If SubstitutionMap is asked to form a substitution for a generic
parameter that has been made concrete by the generic signature,
substitute into the concrete type. This allows us to better deal with
non-canonical types.
We previously used a simple heuristic of visiting the disjunction with
the fewest number of elements in it.
Instead, this commit introduces a new way to select the best disjunction
to explore based on attempting disjunctions based on either how much
information we have about the associated type variables (e.g. how many
of the arguments of a function call in the case of bind overload
disjunctions) or by other criteria that help constraint the search or
split the connected components (e.g. coercions and calls to
initializers).
A key part of the improvement here is allowing the type checker to
attempt bindings of types to type variables when there are still
argument conversion constraints between type variables in the
system. The insight here is if there are no other constraints blocking
an attempt to bind types and we have types to try, we will be able to
test those types against active conformance constraints and potentially
fail much earlier while solving.
Visiting disjunctions in a different order exposed some other problems
with the type checker including a couple cases where map is used that
are now considered ambiguous due to problems in how we are ranking
solutions.
I measured an 11-13% reduction in type checking time for the standard
library using a release build of the compiler.
A variety of enhancements from SE-154 and SE-165:
- Custom Keys and Values collections for Dictionary
- Two sequence-based Dictionary initializers
- Merging methods for Dictionary
- Capacity checking and reserving for Dictionary and Set
- Type-specific filter(_:) methods for Dictionary and Set
- A mapValues(_:) method for Dictionary
- A grouping Dictionary initializer
Previously we had more ad hoc logic that tried to decide if it was
worth desugaring a type based on its structure. Now we instead look
for a typealias that might actually benefit from desugaring, and if
we don't find one we won't show the 'aka' note.
* [stdlib] String : RangeReplaceableCollection & BidirectionalCollection
* Add source compatibility hack for Swift.max
* Add source compatibility hack for Swift.min
* Remove redundant conformance in benchmarks
* Fix stupid typo I thought I'd already pushed
* XFAIL testing now-redundant conformance
* XFAIL an IDE test for now
Consider the following setup:
struct GenericStruct<T> {
typealias Dependent = T
typealias Concrete = Int
}
We have no way to model 'GenericStruct.Dependent' in the AST, so the
reference would crash. Instead, produce a diagnostic suggesting to
insert generic parameters, like 'GenericStruct<Int>.Dependent'.
The reference 'GenericStruct.Concrete' is fine though, and should
not crash; add a test that it works.
Fixes <https://bugs.swift.org/browse/SR-4390>, <rdar://problem/31480755>.
Eliminate the vestigial type `LazyFilterIndex`, which was
necessary pre-Swift-3 to allow the index to move. Swift 3's indexing
model means that the movement of indices is on the collection itself,
so we no longer need `LazyFilterIndex`: instead, the `Index` type of
the lazy filtered collection is simply the `Index` type of the base
collection, which is a nice convenience: it means you can take indices
from a lazy wrapper around a given collection C and use them with the
collection C (and, with care, vice-versa) without jumping through
extra hoops.
Address ABI FIXME #68 by using same-type constraints directly on an
associated type to describe the requirements on the Indices associated
type of the Collection protocol. ABI FIXMEs #89, #90, #91 are all in
StdlibUnittest, and provoke warnings once #68 is fixed, but it's nice
to clear them out.
Fixes SR-2121.
The core substitution routine for the archetypes-to-interface types
substitution was attempting to provide mappings for nested archetypes,
when in fact these archetypes would (1) always be resolvable via their
parent, and (2) could in fact cause infinite recursion, as with the
new test case. Fixes SR-4617 / rdar://problem/31673819.
Fix an odd corner case when UseErrorTypes was off; we would
return the empty type if dependent member type substitution
failed, but otherwise return the original type if it was a
generic type parameter or an archetype.
Now, if UseErrorTypes is off, return the empty type in both
cases, even if the original type is 'primary'.
Reimplement isSelfDerivedSource() in terms of the new
visitPotentialArchetypesAlongPath(). Aside from being clearer and
shorter, this formulation eliminates some hackery (the direct
"NestedType" lookup) that was mostly masking a bug in
updateNestedTypeForConformance() where we created a
nested-type-name-match source with an incorrect source for the
purposes of concretizing the nested type.
Fixes SR-4458 / rdar://problem/31375569.
When an otherwise abstract conformance constraint is derived from a
concrete conformance, retain the abstract conformance by removing the
requirement source that involves the concrete conformance. This
eliminates our reliance on the concrete conformance, which is not
retained as part of the generic signature.
Fixes rdar://problem/31163470 and rdar://problem/31520386.
These had previously been removed in
9c2bc50acd because of differences in the
diagnostics being emitted across different platforms.
It looks like we're always emitting the same diagnostic now, so restore
a specific message.
Resolves: rdar://problem/19677545
