GitLab

Add kind information to type declarations in FlatCurry

See merge request !32

Revert higher rank polymorphism for now

Closes #43

See merge request !31

See curry-base!19 (comment 8669)
for rationale.

The type representation in the IL does not support higher-kinded-types,
thus this should be fine.

Most notably this includes migrating all messages from positions to
spans.

Perform qualified type constructor lookup in WarnCheck

Closes #85

See merge request !30

Provide GHC-like line snippets in messages

Closes #87

See merge request !28

Provide span information in messages

Closes #90

See merge request !26

- Remove implicit prelude import from AST (after imports are resolved)
- Fix restricted re-export of record labels
- Fix wrong error message when a label in a RecordUpdate is missing
Co-authored-by: Kai-Oliver Prott <kai.prott@hotmail.de>

Currently, the frontend crashes when the user tries to compile something
like:

h :: Show a => f a
h = "abc"

This happens because the 'type signature too general' warning is emitted
after the predicate set is expected to only contain type variables
(despite it actually containing a type application).

Output the non-unifiable kinds.

- Improve normalization of type variables in FlatCurry
- Fix KindInference
- Remove trExpr QualTypeExpr branch
  that caused the tests to run indefinitely
Co-authored-by: Finn Teegen <fte@informatik.uni-kiel.de>

Update the readme file

Closes #91

See merge request !27

Remove AmbiguousTypeVariable and EscapingTypeVariable from the list
of expected failures since those tests are neither present nor currently
possible since they require the RankNTypes extension.

Co-authored-by: Kai-Oliver Prott <kai.prott@hotmail.de>

Co-authored-by: Jan-Hendrik Matthes <stu113804@mail.uni-kiel.de>

Co-authored-by: Kai-Oliver Prott <kai.prott@hotmail.de>

- Fix instance check for data-class (again)
- Correct typing of functional patterns
Co-authored-by: Kai Prott <kai.prott@hotmail.de>

- Read currysource.css during compile-time
- Fix internal error when typechecking module with hidden class methods
- FIx fail when embedding data file for the first time
Co-authored-by: Kai-Oliver Prott <kai.prott@hotmail.de>

- Fix compilation issues in TestFrontend
- Fix some tests that fail with the new data-class
- Fix missing Data context when inferring type of partial record constr.
- Simplify detection of derivable data instances
- Add Data context to (=:=), etc.
- Correct comparison of inferred types with type signatures (fixes #43)
Co-authored-by: Kai-Oliver Prott <kai.prott@hotmail.de>

- Fix typing of ExternalDecls with Contexts
- Export AST unqualified
- Keep SpanInfos for local idents in SyntaxCheck
- Fix  'no value found' when a typeclass is not imported with all methods
- Include more LayoutInfo fixes
- Fix SpanInfo of renamed TypeVars
- Remove transformation of over-applied FunctionPatterns
- Respect LayoutInfo in checks and transformations
Co-authored-by: Kai-Oliver Prott <kai.prott@hotmail.de>

Remove support for existential quantified type variables in data type declarations

See merge request !13Co-authored-by: Finn Teegen <fte@informatik.uni-kiel.de>

- Find case/function guards in WarnCheck
- Find case and function guards and pass them into 'checkPatternMatching'.
  This allows for more fine-grained exhaustion/overlap-checks.
- Process guarded wildcards as non-exhaustive in WarnCheck
- Treat guarded equations with a wildcard pattern as non-exhaustive
  (continue processing the rest of the equations).
- Apply basic heuristic to check whether guard is always true in WarnCheck
- When checking whether a case arm is reachable, consider wildcard
  patterns with a guard containing 'True', 'success' or 'otherwise' to be
  exhaustive.
- Process guards with 'success' or 'otherwise' correctly in WarnCheck
- Emit warning correctly for multi-guard case branches
- Consider case branches where any guard is always true to be
  exhaustive.