Reimplement constraint merging for correctness #13292
Merged
Conversation
This file contains 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
Otherwise they will keep accumulating across multiple compiler instances since Stats is global.
When printing a constraint, we print types which might refer to type variables defined in the constraint, but the type printer will rely on ctx.typerState.constraint to determine if these type variables are instantiated, and this might be a different constraint than the one we're trying to print, leading to an incorrect output. This commit fixes this by temporarily setting ctx.typerState.constraint to the current constraint when printing it, this required moving the printing logic from OrderingConstraint to PlainPrinter. At the same time, we drop the distinction between `toText` and `contentsToString` (the former wrapped the printed output in "Constraint(...)" and the latter didn't, now we never do) because preserving it would have been complicated and it didn't seem worth it. Also fix Ordering#toString to correctly print the bounds (the logic was there but it was dead code).
The previous implementation simply combined the content of both constraints into one, but this is not enough since it meant that bounds were not propagated and so transitivity was violated. For example, when merging a constraint containing `?S <: ?T` and one containing `?T <: ?R`, the result did not verify `?S <:< ?R` (see the unit tests added in ConstraintsTest). The new implementation simply starts with one set of constraints and then adds the constraints from the other set one by one using `<:<` which takes care of propagating bounds. This is likely to be more expensive than the previous implementation but it turns out that `TyperState#mergeConstraintWith` is a rare operation (after the previous commit it is only called 43 times when compiling scala3-compiler), so the difference shouldn't be significant. This also incidentally fixes scala#12730 because the previous logic for checking if merging succeeded was flawed.
|
test performance please |
|
performance test scheduled: 1 job(s) in queue, 0 running. |
|
Performance test finished successfully: Visit https://dotty-bench.epfl.ch/13292/ to see the changes. Benchmarks is based on merging with master (e4b421c) |
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.
The previous implementation simply combined the content of both
constraints into one, but this is not enough since it meant that bounds
were not propagated and so transitivity was violated. For example, when
merging a constraint containing
?S <: ?Tand one containing?T <: ?R,the result did not verify
?S <:< ?R(see the unit tests added inConstraintsTest).
The new implementation simply starts with one set of constraints and
then adds the constraints from the other set one by one using
<:<which takes care of propagating bounds. This is likely to be more
expensive than the previous implementation but it turns out that
TyperState#mergeConstraintWithis a rare operation (it isonly called 43 times when compiling scala3-compiler), so
the difference shouldn't be significant.
This also incidentally fixes #12730 because the previous logic for
checking if merging succeeded was flawed.