||
425 | .filter_map(|c| c.as_int_range()) | ||
426 | .any(|other| range.is_covered_by(other)), | ||
427 | Slice(slice) => used_ctors | ||
428 | .iter() | ||
429 | .filter_map(|c| c.as_slice()) | ||
430 | .any(|other| slice.is_covered_by(other)), | ||
431 | // This constructor is never covered by anything else | ||
432 | NonExhaustive => false, | ||
433 | Str(..) | FloatRange(..) | Opaque | Missing | Wildcard => { | ||
434 | pcx.cx.bug(&format!("found unexpected ctor in all_ctors: {:?}", self)) | ||
435 | } | ||
436 | } | ||
437 | } | ||
438 | } | ||
439 | |||
440 | /// A wildcard constructor that we split relative to the constructors in the matrix, as explained | ||
441 | /// at the top of the file. | ||
442 | /// | ||
443 | /// A constructor that is not present in the matrix rows will only be covered by the rows that have | ||
444 | /// wildcards. Thus we can group all of those constructors together; we call them "missing | ||
445 | /// constructors". Splitting a wildcard would therefore list all present constructors individually | ||
446 | /// (or grouped if they are integers or slices), and then all missing constructors together as a | ||
447 | /// group. | ||
448 | /// | ||
449 | /// However we can go further: since any constructor will match the wildcard rows, and having more | ||
450 | /// rows can only reduce the amount of usefulness witnesses, we can skip the present constructors | ||
451 | /// and only try the missing ones. | ||
452 | /// This will not preserve the whole list of witnesses, but will preserve whether the list is empty | ||
453 | /// or not. In fact this is quite natural from the point of view of diagnostics too. This is done | ||
454 | /// in `to_ctors`: in some cases we only return `Missing`. | ||
455 | #[derive(Debug)] | ||
456 | pub(super) struct SplitWildcard { | ||
457 | /// Constructors seen in the matrix. | ||
458 | matrix_ctors: Vec<Constructor>, | ||
459 | /// All the constructors for this type | ||
460 | all_ctors: SmallVec<[Constructor; 1]>, | ||
461 | } | ||
462 | |||
463 | impl SplitWildcard { | ||
464 | pub(super) fn new(pcx: PatCtxt<'_>) -> Self { | ||
465 | let cx = pcx.cx; | ||
466 | let make_range = |start, end, scalar| IntRange(IntRange::from_range(start, end, scalar)); | ||
467 | |||
468 | // Unhandled types are treated as non-exhaustive. Being explicit here instead of falling | ||
469 | // to catchall arm to ease further implementation. | ||
470 | let unhandled = || smallvec![NonExhaustive]; | ||
471 | |||
472 | // This determines the set of all possible constructors for the type `pcx.ty`. For numbers, | ||
473 | // arrays and slices we use ranges and variable-length slices when appropriate. | ||
474 | // | ||
475 | // If the `exhaustive_patterns` feature is enabled, we make sure to omit constructors that | ||
476 | // are statically impossible. E.g., for `Option<!>`, we do not include `Some(_)` in the | ||
477 | // returned list of constructors. | ||
478 | // Invariant: this is empty if and only if the type is uninhabited (as determined by | ||
479 | // `cx.is_uninhabited()`). | ||
480 | let all_ctors = match pcx.ty.kind(&Interner) { | ||
481 | TyKind::Scalar(Scalar::Bool) => smallvec![make_range(0, 1, Scalar::Bool)], | ||
482 | // TyKind::Array(..) if ... => unhandled(), | ||
483 | TyKind::Array(..) | TyKind::Slice(..) => unhandled(), | ||
484 | &TyKind::Adt(AdtId(hir_def::AdtId::EnumId(enum_id)), ref _substs) => { | ||
485 | let enum_data = cx.db.enum_data(enum_id); | ||
486 | |||
487 | // If the enum is declared as `#[non_exhaustive]`, we treat it as if it had an | ||
488 | // additional "unknown" constructor. | ||
489 | // There is no point in enumerating all possible variants, because the user can't | ||
490 | // actually match against them all themselves. So we always return only the fictitious | ||
491 | // constructor. | ||
492 | // E.g., in an example like: | ||
493 | // | ||
494 | // ``` | ||
495 | // let err: io::ErrorKind = ...; | ||
496 | // match err { | ||
497 | // io::ErrorKind::NotFound => {}, | ||
498 | // } | ||
499 | // ``` | ||
500 | // | ||
501 | // we don't want to show every possible IO error, but instead have only `_` as the | ||
502 | // witness. | ||
503 | let is_declared_nonexhaustive = cx.is_foreign_non_exhaustive_enum(enum_id); | ||
504 | |||
505 | // If `exhaustive_patterns` is disabled and our scrutinee is an empty enum, we treat it | ||
506 | // as though it had an "unknown" constructor to avoid exposing its emptiness. The | ||
507 | // exception is if the pattern is at the top level, because we want empty matches to be | ||
508 | // considered exhaustive. | ||
509 | let is_secretly_empty = enum_data.variants.is_empty() | ||
510 | && !cx.feature_exhaustive_patterns() | ||
511 | && !pcx.is_top_level; | ||
512 | |||
513 | if is_secretly_empty || is_declared_nonexhaustive { | ||
514 | smallvec![NonExhaustive] | ||
515 | } else if cx.feature_exhaustive_patterns() { | ||
516 | unimplemented!() // see MatchCheckCtx.feature_exhaustive_patterns() | ||
517 | } else { | ||
518 | enum_data | ||
519 | .variants | ||
520 | .iter() | ||
521 | .map(|(local_id, ..)| Variant(EnumVariantId { parent: enum_id, local_id })) | ||
522 | .collect() | ||
523 | } | ||
524 | } | ||
525 | TyKind::Scalar(Scalar::Char) => unhandled(), | ||
526 | TyKind::Scalar(Scalar::Int(..)) | TyKind::Scalar(Scalar::Uint(..)) => unhandled(), | ||
527 | TyKind::Never if !cx.feature_exhaustive_patterns() && !pcx.is_top_level => { | ||
528 | smallvec![NonExhaustive] | ||
529 | } | ||
530 | TyKind::Never => SmallVec::new(), | ||
531 | _ if cx.is_uninhabited(&pcx.ty) => SmallVec::new(), | ||
532 | TyKind::Adt(..) | TyKind::Tuple(..) | TyKind::Ref(..) => smallvec![Single], | ||
533 | // This type is one for which we cannot list constructors, like `str` or `f64`. | ||
534 | _ => smallvec![NonExhaustive], | ||
535 | }; | ||
536 | SplitWildcard { matrix_ctors: Vec::new(), all_ctors } | ||
537 | } | ||
538 | |||
539 | /// Pass a set of constructors relative to which to split this one. Don't call twice, it won't | ||
540 | /// do what you want. | ||
541 | pub(super) fn split<'a>( | ||
542 | &mut self, | ||
543 | pcx: PatCtxt<'_>, | ||
544 | ctors: impl Iterator<Item = &'a Constructor> + Clone, | ||
545 | ) { | ||
546 | // Since `all_ctors` never contains wildcards, this won't recurse further. | ||
547 | self.all_ctors = | ||
548 | self.all_ctors.iter().flat_map(|ctor| ctor.split(pcx, ctors.clone())).collect(); | ||
549 | self.matrix_ctors = ctors.filter(|c| !c.is_wildcard()).cloned().collect(); | ||
550 | } | ||
551 | |||
552 | /// Whether there are any value constructors for this type that are not present in the matrix. | ||
553 | fn any_missing(&self, pcx: PatCtxt<'_>) -> bool { | ||
554 | self.iter_missing(pcx).next().is_some() | ||
555 | } | ||
556 | |||
557 | /// Iterate over the constructors for this type that are not present in the matrix. | ||
558 | pub(super) fn iter_missing<'a>( | ||
559 | &'a self, | ||
560 | pcx: PatCtxt<'a>, | ||
561 | ) -> impl Iterator<Item = &'a Constructor> { | ||
562 | self.all_ctors.iter().filter(move |ctor| !ctor.is_covered_by_any(pcx, &self.matrix_ctors)) | ||
563 | } | ||
564 | |||
565 | /// Return the set of constructors resulting from splitting the wildcard. As explained at the | ||
566 | /// top of the file, if any constructors are missing we can ignore the present ones. | ||
567 | fn into_ctors(self, pcx: PatCtxt<'_>) -> SmallVec<[Constructor; 1]> { | ||
568 | if self.any_missing(pcx) { | ||
569 | // Some constructors are missing, thus we can specialize with the special `Missing` | ||
570 | // constructor, which stands for those constructors that are not seen in the matrix, | ||
571 | // and matches the same rows as any of them (namely the wildcard rows). See the top of | ||
572 | // the file for details. | ||
573 | // However, when all constructors are missing we can also specialize with the full | ||
574 | // `Wildcard` constructor. The difference will depend on what we want in diagnostics. | ||
575 | |||
576 | // If some constructors are missing, we typically want to report those constructors, | ||
577 | // e.g.: | ||
578 | // ``` | ||
579 | // enum Direction { N, S, E, W } | ||
580 | // let Direction::N = ...; | ||
581 | // ``` | ||
582 | // we can report 3 witnesses: `S`, `E`, and `W`. | ||
583 | // | ||
584 | // However, if the user didn't actually specify a constructor | ||
585 | // in this arm, e.g., in | ||
586 | // ``` | ||
587 | // let x: (Direction, Direction, bool) = ...; | ||
588 | // let (_, _, false) = x; | ||
589 | // ``` | ||
590 | // we don't want to show all 16 possible witnesses `(<direction-1>, <direction-2>, | ||
591 | // true)` - we are satisfied with `(_, _, true)`. So if all constructors are missing we | ||
592 | // prefer to report just a wildcard `_`. | ||
593 | // | ||
594 | // The exception is: if we are at the top-level, for example in an empty match, we | ||
595 | // sometimes prefer reporting the list of constructors instead of just `_`. | ||
596 | let report_when_all_missing = pcx.is_top_level && !IntRange::is_integral(pcx.ty); | ||
597 | let ctor = if !self.matrix_ctors.is_empty() || report_when_all_missing { | ||
598 | Missing | ||
599 | } else { | ||
600 | Wildcard | ||
601 | }; | ||
602 | return smallvec![ctor]; | ||
603 | } | ||
604 | |||
605 | // All the constructors are present in the matrix, so we just go through them all. | ||
606 | self.all_ctors | ||
607 | } | ||
608 | } | ||
609 | |||
610 | /// A value can be decomposed into a constructor applied to some fields. This struct represents | ||
611 | /// those fields, generalized to allow patterns in each field. See also `Constructor`. | ||
612 | /// This is constructed from a constructor using [`Fields::wildcards()`]. | ||
613 | /// | ||
614 | /// If a private or `non_exhaustive` field is uninhabited, the code mustn't observe that it is | ||
615 | /// uninhabited. For that, we filter these fields out of the matrix. This is handled automatically | ||
616 | /// in `Fields`. This filtering is uncommon in practice, because uninhabited fields are rarely used, | ||
617 | /// so we avoid it when possible to preserve performance. | ||
618 | #[derive(Debug, Clone)] | ||
619 | pub(super) enum Fields { | ||
620 | /// Lists of patterns that don't contain any filtered fields. | ||
621 | /// `Slice` and `Vec` behave the same; the difference is only to avoid allocating and | ||
622 | /// triple-dereferences when possible. Frankly this is premature optimization, I (Nadrieril) | ||
623 | /// have not measured if it really made a difference. | ||
624 | Vec(SmallVec<[PatId; 2]>), | ||
625 | } | ||
626 | |||
627 | impl Fields { | ||
628 | /// Internal use. Use `Fields::wildcards()` instead. | ||
629 | /// Must not be used if the pattern is a field of a struct/tuple/variant. | ||
630 | fn from_single_pattern(pat: PatId) -> Self { | ||
631 | Fields::Vec(smallvec![pat]) | ||
632 | } | ||
633 | |||
634 | /// Convenience; internal use. | ||
635 | fn wildcards_from_tys(cx: &MatchCheckCtx<'_>, tys: impl IntoIterator<Item = Ty>) -> Self { | ||
636 | let wilds = tys.into_iter().map(Pat::wildcard_from_ty); | ||
637 | let pats = wilds.map(|pat| cx.alloc_pat(pat)).collect(); | ||
638 | Fields::Vec(pats) | ||
639 | } | ||
640 | |||
641 | /// Creates a new list of wildcard fields for a given constructor. | ||
642 | pub(crate) fn wildcards(pcx: PatCtxt<'_>, constructor: &Constructor) -> Self { | ||
643 | let ty = pcx.ty; | ||
644 | let cx = pcx.cx; | ||
645 | let wildcard_from_ty = |ty: &Ty| cx.alloc_pat(Pat::wildcard_from_ty(ty.clone())); | ||
646 | |||
647 | let ret = match constructor { | ||
648 | Single | Variant(_) => match ty.kind(&Interner) { | ||
649 | TyKind::Tuple(_, substs) => { | ||
650 | let tys = substs.iter(&Interner).map(|ty| ty.assert_ty_ref(&Interner)); | ||
651 | Fields::wildcards_from_tys(cx, tys.cloned()) | ||
652 | } | ||
653 | TyKind::Ref(.., rty) => Fields::from_single_pattern(wildcard_from_ty(rty)), | ||
654 | &TyKind::Adt(AdtId(adt), ref substs) => { | ||
655 | if adt_is_box(adt, cx) { | ||
656 | // Use T as the sub pattern type of Box<T>. | ||
657 | let subst_ty = substs.at(&Interner, 0).assert_ty_ref(&Interner); | ||
658 | Fields::from_single_pattern(wildcard_from_ty(subst_ty)) | ||
659 | } else { | ||
660 | let variant_id = constructor.variant_id_for_adt(adt); | ||
661 | let adt_is_local = | ||
662 | variant_id.module(cx.db.upcast()).krate() == cx.module.krate(); | ||
663 | // Whether we must not match the fields of this variant exhaustively. | ||
664 | let is_non_exhaustive = | ||
665 | is_field_list_non_exhaustive(variant_id, cx) && !adt_is_local; | ||
666 | |||
667 | cov_mark::hit!(match_check_wildcard_expanded_to_substitutions); | ||
668 | let field_ty_data = cx.db.field_types(variant_id); | ||
669 | let field_tys = || { | ||
670 | field_ty_data | ||
671 | .iter() | ||
672 | .map(|(_, binders)| binders.clone().substitute(&Interner, substs)) | ||
673 | }; | ||
674 | |||
675 | // In the following cases, we don't need to filter out any fields. This is | ||
676 | // the vast majority of real cases, since uninhabited fields are uncommon. | ||
677 | let has_no_hidden_fields = (matches!(adt, hir_def::AdtId::EnumId(_)) | ||
678 | && !is_non_exhaustive) | ||
679 | || !field_tys().any(|ty| cx.is_uninhabited(&ty)); | ||
680 | |||
681 | if has_no_hidden_fields { | ||
682 | Fields::wildcards_from_tys(cx, field_tys()) | ||
683 | } else { | ||
684 | //FIXME(iDawer): see MatchCheckCtx::is_uninhabited, has_no_hidden_fields is always true | ||
685 | unimplemented!("exhaustive_patterns feature") | ||
686 | } | ||
687 | } | ||
688 | } | ||
689 | ty_kind => { | ||
690 | cx.bug(&format!("Unexpected type for `Single` constructor: {:?}", ty_kind)) | ||
691 | } | ||
692 | }, | ||
693 | Slice(..) => { | ||
694 | unimplemented!() | ||
695 | } | ||
696 | Str(..) | FloatRange(..) | IntRange(..) | NonExhaustive | Opaque | Missing | ||
697 | | Wildcard => Fields::Vec(Default::default()), | ||
698 | }; | ||
699 | ret | ||
700 | } | ||
701 | |||
702 | /// Apply a constructor to a list of patterns, yielding a new pattern. `self` | ||
703 | /// must have as many elements as this constructor's arity. | ||
704 | /// | ||
705 | /// This is roughly the inverse of `specialize_constructor`. | ||
706 | /// | ||
707 | /// Examples: | ||
708 | /// `ctor`: `Constructor::Single` | ||
709 | /// `ty`: `Foo(u32, u32, u32)` | ||
710 | /// `self`: `[10, 20, _]` | ||
711 | /// returns `Foo(10, 20, _)` | ||
712 | /// | ||
713 | /// `ctor`: `Constructor::Variant(Option::Some)` | ||
714 | /// `ty`: `Option<bool>` | ||
715 | /// `self`: `[false]` | ||
716 | /// returns `Some(false)` | ||
717 | pub(super) fn apply(self, pcx: PatCtxt<'_>, ctor: &Constructor) -> Pat { | ||
718 | let subpatterns_and_indices = self.patterns_and_indices(); | ||
719 | let mut subpatterns = | ||
720 | subpatterns_and_indices.iter().map(|&(_, p)| pcx.cx.pattern_arena.borrow()[p].clone()); | ||
721 | // FIXME(iDawer) witnesses are not yet used | ||
722 | const UNHANDLED: PatKind = PatKind::Wild; | ||
723 | |||
724 | let pat = match ctor { | ||
725 | Single | Variant(_) => match pcx.ty.kind(&Interner) { | ||
726 | TyKind::Adt(..) | TyKind::Tuple(..) => { | ||
727 | // We want the real indices here. | ||
728 | let subpatterns = subpatterns_and_indices | ||
729 | .iter() | ||
730 | .map(|&(field, pat)| FieldPat { | ||
731 | field, | ||
732 | pattern: pcx.cx.pattern_arena.borrow()[pat].clone(), | ||
733 | }) | ||
734 | .collect(); | ||
735 | |||
736 | if let Some((adt, substs)) = pcx.ty.as_adt() { | ||
737 | if let hir_def::AdtId::EnumId(_) = adt { | ||
738 | let enum_variant = match ctor { | ||
739 | &Variant(id) => id, | ||
740 | _ => unreachable!(), | ||
741 | }; | ||
742 | PatKind::Variant { substs: substs.clone(), enum_variant, subpatterns } | ||
743 | } else { | ||
744 | PatKind::Leaf { subpatterns } | ||
745 | } | ||
746 | } else { | ||
747 | PatKind::Leaf { subpatterns } | ||
748 | } | ||
749 | } | ||
750 | // Note: given the expansion of `&str` patterns done in `expand_pattern`, we should | ||
751 | // be careful to reconstruct the correct constant pattern here. However a string | ||
752 | // literal pattern will never be reported as a non-exhaustiveness witness, so we | ||
753 | // can ignore this issue. | ||
754 | TyKind::Ref(..) => PatKind::Deref { subpattern: subpatterns.next().unwrap() }, | ||
755 | TyKind::Slice(..) | TyKind::Array(..) => { | ||
756 | pcx.cx.bug(&format!("bad slice pattern {:?} {:?}", ctor, pcx.ty)) | ||
757 | } | ||
758 | _ => PatKind::Wild, | ||
759 | }, | ||
760 | Constructor::Slice(_) => UNHANDLED, | ||
761 | Str(_) => UNHANDLED, | ||
762 | FloatRange(..) => UNHANDLED, | ||
763 | Constructor::IntRange(_) => UNHANDLED, | ||
764 | NonExhaustive => PatKind::Wild, | ||
765 | Wildcard => return Pat::wildcard_from_ty(pcx.ty.clone()), | ||
766 | Opaque => pcx.cx.bug("we should not try to apply an opaque constructor"), | ||
767 | Missing => pcx.cx.bug( | ||
768 | "trying to apply the `Missing` constructor;\ | ||
769 | this should have been done in `apply_constructors`", | ||
770 | ), | ||
771 | }; | ||
772 | |||
773 | Pat { ty: pcx.ty.clone(), kind: Box::new(pat) } | ||
774 | } | ||
775 | |||
776 | /// Returns the number of patterns. This is the same as the arity of the constructor used to | ||
777 | /// construct `self`. | ||
778 | pub(super) fn len(&self) -> usize { | ||
779 | match self { | ||
780 | Fields::Vec(pats) => pats.len(), | ||
781 | } | ||
782 | } | ||
783 | |||
784 | /// Returns the list of patterns along with the corresponding field indices. | ||
785 | fn patterns_and_indices(&self) -> SmallVec<[(LocalFieldId, PatId); 2]> { | ||
786 | match self { | ||
787 | Fields::Vec(pats) => pats | ||
788 | .iter() | ||
789 | .copied() | ||
790 | .enumerate() | ||
791 | .map(|(i, p)| (LocalFieldId::from_raw((i as u32).into()), p)) | ||
792 | .collect(), | ||
793 | } | ||
794 | } | ||
795 | |||
796 | pub(super) fn into_patterns(self) -> SmallVec<[PatId; 2]> { | ||
797 | match self { | ||
798 | Fields::Vec(pats) => pats, | ||
799 | } | ||
800 | } | ||
801 | |||
802 | /// Overrides some of the fields with the provided patterns. Exactly like | ||
803 | /// `replace_fields_indexed`, except that it takes `FieldPat`s as input. | ||
804 | fn replace_with_fieldpats( | ||
805 | &self, | ||
806 | new_pats: impl IntoIterator<Item = (LocalFieldId, PatId)>, | ||
807 | ) -> Self { | ||
808 | self.replace_fields_indexed( | ||
809 | new_pats.into_iter().map(|(field, pat)| (u32::from(field.into_raw()) as usize, pat)), | ||
810 | ) | ||
811 | } | ||
812 | |||
813 | /// Overrides some of the fields with the provided patterns. This is used when a pattern | ||
814 | /// defines some fields but not all, for example `Foo { field1: Some(_), .. }`: here we start | ||
815 | /// with a `Fields` that is just one wildcard per field of the `Foo` struct, and override the | ||
816 | /// entry corresponding to `field1` with the pattern `Some(_)`. This is also used for slice | ||
817 | /// patterns for the same reason. | ||
818 | fn replace_fields_indexed(&self, new_pats: impl IntoIterator<Item = (usize, PatId)>) -> Self { | ||
819 | let mut fields = self.clone(); | ||
820 | |||
821 | match &mut fields { | ||
822 | Fields::Vec(pats) => { | ||
823 | for (i, pat) in new_pats { | ||
824 | if let Some(p) = pats.get_mut(i) { | ||
825 | *p = pat; | ||
826 | } | ||
827 | } | ||
828 | } | ||
829 | } | ||
830 | fields | ||
831 | } | ||
832 | |||
833 | /// Replaces contained fields with the given list of patterns. There must be `len()` patterns | ||
834 | /// in `pats`. | ||
835 | pub(super) fn replace_fields( | ||
836 | &self, | ||
837 | cx: &MatchCheckCtx<'_>, | ||
838 | pats: impl IntoIterator<Item = Pat>, | ||
839 | ) -> Self { | ||
840 | let pats = pats.into_iter().map(|pat| cx.alloc_pat(pat)).collect(); | ||
841 | |||
842 | match self { | ||
843 | Fields::Vec(_) => Fields::Vec(pats), | ||
844 | } | ||
845 | } | ||
846 | |||
847 | /// Replaces contained fields with the arguments of the given pattern. Only use on a pattern | ||
848 | /// that is compatible with the constructor used to build `self`. | ||
849 | /// This is meant to be used on the result of `Fields::wildcards()`. The idea is that | ||
850 | /// `wildcards` constructs a list of fields where all entries are wildcards, and the pattern | ||
851 | /// provided to this function fills some of the fields with non-wildcards. | ||
852 | /// In the following example `Fields::wildcards` would return `[_, _, _, _]`. If we call | ||
853 | /// `replace_with_pattern_arguments` on it with the pattern, the result will be `[Some(0), _, | ||
854 | /// _, _]`. | ||
855 | /// ```rust | ||
856 | /// let x: [Option<u8>; 4] = foo(); | ||
857 | /// match x { | ||
858 | /// [Some(0), ..] => {} | ||
859 | /// } | ||
860 | /// ``` | ||
861 | /// This is guaranteed to preserve the number of patterns in `self`. | ||
862 | pub(super) fn replace_with_pattern_arguments( | ||
863 | &self, | ||
864 | pat: PatId, | ||
865 | cx: &MatchCheckCtx<'_>, | ||
866 | ) -> Self { | ||
867 | // FIXME(iDawer): Factor out pattern deep cloning. See discussion: | ||
868 | // https://github.com/rust-analyzer/rust-analyzer/pull/8717#discussion_r633086640 | ||
869 | let mut arena = cx.pattern_arena.borrow_mut(); | ||
870 | match arena[pat].kind.as_ref() { | ||
871 | PatKind::Deref { subpattern } => { | ||
872 | assert_eq!(self.len(), 1); | ||
873 | let subpattern = subpattern.clone(); | ||
874 | Fields::from_single_pattern(arena.alloc(subpattern)) | ||
875 | } | ||
876 | PatKind::Leaf { subpatterns } | PatKind::Variant { subpatterns, .. } => { | ||
877 | let subpatterns = subpatterns.clone(); | ||
878 | let subpatterns = subpatterns | ||
879 | .iter() | ||
880 | .map(|field_pat| (field_pat.field, arena.alloc(field_pat.pattern.clone()))); | ||
881 | self.replace_with_fieldpats(subpatterns) | ||
882 | } | ||
883 | |||
884 | PatKind::Wild | ||
885 | | PatKind::Binding { .. } | ||
886 | | PatKind::LiteralBool { .. } | ||
887 | | PatKind::Or { .. } => self.clone(), | ||
888 | } | ||
889 | } | ||
890 | } | ||
891 | |||
892 | fn is_field_list_non_exhaustive(variant_id: VariantId, cx: &MatchCheckCtx<'_>) -> bool { | ||
893 | let attr_def_id = match variant_id { | ||
894 | VariantId::EnumVariantId(id) => id.into(), | ||
895 | VariantId::StructId(id) => id.into(), | ||
896 | VariantId::UnionId(id) => id.into(), | ||
897 | }; | ||
898 | cx.db.attrs(attr_def_id).by_key("non_exhaustive").exists() | ||
899 | } | ||
900 | |||
901 | fn adt_is_box(adt: hir_def::AdtId, cx: &MatchCheckCtx<'_>) -> bool { | ||
902 | use hir_def::lang_item::LangItemTarget; | ||
903 | match cx.db.lang_item(cx.module.krate(), "owned_box".into()) { | ||
904 | Some(LangItemTarget::StructId(box_id)) => adt == box_id.into(), | ||
905 | _ => false, | ||
906 | } | ||
907 | } | ||
diff --git a/crates/hir_ty/src/diagnostics/match_check/pat_util.rs b/crates/hir_ty/src/diagnostics/match_check/pat_util.rs new file mode 100644 index 000000000..b89b4f2bf --- /dev/null +++ b/crates/hir_ty/src/diagnostics/match_check/pat_util.rs | |||
@@ -0,0 +1,56 @@ | |||
1 | //! Pattern untilities. | ||
2 | //! | ||
3 | //! Originates from `rustc_hir::pat_util` | ||
4 | |||
5 | use std::iter::{Enumerate, ExactSizeIterator}; | ||
6 | |||
7 | pub(crate) struct EnumerateAndAdjust<I> { | ||
8 | enumerate: Enumerate<I>, | ||
9 | gap_pos: usize, | ||
10 | gap_len: usize, | ||
11 | } | ||
12 | |||
13 | impl<I> Iterator for EnumerateAndAdjust<I> | ||
14 | where | ||
15 | I: Iterator, | ||
16 | { | ||
17 | type Item = (usize, <I as Iterator>::Item); | ||
18 | |||
19 | fn next(&mut self) -> Option<(usize, <I as Iterator>::Item)> { | ||
20 | self.enumerate | ||
21 | .next() | ||
22 | .map(|(i, elem)| (if i < self.gap_pos { i } else { i + self.gap_len }, elem)) | ||
23 | } | ||
24 | |||
25 | fn size_hint(&self) -> (usize, Option<usize>) { | ||
26 | self.enumerate.size_hint() | ||
27 | } | ||
28 | } | ||
29 | |||
30 | pub(crate) trait EnumerateAndAdjustIterator { | ||
31 | fn enumerate_and_adjust( | ||
32 | self, | ||
33 | expected_len: usize, | ||
34 | gap_pos: Option<usize>, | ||
35 | ) -> EnumerateAndAdjust<Self> | ||
36 | where | ||
37 | Self: Sized; | ||
38 | } | ||
39 | |||
40 | impl<T: ExactSizeIterator> EnumerateAndAdjustIterator for T { | ||
41 | fn enumerate_and_adjust( | ||
42 | self, | ||
43 | expected_len: usize, | ||
44 | gap_pos: Option<usize>, | ||
45 | ) -> EnumerateAndAdjust<Self> | ||
46 | where | ||
47 | Self: Sized, | ||
48 | { | ||
49 | let actual_len = self.len(); | ||
50 | EnumerateAndAdjust { | ||
51 | enumerate: self.enumerate(), | ||
52 | gap_pos: gap_pos.unwrap_or(expected_len), | ||
53 | gap_len: expected_len - actual_len, | ||
54 | } | ||
55 | } | ||
56 | } | ||
diff --git a/crates/hir_ty/src/diagnostics/match_check/usefulness.rs b/crates/hir_ty/src/diagnostics/match_check/usefulness.rs new file mode 100644 index 000000000..83b094a89 --- /dev/null +++ b/crates/hir_ty/src/diagnostics/match_check/usefulness.rs | |||
@@ -0,0 +1,1188 @@ | |||
1 | //! Based on rust-lang/rust 1.52.0-nightly (25c15cdbe 2021-04-22) | ||
2 | //! https://github.com/rust-lang/rust/blob/25c15cdbe/compiler/rustc_mir_build/src/thir/pattern/usefulness.rs | ||
3 | //! | ||
4 | //! ----- | ||
5 | //! | ||
6 | //! This file includes the logic for exhaustiveness and reachability checking for pattern-matching. | ||
7 | //! Specifically, given a list of patterns for a type, we can tell whether: | ||
8 | //! (a) each pattern is reachable (reachability) | ||
9 | //! (b) the patterns cover every possible value for the type (exhaustiveness) | ||
10 | //! | ||
11 | //! The algorithm implemented here is a modified version of the one described in [this | ||
12 | //! paper](http://moscova.inria.fr/~maranget/papers/warn/index.html). We have however generalized | ||
13 | //! it to accommodate the variety of patterns that Rust supports. We thus explain our version here, | ||
14 | //! without being as rigorous. | ||
15 | //! | ||
16 | //! | ||
17 | //! # Summary | ||
18 | //! | ||
19 | //! The core of the algorithm is the notion of "usefulness". A pattern `q` is said to be *useful* | ||
20 | //! relative to another pattern `p` of the same type if there is a value that is matched by `q` and | ||
21 | //! not matched by `p`. This generalizes to many `p`s: `q` is useful w.r.t. a list of patterns | ||
22 | //! `p_1 .. p_n` if there is a value that is matched by `q` and by none of the `p_i`. We write | ||
23 | //! `usefulness(p_1 .. p_n, q)` for a function that returns a list of such values. The aim of this | ||
24 | //! file is to compute it efficiently. | ||
25 | //! | ||
26 | //! This is enough to compute reachability: a pattern in a `match` expression is reachable iff it | ||
27 | //! is useful w.r.t. the patterns above it: | ||
28 | //! ```rust | ||
29 | //! match x { | ||
30 | //! Some(_) => ..., | ||
31 | //! None => ..., // reachable: `None` is matched by this but not the branch above | ||
32 | //! Some(0) => ..., // unreachable: all the values this matches are already matched by | ||
33 | //! // `Some(_)` above | ||
34 | //! } | ||
35 | //! ``` | ||
36 | //! | ||
37 | //! This is also enough to compute exhaustiveness: a match is exhaustive iff the wildcard `_` | ||
38 | //! pattern is _not_ useful w.r.t. the patterns in the match. The values returned by `usefulness` | ||
39 | //! are used to tell the user which values are missing. | ||
40 | //! ```rust | ||
41 | //! match x { | ||
42 | //! Some(0) => ..., | ||
43 | //! None => ..., | ||
44 | //! // not exhaustive: `_` is useful because it matches `Some(1)` | ||
45 | //! } | ||
46 | //! ``` | ||
47 | //! | ||
48 | //! The entrypoint of this file is the [`compute_match_usefulness`] function, which computes | ||
49 | //! reachability for each match branch and exhaustiveness for the whole match. | ||
50 | //! | ||
51 | //! | ||
52 | //! # Constructors and fields | ||
53 | //! | ||
54 | //! Note: we will often abbreviate "constructor" as "ctor". | ||
55 | //! | ||
56 | //! The idea that powers everything that is done in this file is the following: a (matcheable) | ||
57 | //! value is made from a constructor applied to a number of subvalues. Examples of constructors are | ||
58 | //! `Some`, `None`, `(,)` (the 2-tuple constructor), `Foo {..}` (the constructor for a struct | ||
59 | //! `Foo`), and `2` (the constructor for the number `2`). This is natural when we think of | ||
60 | //! pattern-matching, and this is the basis for what follows. | ||
61 | //! | ||
62 | //! Some of the ctors listed above might feel weird: `None` and `2` don't take any arguments. | ||
63 | //! That's ok: those are ctors that take a list of 0 arguments; they are the simplest case of | ||
64 | //! ctors. We treat `2` as a ctor because `u64` and other number types behave exactly like a huge | ||
65 | //! `enum`, with one variant for each number. This allows us to see any matcheable value as made up | ||
66 | //! from a tree of ctors, each having a set number of children. For example: `Foo { bar: None, | ||
67 | //! baz: Ok(0) }` is made from 4 different ctors, namely `Foo{..}`, `None`, `Ok` and `0`. | ||
68 | //! | ||
69 | //! This idea can be extended to patterns: they are also made from constructors applied to fields. | ||
70 | //! A pattern for a given type is allowed to use all the ctors for values of that type (which we | ||
71 | //! call "value constructors"), but there are also pattern-only ctors. The most important one is | ||
72 | //! the wildcard (`_`), and the others are integer ranges (`0..=10`), variable-length slices (`[x, | ||
73 | //! ..]`), and or-patterns (`Ok(0) | Err(_)`). Examples of valid patterns are `42`, `Some(_)`, `Foo | ||
74 | //! { bar: Some(0) | None, baz: _ }`. Note that a binder in a pattern (e.g. `Some(x)`) matches the | ||
75 | //! same values as a wildcard (e.g. `Some(_)`), so we treat both as wildcards. | ||
76 | //! | ||
77 | //! From this deconstruction we can compute whether a given value matches a given pattern; we | ||
78 | //! simply look at ctors one at a time. Given a pattern `p` and a value `v`, we want to compute | ||
79 | //! `matches!(v, p)`. It's mostly straightforward: we compare the head ctors and when they match | ||
80 | //! we compare their fields recursively. A few representative examples: | ||
81 | //! | ||
82 | //! - `matches!(v, _) := true` | ||
83 | //! - `matches!((v0, v1), (p0, p1)) := matches!(v0, p0) && matches!(v1, p1)` | ||
84 | //! - `matches!(Foo { bar: v0, baz: v1 }, Foo { bar: p0, baz: p1 }) := matches!(v0, p0) && matches!(v1, p1)` | ||
85 | //! - `matches!(Ok(v0), Ok(p0)) := matches!(v0, p0)` | ||
86 | //! - `matches!(Ok(v0), Err(p0)) := false` (incompatible variants) | ||
87 | //! - `matches!(v, 1..=100) := matches!(v, 1) || ... || matches!(v, 100)` | ||
88 | //! - `matches!([v0], [p0, .., p1]) := false` (incompatible lengths) | ||
89 | //! - `matches!([v0, v1, v2], [p0, .., p1]) := matches!(v0, p0) && matches!(v2, p1)` | ||
90 | //! - `matches!(v, p0 | p1) := matches!(v, p0) || matches!(v, p1)` | ||
91 | //! | ||
92 | //! Constructors, fields and relevant operations are defined in the [`super::deconstruct_pat`] module. | ||
93 | //! | ||
94 | //! Note: this constructors/fields distinction may not straightforwardly apply to every Rust type. | ||
95 | //! For example a value of type `Rc<u64>` can't be deconstructed that way, and `&str` has an | ||
96 | //! infinitude of constructors. There are also subtleties with visibility of fields and | ||
97 | //! uninhabitedness and various other things. The constructors idea can be extended to handle most | ||
98 | //! of these subtleties though; caveats are documented where relevant throughout the code. | ||
99 | //! | ||
100 | //! Whether constructors cover each other is computed by [`Constructor::is_covered_by`]. | ||
101 | //! | ||
102 | //! | ||
103 | //! # Specialization | ||
104 | //! | ||
105 | //! Recall that we wish to compute `usefulness(p_1 .. p_n, q)`: given a list of patterns `p_1 .. | ||
106 | //! p_n` and a pattern `q`, all of the same type, we want to find a list of values (called | ||
107 | //! "witnesses") that are matched by `q` and by none of the `p_i`. We obviously don't just | ||
108 | //! enumerate all possible values. From the discussion above we see that we can proceed | ||
109 | //! ctor-by-ctor: for each value ctor of the given type, we ask "is there a value that starts with | ||
110 | //! this constructor and matches `q` and none of the `p_i`?". As we saw above, there's a lot we can | ||
111 | //! say from knowing only the first constructor of our candidate value. | ||
112 | //! | ||
113 | //! Let's take the following example: | ||
114 | //! ``` | ||
115 | //! match x { | ||
116 | //! Enum::Variant1(_) => {} // `p1` | ||
117 | //! Enum::Variant2(None, 0) => {} // `p2` | ||
118 | //! Enum::Variant2(Some(_), 0) => {} // `q` | ||
119 | //! } | ||
120 | //! ``` | ||
121 | //! | ||
122 | //! We can easily see that if our candidate value `v` starts with `Variant1` it will not match `q`. | ||
123 | //! If `v = Variant2(v0, v1)` however, whether or not it matches `p2` and `q` will depend on `v0` | ||
124 | //! and `v1`. In fact, such a `v` will be a witness of usefulness of `q` exactly when the tuple | ||
125 | //! `(v0, v1)` is a witness of usefulness of `q'` in the following reduced match: | ||
126 | //! | ||
127 | //! ``` | ||
128 | //! match x { | ||
129 | //! (None, 0) => {} // `p2'` | ||
130 | //! (Some(_), 0) => {} // `q'` | ||
131 | //! } | ||
132 | //! ``` | ||
133 | //! | ||
134 | //! This motivates a new step in computing usefulness, that we call _specialization_. | ||
135 | //! Specialization consist of filtering a list of patterns for those that match a constructor, and | ||
136 | //! then looking into the constructor's fields. This enables usefulness to be computed recursively. | ||
137 | //! | ||
138 | //! Instead of acting on a single pattern in each row, we will consider a list of patterns for each | ||
139 | //! row, and we call such a list a _pattern-stack_. The idea is that we will specialize the | ||
140 | //! leftmost pattern, which amounts to popping the constructor and pushing its fields, which feels | ||
141 | //! like a stack. We note a pattern-stack simply with `[p_1 ... p_n]`. | ||
142 | //! Here's a sequence of specializations of a list of pattern-stacks, to illustrate what's | ||
143 | //! happening: | ||
144 | //! ``` | ||
145 | //! [Enum::Variant1(_)] | ||
146 | //! [Enum::Variant2(None, 0)] | ||
147 | //! [Enum::Variant2(Some(_), 0)] | ||
148 | //! //==>> specialize with `Variant2` | ||
149 | //! [None, 0] | ||
150 | //! [Some(_), 0] | ||
151 | //! //==>> specialize with `Some` | ||
152 | //! [_, 0] | ||
153 | //! //==>> specialize with `true` (say the type was `bool`) | ||
154 | //! [0] | ||
155 | //! //==>> specialize with `0` | ||
156 | //! [] | ||
157 | //! ``` | ||
158 | //! | ||
159 | //! The function `specialize(c, p)` takes a value constructor `c` and a pattern `p`, and returns 0 | ||
160 | //! or more pattern-stacks. If `c` does not match the head constructor of `p`, it returns nothing; | ||
161 | //! otherwise if returns the fields of the constructor. This only returns more than one | ||
162 | //! pattern-stack if `p` has a pattern-only constructor. | ||
163 | //! | ||
164 | //! - Specializing for the wrong constructor returns nothing | ||
165 | //! | ||
166 | //! `specialize(None, Some(p0)) := []` | ||
167 | //! | ||
168 | //! - Specializing for the correct constructor returns a single row with the fields | ||
169 | //! | ||
170 | //! `specialize(Variant1, Variant1(p0, p1, p2)) := [[p0, p1, p2]]` | ||
171 | //! | ||
172 | //! `specialize(Foo{..}, Foo { bar: p0, baz: p1 }) := [[p0, p1]]` | ||
173 | //! | ||
174 | //! - For or-patterns, we specialize each branch and concatenate the results | ||
175 | //! | ||
176 | //! `specialize(c, p0 | p1) := specialize(c, p0) ++ specialize(c, p1)` | ||
177 | //! | ||
178 | //! - We treat the other pattern constructors as if they were a large or-pattern of all the | ||
179 | //! possibilities: | ||
180 | //! | ||
181 | //! `specialize(c, _) := specialize(c, Variant1(_) | Variant2(_, _) | ...)` | ||
182 | //! | ||
183 | //! `specialize(c, 1..=100) := specialize(c, 1 | ... | 100)` | ||
184 | //! | ||
185 | //! `specialize(c, [p0, .., p1]) := specialize(c, [p0, p1] | [p0, _, p1] | [p0, _, _, p1] | ...)` | ||
186 | //! | ||
187 | //! - If `c` is a pattern-only constructor, `specialize` is defined on a case-by-case basis. See | ||
188 | //! the discussion about constructor splitting in [`super::deconstruct_pat`]. | ||
189 | //! | ||
190 | //! | ||
191 | //! We then extend this function to work with pattern-stacks as input, by acting on the first | ||
192 | //! column and keeping the other columns untouched. | ||
193 | //! | ||
194 | //! Specialization for the whole matrix is done in [`Matrix::specialize_constructor`]. Note that | ||
195 | //! or-patterns in the first column are expanded before being stored in the matrix. Specialization | ||
196 | //! for a single patstack is done from a combination of [`Constructor::is_covered_by`] and | ||
197 | //! [`PatStack::pop_head_constructor`]. The internals of how it's done mostly live in the | ||
198 | //! [`Fields`] struct. | ||
199 | //! | ||
200 | //! | ||
201 | //! # Computing usefulness | ||
202 | //! | ||
203 | //! We now have all we need to compute usefulness. The inputs to usefulness are a list of | ||
204 | //! pattern-stacks `p_1 ... p_n` (one per row), and a new pattern_stack `q`. The paper and this | ||
205 | //! file calls the list of patstacks a _matrix_. They must all have the same number of columns and | ||
206 | //! the patterns in a given column must all have the same type. `usefulness` returns a (possibly | ||
207 | //! empty) list of witnesses of usefulness. These witnesses will also be pattern-stacks. | ||
208 | //! | ||
209 | //! - base case: `n_columns == 0`. | ||
210 | //! Since a pattern-stack functions like a tuple of patterns, an empty one functions like the | ||
211 | //! unit type. Thus `q` is useful iff there are no rows above it, i.e. if `n == 0`. | ||
212 | //! | ||
213 | //! - inductive case: `n_columns > 0`. | ||
214 | //! We need a way to list the constructors we want to try. We will be more clever in the next | ||
215 | //! section but for now assume we list all value constructors for the type of the first column. | ||
216 | //! | ||
217 | //! - for each such ctor `c`: | ||
218 | //! | ||
219 | //! - for each `q'` returned by `specialize(c, q)`: | ||
220 | //! | ||
221 | //! - we compute `usefulness(specialize(c, p_1) ... specialize(c, p_n), q')` | ||
222 | //! | ||
223 | //! - for each witness found, we revert specialization by pushing the constructor `c` on top. | ||
224 | //! | ||
225 | //! - We return the concatenation of all the witnesses found, if any. | ||
226 | //! | ||
227 | //! Example: | ||
228 | //! ``` | ||
229 | //! [Some(true)] // p_1 | ||
230 | //! [None] // p_2 | ||
231 | //! [Some(_)] // q | ||
232 | //! //==>> try `None`: `specialize(None, q)` returns nothing | ||
233 | //! //==>> try `Some`: `specialize(Some, q)` returns a single row | ||
234 | //! [true] // p_1' | ||
235 | //! [_] // q' | ||
236 | //! //==>> try `true`: `specialize(true, q')` returns a single row | ||
237 | //! [] // p_1'' | ||
238 | //! [] // q'' | ||
239 | //! //==>> base case; `n != 0` so `q''` is not useful. | ||
240 | //! //==>> go back up a step | ||
241 | //! [true] // p_1' | ||
242 | //! [_] // q' | ||
243 | //! //==>> try `false`: `specialize(false, q')` returns a single row | ||
244 | //! [] // q'' | ||
245 | //! //==>> base case; `n == 0` so `q''` is useful. We return the single witness `[]` | ||
246 | //! witnesses: | ||
247 | //! [] | ||
248 | //! //==>> undo the specialization with `false` | ||
249 | //! witnesses: | ||
250 | //! [false] | ||
251 | //! //==>> undo the specialization with `Some` | ||
252 | //! witnesses: | ||
253 | //! [Some(false)] | ||
254 | //! //==>> we have tried all the constructors. The output is the single witness `[Some(false)]`. | ||
255 | //! ``` | ||
256 | //! | ||
257 | //! This computation is done in [`is_useful`]. In practice we don't care about the list of | ||
258 | //! witnesses when computing reachability; we only need to know whether any exist. We do keep the | ||
259 | //! witnesses when computing exhaustiveness to report them to the user. | ||
260 | //! | ||
261 | //! | ||
262 | //! # Making usefulness tractable: constructor splitting | ||
263 | //! | ||
264 | //! We're missing one last detail: which constructors do we list? Naively listing all value | ||
265 | //! constructors cannot work for types like `u64` or `&str`, so we need to be more clever. The | ||
266 | //! first obvious insight is that we only want to list constructors that are covered by the head | ||
267 | //! constructor of `q`. If it's a value constructor, we only try that one. If it's a pattern-only | ||
268 | //! constructor, we use the final clever idea for this algorithm: _constructor splitting_, where we | ||
269 | //! group together constructors that behave the same. | ||
270 | //! | ||
271 | //! The details are not necessary to understand this file, so we explain them in | ||
272 | //! [`super::deconstruct_pat`]. Splitting is done by the [`Constructor::split`] function. | ||
273 | |||
274 | use std::{cell::RefCell, iter::FromIterator}; | ||
275 | |||
276 | use hir_def::{expr::ExprId, HasModule, ModuleId}; | ||
277 | use la_arena::Arena; | ||
278 | use once_cell::unsync::OnceCell; | ||
279 | use rustc_hash::FxHashMap; | ||
280 | use smallvec::{smallvec, SmallVec}; | ||
281 | |||
282 | use crate::{db::HirDatabase, InferenceResult, Interner, Ty}; | ||
283 | |||
284 | use super::{ | ||
285 | deconstruct_pat::{Constructor, Fields, SplitWildcard}, | ||
286 | Pat, PatId, PatKind, PatternFoldable, PatternFolder, | ||
287 | }; | ||
288 | |||
289 | use self::{helper::PatIdExt, Usefulness::*, WitnessPreference::*}; | ||
290 | |||
291 | pub(crate) struct MatchCheckCtx<'a> { | ||
292 | pub(crate) module: ModuleId, | ||
293 | pub(crate) match_expr: ExprId, | ||
294 | pub(crate) infer: &'a InferenceResult, | ||
295 | pub(crate) db: &'a dyn HirDatabase, | ||
296 | /// Lowered patterns from arms plus generated by the check. | ||
297 | pub(crate) pattern_arena: &'a RefCell<PatternArena>, | ||
298 | pub(crate) eprint_panic_context: &'a dyn Fn(), | ||
299 | } | ||
300 | |||
301 | impl<'a> MatchCheckCtx<'a> { | ||
302 | pub(super) fn is_uninhabited(&self, _ty: &Ty) -> bool { | ||
303 | // FIXME(iDawer) implement exhaustive_patterns feature. More info in: | ||
304 | // Tracking issue for RFC 1872: exhaustive_patterns feature https://github.com/rust-lang/rust/issues/51085 | ||
305 | false | ||
306 | } | ||
307 | |||
308 | /// Returns whether the given type is an enum from another crate declared `#[non_exhaustive]`. | ||
309 | pub(super) fn is_foreign_non_exhaustive_enum(&self, enum_id: hir_def::EnumId) -> bool { | ||
310 | let has_non_exhaustive_attr = | ||
311 | self.db.attrs(enum_id.into()).by_key("non_exhaustive").exists(); | ||
312 | let is_local = | ||
313 | hir_def::AdtId::from(enum_id).module(self.db.upcast()).krate() == self.module.krate(); | ||
314 | has_non_exhaustive_attr && !is_local | ||
315 | } | ||
316 | |||
317 | // Rust feature described as "Allows exhaustive pattern matching on types that contain uninhabited types." | ||
318 | pub(super) fn feature_exhaustive_patterns(&self) -> bool { | ||
319 | // FIXME see MatchCheckCtx::is_uninhabited | ||
320 | false | ||
321 | } | ||
322 | |||
323 | pub(super) fn alloc_pat(&self, pat: Pat) -> PatId { | ||
324 | self.pattern_arena.borrow_mut().alloc(pat) | ||
325 | } | ||
326 | |||
327 | /// Get type of a pattern. Handles expanded patterns. | ||
328 | pub(super) fn type_of(&self, pat: PatId) -> Ty { | ||
329 | self.pattern_arena.borrow()[pat].ty.clone() | ||
330 | } | ||
331 | |||
332 | #[track_caller] | ||
333 | pub(super) fn bug(&self, info: &str) -> ! { | ||
334 | (self.eprint_panic_context)(); | ||
335 | panic!("bug: {}", info); | ||
336 | } | ||
337 | } | ||
338 | |||
339 | #[derive(Copy, Clone)] | ||
340 | pub(super) struct PatCtxt<'a> { | ||
341 | pub(super) cx: &'a MatchCheckCtx<'a>, | ||
342 | /// Type of the current column under investigation. | ||
343 | pub(super) ty: &'a Ty, | ||
344 | /// Whether the current pattern is the whole pattern as found in a match arm, or if it's a | ||
345 | /// subpattern. | ||
346 | pub(super) is_top_level: bool, | ||
347 | } | ||
348 | |||
349 | pub(crate) fn expand_pattern(pat: Pat) -> Pat { | ||
350 | LiteralExpander.fold_pattern(&pat) | ||
351 | } | ||
352 | |||
353 | struct LiteralExpander; | ||
354 | |||
355 | impl PatternFolder for LiteralExpander { | ||
356 | fn fold_pattern(&mut self, pat: &Pat) -> Pat { | ||
357 | match (pat.ty.kind(&Interner), pat.kind.as_ref()) { | ||
358 | (_, PatKind::Binding { subpattern: Some(s), .. }) => s.fold_with(self), | ||
359 | _ => pat.super_fold_with(self), | ||
360 | } | ||
361 | } | ||
362 | } | ||
363 | |||
364 | impl Pat { | ||
365 | fn _is_wildcard(&self) -> bool { | ||
366 | matches!(*self.kind, PatKind::Binding { subpattern: None, .. } | PatKind::Wild) | ||
367 | } | ||
368 | } | ||
369 | |||
370 | impl PatIdExt for PatId { | ||
371 | fn is_or_pat(self, cx: &MatchCheckCtx<'_>) -> bool { | ||
372 | matches!(*cx.pattern_arena.borrow()[self].kind, PatKind::Or { .. }) | ||
373 | } | ||
374 | |||
375 | /// Recursively expand this pattern into its subpatterns. Only useful for or-patterns. | ||
376 | fn expand_or_pat(self, cx: &MatchCheckCtx<'_>) -> Vec<Self> { | ||
377 | fn expand(pat: PatId, vec: &mut Vec<PatId>, pat_arena: &mut PatternArena) { | ||
378 | if let PatKind::Or { pats } = pat_arena[pat].kind.as_ref() { | ||
379 | // FIXME(iDawer): Factor out pattern deep cloning. See discussion: | ||
380 | // https://github.com/rust-analyzer/rust-analyzer/pull/8717#discussion_r633086640 | ||
381 | let pats = pats.clone(); | ||
382 | for pat in pats { | ||
383 | let pat = pat_arena.alloc(pat.clone()); | ||
384 | expand(pat, vec, pat_arena); | ||
385 | } | ||
386 | } else { | ||
387 | vec.push(pat) | ||
388 | } | ||
389 | } | ||
390 | |||
391 | let mut pat_arena = cx.pattern_arena.borrow_mut(); | ||
392 | let mut pats = Vec::new(); | ||
393 | expand(self, &mut pats, &mut pat_arena); | ||
394 | pats | ||
395 | } | ||
396 | } | ||
397 | |||
398 | /// A row of a matrix. Rows of len 1 are very common, which is why `SmallVec[_; 2]` | ||
399 | /// works well. | ||
400 | #[derive(Clone)] | ||
401 | pub(super) struct PatStack { | ||
402 | pats: SmallVec<[PatId; 2]>, | ||
403 | /// Cache for the constructor of the head | ||
404 | head_ctor: OnceCell<Constructor>, | ||
405 | } | ||
406 | |||
407 | impl PatStack { | ||
408 | fn from_pattern(pat: PatId) -> Self { | ||
409 | Self::from_vec(smallvec![pat]) | ||
410 | } | ||
411 | |||
412 | fn from_vec(vec: SmallVec<[PatId; 2]>) -> Self { | ||
413 | PatStack { pats: vec, head_ctor: OnceCell::new() } | ||
414 | } | ||
415 | |||
416 | fn is_empty(&self) -> bool { | ||
417 | self.pats.is_empty() | ||
418 | } | ||
419 | |||
420 | fn len(&self) -> usize { | ||
421 | self.pats.len() | ||
422 | } | ||
423 | |||
424 | fn head(&self) -> PatId { | ||
425 | self.pats[0] | ||
426 | } | ||
427 | |||
428 | #[inline] | ||
429 | fn head_ctor(&self, cx: &MatchCheckCtx<'_>) -> &Constructor { | ||
430 | self.head_ctor.get_or_init(|| Constructor::from_pat(cx, self.head())) | ||
431 | } | ||
432 | |||
433 | // Recursively expand the first pattern into its subpatterns. Only useful if the pattern is an | ||
434 | // or-pattern. Panics if `self` is empty. | ||
435 | fn expand_or_pat(&self, cx: &MatchCheckCtx<'_>) -> impl Iterator<Item = PatStack> + '_ { | ||
436 | self.head().expand_or_pat(cx).into_iter().map(move |pat| { | ||
437 | let mut new_patstack = PatStack::from_pattern(pat); | ||
438 | new_patstack.pats.extend_from_slice(&self.pats[1..]); | ||
439 | new_patstack | ||
440 | }) | ||
441 | } | ||
442 | |||
443 | /// This computes `S(self.head_ctor(), self)`. See top of the file for explanations. | ||
444 | /// | ||
445 | /// Structure patterns with a partial wild pattern (Foo { a: 42, .. }) have their missing | ||
446 | /// fields filled with wild patterns. | ||
447 | /// | ||
448 | /// This is roughly the inverse of `Constructor::apply`. | ||
449 | fn pop_head_constructor( | ||
450 | &self, | ||
451 | ctor_wild_subpatterns: &Fields, | ||
452 | cx: &MatchCheckCtx<'_>, | ||
453 | ) -> PatStack { | ||
454 | // We pop the head pattern and push the new fields extracted from the arguments of | ||
455 | // `self.head()`. | ||
456 | let mut new_fields = | ||
457 | ctor_wild_subpatterns.replace_with_pattern_arguments(self.head(), cx).into_patterns(); | ||
458 | new_fields.extend_from_slice(&self.pats[1..]); | ||
459 | PatStack::from_vec(new_fields) | ||
460 | } | ||
461 | } | ||
462 | |||
463 | impl Default for PatStack { | ||
464 | fn default() -> Self { | ||
465 | Self::from_vec(smallvec![]) | ||
466 | } | ||
467 | } | ||
468 | |||
469 | impl PartialEq for PatStack { | ||
470 | fn eq(&self, other: &Self) -> bool { | ||
471 | self.pats == other.pats | ||
472 | } | ||
473 | } | ||
474 | |||
475 | impl FromIterator<PatId> for PatStack { | ||
476 | fn from_iter<T>(iter: T) -> Self | ||
477 | where | ||
478 | T: IntoIterator<Item = PatId>, | ||
479 | { | ||
480 | Self::from_vec(iter.into_iter().collect()) | ||
481 | } | ||
482 | } | ||
483 | |||
484 | /// A 2D matrix. | ||
485 | #[derive(Clone)] | ||
486 | pub(super) struct Matrix { | ||
487 | patterns: Vec<PatStack>, | ||
488 | } | ||
489 | |||
490 | impl Matrix { | ||
491 | fn empty() -> Self { | ||
492 | Matrix { patterns: vec![] } | ||
493 | } | ||
494 | |||
495 | /// Number of columns of this matrix. `None` is the matrix is empty. | ||
496 | pub(super) fn _column_count(&self) -> Option<usize> { | ||
497 | self.patterns.get(0).map(|r| r.len()) | ||
498 | } | ||
499 | |||
500 | /// Pushes a new row to the matrix. If the row starts with an or-pattern, this recursively | ||
501 | /// expands it. | ||
502 | fn push(&mut self, row: PatStack, cx: &MatchCheckCtx<'_>) { | ||
503 | if !row.is_empty() && row.head().is_or_pat(cx) { | ||
504 | for row in row.expand_or_pat(cx) { | ||
505 | self.patterns.push(row); | ||
506 | } | ||
507 | } else { | ||
508 | self.patterns.push(row); | ||
509 | } | ||
510 | } | ||
511 | |||
512 | /// Iterate over the first component of each row | ||
513 | fn heads(&self) -> impl Iterator<Item = PatId> + '_ { | ||
514 | self.patterns.iter().map(|r| r.head()) | ||
515 | } | ||
516 | |||
517 | /// Iterate over the first constructor of each row. | ||
518 | fn head_ctors<'a>( | ||
519 | &'a self, | ||
520 | cx: &'a MatchCheckCtx<'_>, | ||
521 | ) -> impl Iterator<Item = &'a Constructor> + Clone { | ||
522 | self.patterns.iter().map(move |r| r.head_ctor(cx)) | ||
523 | } | ||
524 | |||
525 | /// This computes `S(constructor, self)`. See top of the file for explanations. | ||
526 | fn specialize_constructor( | ||
527 | &self, | ||
528 | pcx: PatCtxt<'_>, | ||
529 | ctor: &Constructor, | ||
530 | ctor_wild_subpatterns: &Fields, | ||
531 | ) -> Matrix { | ||
532 | let rows = self | ||
533 | .patterns | ||
534 | .iter() | ||
535 | .filter(|r| ctor.is_covered_by(pcx, r.head_ctor(pcx.cx))) | ||
536 | .map(|r| r.pop_head_constructor(ctor_wild_subpatterns, pcx.cx)); | ||
537 | Matrix::from_iter(rows, pcx.cx) | ||
538 | } | ||
539 | |||
540 | fn from_iter(rows: impl IntoIterator<Item = PatStack>, cx: &MatchCheckCtx<'_>) -> Matrix { | ||
541 | let mut matrix = Matrix::empty(); | ||
542 | for x in rows { | ||
543 | // Using `push` ensures we correctly expand or-patterns. | ||
544 | matrix.push(x, cx); | ||
545 | } | ||
546 | matrix | ||
547 | } | ||
548 | } | ||
549 | |||
550 | /// Given a pattern or a pattern-stack, this struct captures a set of its subpatterns. We use that | ||
551 | /// to track reachable sub-patterns arising from or-patterns. In the absence of or-patterns this | ||
552 | /// will always be either `Empty` (the whole pattern is unreachable) or `Full` (the whole pattern | ||
553 | /// is reachable). When there are or-patterns, some subpatterns may be reachable while others | ||
554 | /// aren't. In this case the whole pattern still counts as reachable, but we will lint the | ||
555 | /// unreachable subpatterns. | ||
556 | /// | ||
557 | /// This supports a limited set of operations, so not all possible sets of subpatterns can be | ||
558 | /// represented. That's ok, we only want the ones that make sense for our usage. | ||
559 | /// | ||
560 | /// What we're doing is illustrated by this: | ||
561 | /// ``` | ||
562 | /// match (true, 0) { | ||
563 | /// (true, 0) => {} | ||
564 | /// (_, 1) => {} | ||
565 | /// (true | false, 0 | 1) => {} | ||
566 | /// } | ||
567 | /// ``` | ||
568 | /// When we try the alternatives of the `true | false` or-pattern, the last `0` is reachable in the | ||
569 | /// `false` alternative but not the `true`. So overall it is reachable. By contrast, the last `1` | ||
570 | /// is not reachable in either alternative, so we want to signal this to the user. | ||
571 | /// Therefore we take the union of sets of reachable patterns coming from different alternatives in | ||
572 | /// order to figure out which subpatterns are overall reachable. | ||
573 | /// | ||
574 | /// Invariant: we try to construct the smallest representation we can. In particular if | ||
575 | /// `self.is_empty()` we ensure that `self` is `Empty`, and same with `Full`. This is not important | ||
576 | /// for correctness currently. | ||
577 | #[derive(Debug, Clone)] | ||
578 | enum SubPatSet { | ||
579 | /// The empty set. This means the pattern is unreachable. | ||
580 | Empty, | ||
581 | /// The set containing the full pattern. | ||
582 | Full, | ||
583 | /// If the pattern is a pattern with a constructor or a pattern-stack, we store a set for each | ||
584 | /// of its subpatterns. Missing entries in the map are implicitly full, because that's the | ||
585 | /// common case. | ||
586 | Seq { subpats: FxHashMap<usize, SubPatSet> }, | ||
587 | /// If the pattern is an or-pattern, we store a set for each of its alternatives. Missing | ||
588 | /// entries in the map are implicitly empty. Note: we always flatten nested or-patterns. | ||
589 | Alt { | ||
590 | subpats: FxHashMap<usize, SubPatSet>, | ||
591 | /// Counts the total number of alternatives in the pattern | ||
592 | alt_count: usize, | ||
593 | /// We keep the pattern around to retrieve spans. | ||
594 | pat: PatId, | ||
595 | }, | ||
596 | } | ||
597 | |||
598 | impl SubPatSet { | ||
599 | fn full() -> Self { | ||
600 | SubPatSet::Full | ||
601 | } | ||
602 | |||
603 | fn empty() -> Self { | ||
604 | SubPatSet::Empty | ||
605 | } | ||
606 | |||
607 | fn is_empty(&self) -> bool { | ||
608 | match self { | ||
609 | SubPatSet::Empty => true, | ||
610 | SubPatSet::Full => false, | ||
611 | // If any subpattern in a sequence is unreachable, the whole pattern is unreachable. | ||
612 | SubPatSet::Seq { subpats } => subpats.values().any(|set| set.is_empty()), | ||
613 | // An or-pattern is reachable if any of its alternatives is. | ||
614 | SubPatSet::Alt { subpats, .. } => subpats.values().all(|set| set.is_empty()), | ||
615 | } | ||
616 | } | ||
617 | |||
618 | fn is_full(&self) -> bool { | ||
619 | match self { | ||
620 | SubPatSet::Empty => false, | ||
621 | SubPatSet::Full => true, | ||
622 | // The whole pattern is reachable only when all its alternatives are. | ||
623 | SubPatSet::Seq { subpats } => subpats.values().all(|sub_set| sub_set.is_full()), | ||
624 | // The whole or-pattern is reachable only when all its alternatives are. | ||
625 | SubPatSet::Alt { subpats, alt_count, .. } => { | ||
626 | subpats.len() == *alt_count && subpats.values().all(|set| set.is_full()) | ||
627 | } | ||
628 | } | ||
629 | } | ||
630 | |||
631 | /// Union `self` with `other`, mutating `self`. | ||
632 | fn union(&mut self, other: Self) { | ||
633 | use SubPatSet::*; | ||
634 | // Union with full stays full; union with empty changes nothing. | ||
635 | if self.is_full() || other.is_empty() { | ||
636 | return; | ||
637 | } else if self.is_empty() { | ||
638 | *self = other; | ||
639 | return; | ||
640 | } else if other.is_full() { | ||
641 | *self = Full; | ||
642 | return; | ||
643 | } | ||
644 | |||
645 | match (&mut *self, other) { | ||
646 | (Seq { subpats: s_set }, Seq { subpats: mut o_set }) => { | ||
647 | s_set.retain(|i, s_sub_set| { | ||
648 | // Missing entries count as full. | ||
649 | let o_sub_set = o_set.remove(&i).unwrap_or(Full); | ||
650 | s_sub_set.union(o_sub_set); | ||
651 | // We drop full entries. | ||
652 | !s_sub_set.is_full() | ||
653 | }); | ||
654 | // Everything left in `o_set` is missing from `s_set`, i.e. counts as full. Since | ||
655 | // unioning with full returns full, we can drop those entries. | ||
656 | } | ||
657 | (Alt { subpats: s_set, .. }, Alt { subpats: mut o_set, .. }) => { | ||
658 | s_set.retain(|i, s_sub_set| { | ||
659 | // Missing entries count as empty. | ||
660 | let o_sub_set = o_set.remove(&i).unwrap_or(Empty); | ||
661 | s_sub_set.union(o_sub_set); | ||
662 | // We drop empty entries. | ||
663 | !s_sub_set.is_empty() | ||
664 | }); | ||
665 | // Everything left in `o_set` is missing from `s_set`, i.e. counts as empty. Since | ||
666 | // unioning with empty changes nothing, we can take those entries as is. | ||
667 | s_set.extend(o_set); | ||
668 | } | ||
669 | _ => panic!("bug"), | ||
670 | } | ||
671 | |||
672 | if self.is_full() { | ||
673 | *self = Full; | ||
674 | } | ||
675 | } | ||
676 | |||
677 | /// Returns a list of the unreachable subpatterns. If `self` is empty (i.e. the | ||
678 | /// whole pattern is unreachable) we return `None`. | ||
679 | fn list_unreachable_subpatterns(&self, cx: &MatchCheckCtx<'_>) -> Option<Vec<PatId>> { | ||
680 | /// Panics if `set.is_empty()`. | ||
681 | fn fill_subpats( | ||
682 | set: &SubPatSet, | ||
683 | unreachable_pats: &mut Vec<PatId>, | ||
684 | cx: &MatchCheckCtx<'_>, | ||
685 | ) { | ||
686 | match set { | ||
687 | SubPatSet::Empty => panic!("bug"), | ||
688 | SubPatSet::Full => {} | ||
689 | SubPatSet::Seq { subpats } => { | ||
690 | for (_, sub_set) in subpats { | ||
691 | fill_subpats(sub_set, unreachable_pats, cx); | ||
692 | } | ||
693 | } | ||
694 | SubPatSet::Alt { subpats, pat, alt_count, .. } => { | ||
695 | let expanded = pat.expand_or_pat(cx); | ||
696 | for i in 0..*alt_count { | ||
697 | let sub_set = subpats.get(&i).unwrap_or(&SubPatSet::Empty); | ||
698 | if sub_set.is_empty() { | ||
699 | // Found a unreachable subpattern. | ||
700 | unreachable_pats.push(expanded[i]); | ||
701 | } else { | ||
702 | fill_subpats(sub_set, unreachable_pats, cx); | ||
703 | } | ||
704 | } | ||
705 | } | ||
706 | } | ||
707 | } | ||
708 | |||
709 | if self.is_empty() { | ||
710 | return None; | ||
711 | } | ||
712 | if self.is_full() { | ||
713 | // No subpatterns are unreachable. | ||
714 | return Some(Vec::new()); | ||
715 | } | ||
716 | let mut unreachable_pats = Vec::new(); | ||
717 | fill_subpats(self, &mut unreachable_pats, cx); | ||
718 | Some(unreachable_pats) | ||
719 | } | ||
720 | |||
721 | /// When `self` refers to a patstack that was obtained from specialization, after running | ||
722 | /// `unspecialize` it will refer to the original patstack before specialization. | ||
723 | fn unspecialize(self, arity: usize) -> Self { | ||
724 | use SubPatSet::*; | ||
725 | match self { | ||
726 | Full => Full, | ||
727 | Empty => Empty, | ||
728 | Seq { subpats } => { | ||
729 | // We gather the first `arity` subpatterns together and shift the remaining ones. | ||
730 | let mut new_subpats = FxHashMap::default(); | ||
731 | let mut new_subpats_first_col = FxHashMap::default(); | ||
732 | for (i, sub_set) in subpats { | ||
733 | if i < arity { | ||
734 | // The first `arity` indices are now part of the pattern in the first | ||
735 | // column. | ||
736 | new_subpats_first_col.insert(i, sub_set); | ||
737 | } else { | ||
738 | // Indices after `arity` are simply shifted | ||
739 | new_subpats.insert(i - arity + 1, sub_set); | ||
740 | } | ||
741 | } | ||
742 | // If `new_subpats_first_col` has no entries it counts as full, so we can omit it. | ||
743 | if !new_subpats_first_col.is_empty() { | ||
744 | new_subpats.insert(0, Seq { subpats: new_subpats_first_col }); | ||
745 | } | ||
746 | Seq { subpats: new_subpats } | ||
747 | } | ||
748 | Alt { .. } => panic!("bug"), // `self` is a patstack | ||
749 | } | ||
750 | } | ||
751 | |||
752 | /// When `self` refers to a patstack that was obtained from splitting an or-pattern, after | ||
753 | /// running `unspecialize` it will refer to the original patstack before splitting. | ||
754 | /// | ||
755 | /// For example: | ||
756 | /// ``` | ||
757 | /// match Some(true) { | ||
758 | /// Some(true) => {} | ||
759 | /// None | Some(true | false) => {} | ||
760 | /// } | ||
761 | /// ``` | ||
762 | /// Here `None` would return the full set and `Some(true | false)` would return the set | ||
763 | /// containing `false`. After `unsplit_or_pat`, we want the set to contain `None` and `false`. | ||
764 | /// This is what this function does. | ||
765 | fn unsplit_or_pat(mut self, alt_id: usize, alt_count: usize, pat: PatId) -> Self { | ||
766 | use SubPatSet::*; | ||
767 | if self.is_empty() { | ||
768 | return Empty; | ||
769 | } | ||
770 | |||
771 | // Subpatterns coming from inside the or-pattern alternative itself, e.g. in `None | Some(0 | ||
772 | // | 1)`. | ||
773 | let set_first_col = match &mut self { | ||
774 | Full => Full, | ||
775 | Seq { subpats } => subpats.remove(&0).unwrap_or(Full), | ||
776 | Empty => unreachable!(), | ||
777 | Alt { .. } => panic!("bug"), // `self` is a patstack | ||
778 | }; | ||
779 | let mut subpats_first_col = FxHashMap::default(); | ||
780 | subpats_first_col.insert(alt_id, set_first_col); | ||
781 | let set_first_col = Alt { subpats: subpats_first_col, pat, alt_count }; | ||
782 | |||
783 | let mut subpats = match self { | ||
784 | Full => FxHashMap::default(), | ||
785 | Seq { subpats } => subpats, | ||
786 | Empty => unreachable!(), | ||
787 | Alt { .. } => panic!("bug"), // `self` is a patstack | ||
788 | }; | ||
789 | subpats.insert(0, set_first_col); | ||
790 | Seq { subpats } | ||
791 | } | ||
792 | } | ||
793 | |||
794 | /// This carries the results of computing usefulness, as described at the top of the file. When | ||
795 | /// checking usefulness of a match branch, we use the `NoWitnesses` variant, which also keeps track | ||
796 | /// of potential unreachable sub-patterns (in the presence of or-patterns). When checking | ||
797 | /// exhaustiveness of a whole match, we use the `WithWitnesses` variant, which carries a list of | ||
798 | /// witnesses of non-exhaustiveness when there are any. | ||
799 | /// Which variant to use is dictated by `WitnessPreference`. | ||
800 | #[derive(Clone, Debug)] | ||
801 | enum Usefulness { | ||
802 | /// Carries a set of subpatterns that have been found to be reachable. If empty, this indicates | ||
803 | /// the whole pattern is unreachable. If not, this indicates that the pattern is reachable but | ||
804 | /// that some sub-patterns may be unreachable (due to or-patterns). In the absence of | ||
805 | /// or-patterns this will always be either `Empty` (the whole pattern is unreachable) or `Full` | ||
806 | /// (the whole pattern is reachable). | ||
807 | NoWitnesses(SubPatSet), | ||
808 | /// Carries a list of witnesses of non-exhaustiveness. If empty, indicates that the whole | ||
809 | /// pattern is unreachable. | ||
810 | WithWitnesses(Vec<Witness>), | ||
811 | } | ||
812 | |||
813 | impl Usefulness { | ||
814 | fn new_useful(preference: WitnessPreference) -> Self { | ||
815 | match preference { | ||
816 | ConstructWitness => WithWitnesses(vec![Witness(vec![])]), | ||
817 | LeaveOutWitness => NoWitnesses(SubPatSet::full()), | ||
818 | } | ||
819 | } | ||
820 | fn new_not_useful(preference: WitnessPreference) -> Self { | ||
821 | match preference { | ||
822 | ConstructWitness => WithWitnesses(vec![]), | ||
823 | LeaveOutWitness => NoWitnesses(SubPatSet::empty()), | ||
824 | } | ||
825 | } | ||
826 | |||
827 | /// Combine usefulnesses from two branches. This is an associative operation. | ||
828 | fn extend(&mut self, other: Self) { | ||
829 | match (&mut *self, other) { | ||
830 | (WithWitnesses(_), WithWitnesses(o)) if o.is_empty() => {} | ||
831 | (WithWitnesses(s), WithWitnesses(o)) if s.is_empty() => *self = WithWitnesses(o), | ||
832 | (WithWitnesses(s), WithWitnesses(o)) => s.extend(o), | ||
833 | (NoWitnesses(s), NoWitnesses(o)) => s.union(o), | ||
834 | _ => unreachable!(), | ||
835 | } | ||
836 | } | ||
837 | |||
838 | /// When trying several branches and each returns a `Usefulness`, we need to combine the | ||
839 | /// results together. | ||
840 | fn merge(pref: WitnessPreference, usefulnesses: impl Iterator<Item = Self>) -> Self { | ||
841 | let mut ret = Self::new_not_useful(pref); | ||
842 | for u in usefulnesses { | ||
843 | ret.extend(u); | ||
844 | if let NoWitnesses(subpats) = &ret { | ||
845 | if subpats.is_full() { | ||
846 | // Once we reach the full set, more unions won't change the result. | ||
847 | return ret; | ||
848 | } | ||
849 | } | ||
850 | } | ||
851 | ret | ||
852 | } | ||
853 | |||
854 | /// After calculating the usefulness for a branch of an or-pattern, call this to make this | ||
855 | /// usefulness mergeable with those from the other branches. | ||
856 | fn unsplit_or_pat(self, alt_id: usize, alt_count: usize, pat: PatId) -> Self { | ||
857 | match self { | ||
858 | NoWitnesses(subpats) => NoWitnesses(subpats.unsplit_or_pat(alt_id, alt_count, pat)), | ||
859 | WithWitnesses(_) => panic!("bug"), | ||
860 | } | ||
861 | } | ||
862 | |||
863 | /// After calculating usefulness after a specialization, call this to recontruct a usefulness | ||
864 | /// that makes sense for the matrix pre-specialization. This new usefulness can then be merged | ||
865 | /// with the results of specializing with the other constructors. | ||
866 | fn apply_constructor( | ||
867 | self, | ||
868 | pcx: PatCtxt<'_>, | ||
869 | matrix: &Matrix, | ||
870 | ctor: &Constructor, | ||
871 | ctor_wild_subpatterns: &Fields, | ||
872 | ) -> Self { | ||
873 | match self { | ||
874 | WithWitnesses(witnesses) if witnesses.is_empty() => WithWitnesses(witnesses), | ||
875 | WithWitnesses(witnesses) => { | ||
876 | let new_witnesses = if matches!(ctor, Constructor::Missing) { | ||
877 | let mut split_wildcard = SplitWildcard::new(pcx); | ||
878 | split_wildcard.split(pcx, matrix.head_ctors(pcx.cx)); | ||
879 | // Construct for each missing constructor a "wild" version of this | ||
880 | // constructor, that matches everything that can be built with | ||
881 | // it. For example, if `ctor` is a `Constructor::Variant` for | ||
882 | // `Option::Some`, we get the pattern `Some(_)`. | ||
883 | let new_patterns: Vec<_> = split_wildcard | ||
884 | .iter_missing(pcx) | ||
885 | .map(|missing_ctor| { | ||
886 | Fields::wildcards(pcx, missing_ctor).apply(pcx, missing_ctor) | ||
887 | }) | ||
888 | .collect(); | ||
889 | witnesses | ||
890 | .into_iter() | ||
891 | .flat_map(|witness| { | ||
892 | new_patterns.iter().map(move |pat| { | ||
893 | let mut witness = witness.clone(); | ||
894 | witness.0.push(pat.clone()); | ||
895 | witness | ||
896 | }) | ||
897 | }) | ||
898 | .collect() | ||
899 | } else { | ||
900 | witnesses | ||
901 | .into_iter() | ||
902 | .map(|witness| witness.apply_constructor(pcx, &ctor, ctor_wild_subpatterns)) | ||
903 | .collect() | ||
904 | }; | ||
905 | WithWitnesses(new_witnesses) | ||
906 | } | ||
907 | NoWitnesses(subpats) => NoWitnesses(subpats.unspecialize(ctor_wild_subpatterns.len())), | ||
908 | } | ||
909 | } | ||
910 | } | ||
911 | |||
912 | #[derive(Copy, Clone, Debug)] | ||
913 | enum WitnessPreference { | ||
914 | ConstructWitness, | ||
915 | LeaveOutWitness, | ||
916 | } | ||
917 | |||
918 | /// A witness of non-exhaustiveness for error reporting, represented | ||
919 | /// as a list of patterns (in reverse order of construction) with | ||
920 | /// wildcards inside to represent elements that can take any inhabitant | ||
921 | /// of the type as a value. | ||
922 | /// | ||
923 | /// A witness against a list of patterns should have the same types | ||
924 | /// and length as the pattern matched against. Because Rust `match` | ||
925 | /// is always against a single pattern, at the end the witness will | ||
926 | /// have length 1, but in the middle of the algorithm, it can contain | ||
927 | /// multiple patterns. | ||
928 | /// | ||
929 | /// For example, if we are constructing a witness for the match against | ||
930 | /// | ||
931 | /// ``` | ||
932 | /// struct Pair(Option<(u32, u32)>, bool); | ||
933 | /// | ||
934 | /// match (p: Pair) { | ||
935 | /// Pair(None, _) => {} | ||
936 | /// Pair(_, false) => {} | ||
937 | /// } | ||
938 | /// ``` | ||
939 | /// | ||
940 | /// We'll perform the following steps: | ||
941 | /// 1. Start with an empty witness | ||
942 | /// `Witness(vec![])` | ||
943 | /// 2. Push a witness `true` against the `false` | ||
944 | /// `Witness(vec![true])` | ||
945 | /// 3. Push a witness `Some(_)` against the `None` | ||
946 | /// `Witness(vec![true, Some(_)])` | ||
947 | /// 4. Apply the `Pair` constructor to the witnesses | ||
948 | /// `Witness(vec![Pair(Some(_), true)])` | ||
949 | /// | ||
950 | /// The final `Pair(Some(_), true)` is then the resulting witness. | ||
951 | #[derive(Clone, Debug)] | ||
952 | pub(crate) struct Witness(Vec<Pat>); | ||
953 | |||
954 | impl Witness { | ||
955 | /// Asserts that the witness contains a single pattern, and returns it. | ||
956 | fn single_pattern(self) -> Pat { | ||
957 | assert_eq!(self.0.len(), 1); | ||
958 | self.0.into_iter().next().unwrap() | ||
959 | } | ||
960 | |||
961 | /// Constructs a partial witness for a pattern given a list of | ||
962 | /// patterns expanded by the specialization step. | ||
963 | /// | ||
964 | /// When a pattern P is discovered to be useful, this function is used bottom-up | ||
965 | /// to reconstruct a complete witness, e.g., a pattern P' that covers a subset | ||
966 | /// of values, V, where each value in that set is not covered by any previously | ||
967 | /// used patterns and is covered by the pattern P'. Examples: | ||
968 | /// | ||
969 | /// left_ty: tuple of 3 elements | ||
970 | /// pats: [10, 20, _] => (10, 20, _) | ||
971 | /// | ||
972 | /// left_ty: struct X { a: (bool, &'static str), b: usize} | ||
973 | /// pats: [(false, "foo"), 42] => X { a: (false, "foo"), b: 42 } | ||
974 | fn apply_constructor( | ||
975 | mut self, | ||
976 | pcx: PatCtxt<'_>, | ||
977 | ctor: &Constructor, | ||
978 | ctor_wild_subpatterns: &Fields, | ||
979 | ) -> Self { | ||
980 | let pat = { | ||
981 | let len = self.0.len(); | ||
982 | let arity = ctor_wild_subpatterns.len(); | ||
983 | let pats = self.0.drain((len - arity)..).rev(); | ||
984 | ctor_wild_subpatterns.replace_fields(pcx.cx, pats).apply(pcx, ctor) | ||
985 | }; | ||
986 | |||
987 | self.0.push(pat); | ||
988 | |||
989 | self | ||
990 | } | ||
991 | } | ||
992 | |||
993 | /// Algorithm from <http://moscova.inria.fr/~maranget/papers/warn/index.html>. | ||
994 | /// The algorithm from the paper has been modified to correctly handle empty | ||
995 | /// types. The changes are: | ||
996 | /// (0) We don't exit early if the pattern matrix has zero rows. We just | ||
997 | /// continue to recurse over columns. | ||
998 | /// (1) all_constructors will only return constructors that are statically | ||
999 | /// possible. E.g., it will only return `Ok` for `Result<T, !>`. | ||
1000 | /// | ||
1001 | /// This finds whether a (row) vector `v` of patterns is 'useful' in relation | ||
1002 | /// to a set of such vectors `m` - this is defined as there being a set of | ||
1003 | /// inputs that will match `v` but not any of the sets in `m`. | ||
1004 | /// | ||
1005 | /// All the patterns at each column of the `matrix ++ v` matrix must have the same type. | ||
1006 | /// | ||
1007 | /// This is used both for reachability checking (if a pattern isn't useful in | ||
1008 | /// relation to preceding patterns, it is not reachable) and exhaustiveness | ||
1009 | /// checking (if a wildcard pattern is useful in relation to a matrix, the | ||
1010 | /// matrix isn't exhaustive). | ||
1011 | /// | ||
1012 | /// `is_under_guard` is used to inform if the pattern has a guard. If it | ||
1013 | /// has one it must not be inserted into the matrix. This shouldn't be | ||
1014 | /// relied on for soundness. | ||
1015 | fn is_useful( | ||
1016 | cx: &MatchCheckCtx<'_>, | ||
1017 | matrix: &Matrix, | ||
1018 | v: &PatStack, | ||
1019 | witness_preference: WitnessPreference, | ||
1020 | is_under_guard: bool, | ||
1021 | is_top_level: bool, | ||
1022 | ) -> Usefulness { | ||
1023 | let Matrix { patterns: rows, .. } = matrix; | ||
1024 | |||
1025 | // The base case. We are pattern-matching on () and the return value is | ||
1026 | // based on whether our matrix has a row or not. | ||
1027 | // NOTE: This could potentially be optimized by checking rows.is_empty() | ||
1028 | // first and then, if v is non-empty, the return value is based on whether | ||
1029 | // the type of the tuple we're checking is inhabited or not. | ||
1030 | if v.is_empty() { | ||
1031 | let ret = if rows.is_empty() { | ||
1032 | Usefulness::new_useful(witness_preference) | ||
1033 | } else { | ||
1034 | Usefulness::new_not_useful(witness_preference) | ||
1035 | }; | ||
1036 | return ret; | ||
1037 | } | ||
1038 | |||
1039 | assert!(rows.iter().all(|r| r.len() == v.len())); | ||
1040 | |||
1041 | // FIXME(Nadrieril): Hack to work around type normalization issues (see rust-lang/rust#72476). | ||
1042 | let ty = matrix.heads().next().map_or(cx.type_of(v.head()), |r| cx.type_of(r)); | ||
1043 | let pcx = PatCtxt { cx, ty: &ty, is_top_level }; | ||
1044 | |||
1045 | // If the first pattern is an or-pattern, expand it. | ||
1046 | let ret = if v.head().is_or_pat(cx) { | ||
1047 | //expanding or-pattern | ||
1048 | let v_head = v.head(); | ||
1049 | let vs: Vec<_> = v.expand_or_pat(cx).collect(); | ||
1050 | let alt_count = vs.len(); | ||
1051 | // We try each or-pattern branch in turn. | ||
1052 | let mut matrix = matrix.clone(); | ||
1053 | let usefulnesses = vs.into_iter().enumerate().map(|(i, v)| { | ||
1054 | let usefulness = is_useful(cx, &matrix, &v, witness_preference, is_under_guard, false); | ||
1055 | // If pattern has a guard don't add it to the matrix. | ||
1056 | if !is_under_guard { | ||
1057 | // We push the already-seen patterns into the matrix in order to detect redundant | ||
1058 | // branches like `Some(_) | Some(0)`. | ||
1059 | matrix.push(v, cx); | ||
1060 | } | ||
1061 | usefulness.unsplit_or_pat(i, alt_count, v_head) | ||
1062 | }); | ||
1063 | Usefulness::merge(witness_preference, usefulnesses) | ||
1064 | } else { | ||
1065 | let v_ctor = v.head_ctor(cx); | ||
1066 | // if let Constructor::IntRange(ctor_range) = v_ctor { | ||
1067 | // // Lint on likely incorrect range patterns (#63987) | ||
1068 | // ctor_range.lint_overlapping_range_endpoints( | ||
1069 | // pcx, | ||
1070 | // matrix.head_ctors_and_spans(cx), | ||
1071 | // matrix.column_count().unwrap_or(0), | ||
1072 | // hir_id, | ||
1073 | // ) | ||
1074 | // } | ||
1075 | |||
1076 | // We split the head constructor of `v`. | ||
1077 | let split_ctors = v_ctor.split(pcx, matrix.head_ctors(cx)); | ||
1078 | // For each constructor, we compute whether there's a value that starts with it that would | ||
1079 | // witness the usefulness of `v`. | ||
1080 | let start_matrix = matrix; | ||
1081 | let usefulnesses = split_ctors.into_iter().map(|ctor| { | ||
1082 | // debug!("specialize({:?})", ctor); | ||
1083 | // We cache the result of `Fields::wildcards` because it is used a lot. | ||
1084 | let ctor_wild_subpatterns = Fields::wildcards(pcx, &ctor); | ||
1085 | let spec_matrix = | ||
1086 | start_matrix.specialize_constructor(pcx, &ctor, &ctor_wild_subpatterns); | ||
1087 | let v = v.pop_head_constructor(&ctor_wild_subpatterns, cx); | ||
1088 | let usefulness = | ||
1089 | is_useful(cx, &spec_matrix, &v, witness_preference, is_under_guard, false); | ||
1090 | usefulness.apply_constructor(pcx, start_matrix, &ctor, &ctor_wild_subpatterns) | ||
1091 | }); | ||
1092 | Usefulness::merge(witness_preference, usefulnesses) | ||
1093 | }; | ||
1094 | |||
1095 | ret | ||
1096 | } | ||
1097 | |||
1098 | /// The arm of a match expression. | ||
1099 | #[derive(Clone, Copy)] | ||
1100 | pub(crate) struct MatchArm { | ||
1101 | pub(crate) pat: PatId, | ||
1102 | pub(crate) has_guard: bool, | ||
1103 | } | ||
1104 | |||
1105 | /// Indicates whether or not a given arm is reachable. | ||
1106 | #[derive(Clone, Debug)] | ||
1107 | pub(crate) enum Reachability { | ||
1108 | /// The arm is reachable. This additionally carries a set of or-pattern branches that have been | ||
1109 | /// found to be unreachable despite the overall arm being reachable. Used only in the presence | ||
1110 | /// of or-patterns, otherwise it stays empty. | ||
1111 | Reachable(Vec<PatId>), | ||
1112 | /// The arm is unreachable. | ||
1113 | Unreachable, | ||
1114 | } | ||
1115 | |||
1116 | /// The output of checking a match for exhaustiveness and arm reachability. | ||
1117 | pub(crate) struct UsefulnessReport { | ||
1118 | /// For each arm of the input, whether that arm is reachable after the arms above it. | ||
1119 | pub(crate) _arm_usefulness: Vec<(MatchArm, Reachability)>, | ||
1120 | /// If the match is exhaustive, this is empty. If not, this contains witnesses for the lack of | ||
1121 | /// exhaustiveness. | ||
1122 | pub(crate) non_exhaustiveness_witnesses: Vec<Pat>, | ||
1123 | } | ||
1124 | |||
1125 | /// The entrypoint for the usefulness algorithm. Computes whether a match is exhaustive and which | ||
1126 | /// of its arms are reachable. | ||
1127 | /// | ||
1128 | /// Note: the input patterns must have been lowered through | ||
1129 | /// `check_match::MatchVisitor::lower_pattern`. | ||
1130 | pub(crate) fn compute_match_usefulness( | ||
1131 | cx: &MatchCheckCtx<'_>, | ||
1132 | arms: &[MatchArm], | ||
1133 | ) -> UsefulnessReport { | ||
1134 | let mut matrix = Matrix::empty(); | ||
1135 | let arm_usefulness: Vec<_> = arms | ||
1136 | .iter() | ||
1137 | .copied() | ||
1138 | .map(|arm| { | ||
1139 | let v = PatStack::from_pattern(arm.pat); | ||
1140 | let usefulness = is_useful(cx, &matrix, &v, LeaveOutWitness, arm.has_guard, true); | ||
1141 | if !arm.has_guard { | ||
1142 | matrix.push(v, cx); | ||
1143 | } | ||
1144 | let reachability = match usefulness { | ||
1145 | NoWitnesses(subpats) if subpats.is_empty() => Reachability::Unreachable, | ||
1146 | NoWitnesses(subpats) => { | ||
1147 | Reachability::Reachable(subpats.list_unreachable_subpatterns(cx).unwrap()) | ||
1148 | } | ||
1149 | WithWitnesses(..) => panic!("bug"), | ||
1150 | }; | ||
1151 | (arm, reachability) | ||
1152 | }) | ||
1153 | .collect(); | ||
1154 | |||
1155 | let wild_pattern = | ||
1156 | cx.pattern_arena.borrow_mut().alloc(Pat::wildcard_from_ty(cx.infer[cx.match_expr].clone())); | ||
1157 | let v = PatStack::from_pattern(wild_pattern); | ||
1158 | let usefulness = is_useful(cx, &matrix, &v, ConstructWitness, false, true); | ||
1159 | let non_exhaustiveness_witnesses = match usefulness { | ||
1160 | WithWitnesses(pats) => pats.into_iter().map(Witness::single_pattern).collect(), | ||
1161 | NoWitnesses(_) => panic!("bug"), | ||
1162 | }; | ||
1163 | UsefulnessReport { _arm_usefulness: arm_usefulness, non_exhaustiveness_witnesses } | ||
1164 | } | ||
1165 | |||
1166 | pub(crate) type PatternArena = Arena<Pat>; | ||
1167 | |||
1168 | mod helper { | ||
1169 | use super::MatchCheckCtx; | ||
1170 | |||
1171 | pub(super) trait PatIdExt: Sized { | ||
1172 | // fn is_wildcard(self, cx: &MatchCheckCtx<'_>) -> bool; | ||
1173 | fn is_or_pat(self, cx: &MatchCheckCtx<'_>) -> bool; | ||
1174 | fn expand_or_pat(self, cx: &MatchCheckCtx<'_>) -> Vec<Self>; | ||
1175 | } | ||
1176 | |||
1177 | // Copy-pasted from rust/compiler/rustc_data_structures/src/captures.rs | ||
1178 | /// "Signaling" trait used in impl trait to tag lifetimes that you may | ||
1179 | /// need to capture but don't really need for other reasons. | ||
1180 | /// Basically a workaround; see [this comment] for details. | ||
1181 | /// | ||
1182 | /// [this comment]: https://github.com/rust-lang/rust/issues/34511#issuecomment-373423999 | ||
1183 | // FIXME(eddyb) false positive, the lifetime parameter is "phantom" but needed. | ||
1184 | #[allow(unused_lifetimes)] | ||
1185 | pub(crate) trait Captures<'a> {} | ||
1186 | |||
1187 | impl<'a, T: ?Sized> Captures<'a> for T {} | ||
1188 | } | ||