diff options
Diffstat (limited to 'crates/hir_ty/src/diagnostics/match_check/deconstruct_pat.rs')
-rw-r--r-- | crates/hir_ty/src/diagnostics/match_check/deconstruct_pat.rs | 907 |
1 files changed, 907 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 | } | ||