This fixes a regression introduced in https://github.com/swiftlang/swift/pull/82574.
The test case demonstrates the issue: we would incorrectly choose the base class
overload of == if one of the parameters was an archetype or dynamic Self.
Fixes rdar://156454697.
Infix logical operators are usually not overloaded and don't
form disjunctions, but when they do, let's prefer them over
other operators when they have fewer choices because it helps
to split operator chains.
- Expand the inference to include prefix and postfix unary operators
- Recognize previously resolved declaration and member references
in argument positions and record their types.
- Expand reconciliation logic from Double<->Int to include other
floating-point types and `CGFloat`.
If the scores are the same and both disjunctions are operators
they could be ranked purely based on whether the candidates
were speculative or not. The one with more context always wins.
Consider the following situation:
```swift
func test(_: Int) { ... }
func test(_: String) { ... }
test("a" + "b" + "c")
```
In this case we should always prefer ... + "c" over "a" + "b"
because it would fail and prune the other overload if parameter
type (aka contextual type) is `Int`.
This is helpful in situations when all of the chained operators
have literal arguments because it would make sure that every
operator has the same score if there is no contextual type.
These choices could be better than some other non-disfavored ones
in certain situations i.e. when `async` overload is disfavored
but appears in async context it's preferrable to a non-async
overload choice.
Note that the code that mimic old hacks still needs to filter on
`@_disfavoredOverload` in few places to maintain source compatibility.
This matches the behavior of the old hack where favoring choices
were rolled back if `mustConsider` produced `true` which happened
only for protocol requirements and variadic overload choice regardless
of their viability.
When matching candidate like `[Int]` against `Array<Element>`
we need to conservatively assume that if the nominals match
the argument is a viable exact match because otherwise it's
possible to skip some of the valid matches when other overload
choice have generic parameters at the same parameter position.
The problem this is trying to solve is eager selection of operators
over unsupported disjunctions, when matching operators let's take
speculative information into account because it helps to make better
choices in this case.
We need to have a notion of "complete" binding set before
we can allow inference from generic parameters and ternary,
otherwise we'd make a favoring decision that might not be
correct i.e. `v ?? (<<cond>> ? nil : o)` where `o` is `Int`.
`getBindingsFor` doesn't currently infer transitive bindings
which means that for a ternary we'd only have a single
binding - `Int` which could lead to favoring overload of
`??` and has non-optional parameter on the right-hand side.
Some of the disjunctions are not supported by the optimizers but
could still be a better choice than an operator. Using a non-score
based preference mechanism first allows us to make sure that
operator disjunctions are not selected too eagerly in some situations
when i.e. a member (supported or not) could be a better choice.
`isPreferable` currently targets only operators in result builder
contexts but it could be expanded to more uses in the future.
New ranking + selection algorithm suffered from over-eagerly selecting
operator disjunctions vs. unsupported non-operator ones even if the
ranking was based purely on literal candidates.
This change introduces a notion of a speculative candidate - one which
has a type inferred from a literal or an initializer call that has
failable overloads and/or implicit conversions (i.e. Double/CGFloat).
`determineBestChoicesInContext` would reset the score of an operator
disjunction which was computed based on speculative candidates alone
but would preserve favoring information. This way selection algorithm
would not be skewed towards operators and at the same time if there
is no no choice by to select one we'd still have favoring information
available which is important for operator chains that consist purely
of literals.
Thanks to `LinkedExprAnalyzer` unary argument hack was able to
infer matching based on literals and arithmetic operator chains,
let's preserve that behavior in a more principled manner.
`==` and `!=` operators have special overloads that allow matching
`nil` literal on either side even if wrapped type on the other side
doesn't conform to `Equatable`.
If there are no-same type requirements and parameters use
either concrete types or generic parameter types directly,
the optimizer should be able to handle ranking. Currently
candidate arguments are considered in isolation which makes
it impossible to deal with same-type requirements and
complex generic signatures.
Disjunctions with a single element are sometimes introduced after
disfavoring, so we need to make sure that they are always preferred
during disjunction selection.
Having it be part of the other matching wasn't a good idea because
previous "favoring" happened only in a few situations - if argument
was a declaration reference, application or (dynamic) subscript that
had overload choice selected during constraint generation.
Since each candidate and overload choice are considered independenty
there is no way to judge whether non-default literal type is going
to result in a worse solution than non-default one.
For example, `??` operator could produce an optional type
so `test(<<something>> ?? 0) could result in an optional
argument that wraps a type variable. It should be possible
to infer bindings from underlying type variable and restore
optionality.
* Revert "[CSOptimizer] Look through `OptionalEvaluationExpr`s when dealing with unapplied disjunctions"
This reverts commit 72340f39b8.
* Revert "[CSOptimizer] Don't consider disabled overloads when checking whether disjunction is supported"
This reverts commit 6bc23b5057.
* Revert "[CSOptimizer] Disjunctions with IUO overload choices are unsupported"
This reverts commit 471ee21535.
* Revert "[CSOptimizer] MemberImportVisibility: Don't consider overloads that come from implicit imports"
This reverts commit aa4a2b9071.
* Revert "[CSOptimizer] Don't consider CGFloat widening when explicit initializer is used"
This reverts commit 3cc76eacdd.
* Revert "[CSOptimizer] Literal arguments should cause score reset only for operators"
This reverts commit e3987beffb.
* Revert "[CSOptimizer] NFC: check whether a choice is of operator instead of whole disjunction"
This reverts commit 6c82892c3c.
* Revert "[CSOptimizer/Tests] NFC: Add a perf test-case fixed by improved literal array handling"
This reverts commit cfd34e54c4.
* Revert "[CSOptimizer] Extend candidate/parameter matching to support array literals"
This reverts commit 8a304f88c6.
* Revert "[CSOptimizer] Favor choices that don't require application"
This reverts commit 0737542da8.
* Revert "[CSOptimizer] Disable CGFloat -> Double conversion for unary operators"
This reverts commit bc3a15fbe6.
* Revert "[CSOptimizer] Mark bitwise operators as supported"
This reverts commit 860ae08d1b.
* Revert "[CSOptimizer] Simplify handling of non-applied disjunctions"
This reverts commit 43ca7dfff9.
* Revert "[ConstraintSystem] Fix `getEffectiveOverloadType` handling of `mutating` methods"
This reverts commit c767f7aff7.
* Revert "[CSOptimizer] Reduce overload types before ranking"
This reverts commit 95b47aead6.
* Revert "[CSOptimizer] Implement special prioritization rules for result builder contexts"
This reverts commit 56d6635e46.
* Revert "[CSOptimizer] Allow only widening CGFloat->Double conversions while matching candidate arguments"
This reverts commit bf8ae3bc1b.
* Revert "[CSSimplify] CGFloat-Double: Rank narrowing correctly when result is injected into an optional"
This reverts commit cb876cbd9e.
* Revert "[CSBindings] Prevent `BindingSet::isViable` from dropping viable bindings (v2)"
This reverts commit b7e7493076.
* Revert "[CSOptimizer] Add support for chained members without arguments"
This reverts commit 87cd5f8733.
* Revert "[CSOptimizer] Mark compiler synthesized disjunctions as optimized"
This reverts commit 867e64182f.
* Revert "[CSOptimizer] Make a light-weight generic overload check if some requirements are unsatisfiable"
This reverts commit 15c773b9d7.
* Revert "[CSOptimizer] Fix `selectDisjunction` to use favored choices even if disjunction was not optimized"
This reverts commit c2a55886f0.
* Revert "[CSOptimizer] Limit "old" behavior compatibility to unlabeled unary arguments"
This reverts commit 9fb73143f6.
* Revert "[Tests] NFC: Update a couple of type-checker tests"
This reverts commit ff8663ff16.
* Revert "[Tests] NFC: Move simd related test-case from `slow` to `fast`"
This reverts commit 28396a6dce.
* Revert "[CSGen] NFC: Remove obsolete `ConstraintSystem::{get, set}FavoredType`"
This reverts commit 8bd288447f.
* Revert "[CSOptimizer] Allow literal arguments to match parameters that conform to `ExpressibleBy{Integer, Float}Literal`"
This reverts commit 2fdd4b6c35.
* Revert "[CSOptimizer] Adjust `scoreCandidateMatch` to indicate when match cannot be decided"
This reverts commit 9b62c84a4f.
* Revert "[CSOptimizer] Fix Double<->CGFloat implicit conversion support when arguments are literals"
This reverts commit 6caf1ccbb2.
* Revert "[CSOptimizer] A more comprehensive generic overload checking when candidates are fully resolved"
This reverts commit e30587bda4.
* Revert "[CSOptimizer] Restore old hack behavior which used to favor overloads based on arity matches"
This reverts commit a3a3ec4fe0.
* Revert "[CSOptimizer] Desugar types before checking for equality"
This reverts commit 802f5cd105.
* Revert "[ConstraintSystem] Narrowly disable `tryOptimizeGenericDisjunction` when some of the arguments are number literals"
This reverts commit 8d5cb112ef.
* Revert "[CSOptimizer] Infer argument candidates from calls to `Double` and CGFloat constructors"
This reverts commit f2a6677a6d.
* Revert "[CSOptimizer] Score all of the overload choices matching on literals uniformly"
This reverts commit 59109c2d60.
* Revert "[CSOptimizer] Enable ranking of `Int*`, `Float{80}` and `Double` initializers"
This reverts commit 6fb6d1cf90.
* Revert "[CSOptimizer] Rank disjunctions based on score only if both sides are supported"
This reverts commit 8818d399f9.
* Revert "[CSOptimizer] Rank results of operators regardless of whether anything is known about parameters"
This reverts commit 3996b25fbd.
* Revert "[Tests] NFC: Add more test-cases that were previously solved due to old hacks behavior"
This reverts commit d0ff6c81b8.
* Revert "[CSOptimizer] Average score should reflect number of defaulted parameters"
This reverts commit 23589add74.
* Revert "[Tests] NFC: Adjust a couple of improved tests"
This reverts commit 66981364fe.
* Revert "[CSOptimizer] Don't optimize (implicit) calls with code completion arguments"
This reverts commit 8a918e2369.
* Revert "[CSOptimizer] attempt to rank only standard/simd operators and fully concrete overload sets"
This reverts commit deca9b61c5.
* Revert "[CSOptimizer] Record best scores for each disjunction and use them in `selectDisjunction`"
This reverts commit 3819ddfb40.
* Revert "[CSOptimizer] Let `determineBestChoicesInContext` return the best disjunction if one is available"
This reverts commit cf05405eae.
* Revert "[CSOptimizer] Emulate old behavior related to favoring of unary calls to members"
This reverts commit 527de22bec.
* Revert "[Tests] NFC: Add a test-case for rdar://133340307 which is now fast"
This reverts commit 670127abd6.
* Revert "[CSOptimizer] Prefer homogeneous arithmetic operator overloads when argument(s) or result match"
This reverts commit d69b6a0594.
* Revert "[CSOptimizer] Remove an outdated optimization to compare resolved argument types with all else equal"
This reverts commit 1760bd1f1e.
* Revert "[CSOptimizer] NFC: Switch from llvm::Optional to std::optional post-rebase"
This reverts commit c429f5b9ec.
* Revert "[CSOptimizer] Increase score when type matches opaque type"
This reverts commit 2869dff995.
* Revert "[CSOptimizer] NFC: Switch to llvm::Optional"
This reverts commit 0fc6806922.
* Revert "[CSOptimizer] NFC: Adjust conformance check to use `ConstraintSystem::lookupConformance`"
This reverts commit da65333d41.
* Revert "[CSOptimizer] Treat all type parameters equally"
This reverts commit 957a5f4270.
* Revert "[CSStep] Remove disjunction pruning logic from DisjunctionStep"
This reverts commit 2c44e37948.
* Revert "[CSOptimizer] Relax candidate type requirements from equality to set of no-impact conversions"
This reverts commit 11b897b32f.
* Revert "[CSOptimizer] Use `matchCallArguments` to establish argument-to-parameter relationships"
This reverts commit cb1cb2018d.
* Revert "[CSOptimizer] Don't attempt to optimize calls with code completion token(s) in argument position"
This reverts commit 14e2a16fce.
* Revert "[CSOptimizer] Allow generic operator overloads without associated type parameters"
This reverts commit bc5f70a9a3.
* Revert "[CSOptimizer] Make sure that all parameters without arguments are defaulted"
This reverts commit 7c1c46d4e4.
* Revert "[CSStep] Don't favor choices until the disjunction is picked"
This reverts commit e404ed722a.
* Revert "[CSOptimizer] Keep track of mismatches while evaluating candidates"
This reverts commit a094c3ebb0.
* Revert "[CSOptimizer] Favor SIMD related arithmetic operator choices if argument is SIMD<N> type"
This reverts commit c2f7451c7b.
* Revert "[CSOptimizer] Initial implementation of disjunction choice favoring algorithm"
This reverts commit 672ae3d252.
* Revert "[ConstraintSystem] Add skeleton of constraint optimizer"
This reverts commit b5f08a4009.
* Revert "[CSGen] Remove ConstraintOptimizer and all favoring logic"
This reverts commit 4432c51f57.
* Revert "[ConstraintSystem] Remove `shrink`"
This reverts commit 757ca24e8a.
* [TypeChecker] NFC: Remove resurrected use of `SolverShrinkUnsolvedThreshold`
* [TypeChecker] Bring back `SolverDisableShrink`
* [Tests] NFC: Mark tests affected by solver-perf revert as slow
* [Tests] NFC: Adjust async tests that are affected by performance hacks
Ignore declarations that come from implicitly imported modules
when `MemberImportVisibility` feature is enabled otherwise
we might end up favoring an overload that would be diagnosed
as unavailable later.
Resolves: rdar://143582881
The two GatherKinds no longer share any implementation, so there's
no point keeping the logic together. Doing this also allows removing
the acceptConstraintFn from gatherAllConstraints(), which further
simplifies depthFirstSearch().
