Drop witness markers for 'uninteresting' dependent types #5687
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@swift-ci please smoke test |
|
Alright, I clearly need to address this associated type witness inference issue before I have any hope of landing this change. |
Recently I changed the ArchetypeBuilder is minimize requirements in generic signatures. However substitution lists still contained all recursively-expanded nested types. With recursive conformances, this list becomes potentially infinite, so we can't expand it out anymore. Also, it is just a waste of time to have them there.
Extensions of the RandomAccessCollection protocol provide two
mutually-incompatible default implementations for 'indices', one of
which uses DefaultRandomAccessIndices<Self> and the other uses
CountableRange<Index>. Type witness inference (rightly) considers the
two ambiguous, although it didn't before, causing regressions.
Remove the same-type constraint
SubSequence == DefaultRandomAccessIndices<Self>
from the most general extension, so that the more-specialized,
CountableRange version is selected when it's constraints are met, and
the most-general version is selected otherwise.
The down-side to this change is that the most-general version of
'indices', which produces a DefaultRandomAccessIndices<Self>, will be
available in all types that conform to RandomAccessCollection, even if
they have a different Indices type.
…llection. The BidirectionalCollection protocol has no requirements that mention the associated type 'SubSequence', which means it cannot be easily inferred. Duplicate the subscript requirement from Collection into BidirectionalCollection to aid type witness inference.
…e" check. Expand the context conversion failure check used to fix rdar://28909024 to cover the case where we have a single-argument mismatch and there is a type parameter deduction as well.
They're relying on having extraneous witness markers to work around some oddities in the modeling of property behaviors as protocols.
Debugging type witness inference is annoying; add some dumping code to help with debugging.
DougGregor
force-pushed
the
drop-witness-markers
branch
from
66b665d to
eacf223
Compare
DougGregor
changed the title
|
@swift-ci please smoke test and merge |
1 similar comment
|
@swift-ci please smoke test and merge |
…lection. Similar to what we did for BidirectionalCollection, duplicate the subscript-returning-SubSequence requirement in RandomAccessCollection.
|
@swift-ci please smoke test and merge |
1 similar comment
|
@swift-ci please smoke test and merge |
|
@swift-ci please smoke test Linux |
|
I don't think that timeout is related to my changes |
|
@swift-ci please smoke test Linux |
|
@swift-ci please merge |
|
This sped up type checking of the standard library by 11%. Cool. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This pull request removes witness markers for 'uninteresting' dependent types--e.g, those dependent types for which there are no additional protocol conformances required beyond what is implied by other constraints--from substitution lists and the list of "all" dependent types in a generic signature. For example, given something like:
a generic signature
<T: Q>currently would contain witness markers for bothTandT.Assoc, even though everything aboutT.Associs recoverable from the conformance ofTtoQ. With this change, we would only record a witness marker forT.It's a necessary step toward implementing recursive protocol conformances (because the list of "all" witness markers would be infinite were
Assocto conform toQ, i.e.,T.Assoc,T.Assoc.Assoc,T.Assoc.Assoc.Assoc, etc.). It's also a cleaner way to model the problem, and a step toward eliminating witness markers altogether.This pull request is still a work-in-progress for two reasons:
Indiceswitness for a minimalRandomAccessCollectionmodel, and the other involves a change in inference for theSubSequenceof the existential collections. The former is potentially a bug fix to the type checker that may imply a library change, while the latter appears to be an outright bug in the inference of associated type witnesses.