Tickets that Richard E. is interested in

Type system

Easy

• #7494: Allow type synonyms in GADT return types
• #8109: As-patterns in type patterns
• #8634: Dysfunctional dependencies
• #9636: Should F Bool be well-formed, if F is an empty closed type family? (blocked by #9637)
• #9687: Need Typeable (,,,,,,,,,,) and friends
• #10116: Warn on incomplete closed type families
• #10756: Impossible patterns
• #10789: Warn about type families stuck on kinds
• #11620: Kind signatures on classes. But with what syntax?
• #11977: Function as pattern return type

Medium

• #3483: Notation for known-inaccessible code, like () in Agda
• #6018: Injective type families (Jan is working on)
• #7296: Incoherent instance lookup is allowed within an instance declaration (by design)
  • #9820: Another example of the same behavior
• #7495: Allowing list syntax for HList
• #8128: Derived instances sometimes have inaccessible code
• #8165: GND should make associated types, too (not very well specified)
• #8388: Have a consistent story around non-* types in a forall
• #9180: Compile-time staticError function; seems easy, but I don't know how to do this.
• #9427: Break cycles in recursive class/type definitions (the second half of the fix for #9200)
• #9547: Better inference for whether () :: * or () :: Constraint
• #9637: Type-level Error that aborts compilation
• #9649: Proper type-level strings
• #9667: Don't make tyvars untouchable when a GADT pattern-match isn't informative
• #9883: Heterogeneous OverloadedLists
• #10114: Non-* bodies of foralls
• #10318: Allow superclass cycles
  • #10592: Allow superclass cycles
• #10581: Recommend ScopedTypeVariables
• #10836: Continue reporting errors between mut. rec. groups
• #11113: Semantics for type-level reduction
• #11339: Why do pattern bindings always infer types?
• #11342: The Char kind
• #11715: Constraint vs *

Hard

• #8828: Type pattern synonyms
• #10227: Backward reasoning from closed type families
• #11511: Think Hard about non-terminating injective type families
• #11962: Induction recursion

Rocket Science

• #1965: Allow existentials in newtypes
• #2256: Quantify over implication constraints
• #4259: Allow recursive checks for compatibility in type families
  • #8423: Ditto for closed type families (may have different solution)
• #7259: Eta expansion of products
• #7961: Implement "nokinds" (RAE is working on!)
  • #9017: Bad error message b/c of missing kind equality
  • #10379: Can't parse prefix [] in kind
• #8338: Incoherent instances without -XIncoherentInstances
  • #2356: Strangeness about GHC's lazy overlap check
• #9429: An alternative to Any. For example, we want Typeable (forall x. x -> x).
• #9562: Type families + hs-boot files = unsafeCoerce
• #10327: Closed type families should reduce regardless of infinite types
• #10514: Generic for existentials

Typechecker

• #8095: More type-families optimizations, dropping coercions without -dcore-lint
• #9450: Interleave checking against an hs-boot file while typechecking definitions
• #9557: Deriving instances is slow
• #10141: Add a hint about CUSKs to relevant error messages
• #10361: Make DeriveAnyClass work with associated type defaults
• #10381: RebindableSyntax and RankNTypes
• #10808: Type families and record updates

Roles & such

More/better roles

• #8177: Roles for type families (RAE owns, but is not working on)
• #9112: GND with Vector/MVector
• #9118: No eta-reduction possible
• #9123: Need higher-kinded roles

Error messages

• #9518

Front end

• #7169: Warn about incomplete record selectors
• #7668: Better locations in deferred type errors
• #9376: Improve error messages for closed type families that get stuck on the dark corners
• #9378: Make unknown LANGUAGE pragmas warnings
• #9394: :info should show instances of data and type families.
• #9784: Report better error for Foo.'Z
• #10056: Remove the magic from parsing ~. Some open design questions.
• #12477: Left-sections in types

Medium

• #7401: Derive Eq and friends for empty datatypes

Design needed

• #7870: Customized error messages

Optimizations

• #10906: Better SPECIALISE instance

Template Haskell

Easy

• #6089: Allow nested declaration splices
• #7808: Allow reification of a data instance name
• #9022: Fix semicolons in pretty-printer
• #9699: Function to list all names in scope
• #10267: Add holes (jstolarek)
• #10385: Add extra check around TH annotations
• #10486: More addTopDecls
• #10572: Quantify over quoted variables
• #10603: Better parentheses in TH pretty-printer

Unknown

• #9693: Stale state left in failed compilation with GHCi and TH.
• #10279: Issue with package names
• #10330: Better locations in error messages

Design needed

• #1475: Adding imports/exports
  • #10391: Reify exports
• #1831: reify never provides variable info
• #4222: Reifying abstract types
• #5467: Haddock in TH
• #8679: Include value and function signatures in their declarations
• #8761: Pattern synonyms
• #10331: Better interop between TH and HsSyn
• #10541: Reifying tyvars' kinds

Generic programming

• #8560: Generic representation for GADTs
• #10514: Generic representation for existentials

Documentation

• #9247: Document -XDatatypeContexts
• #9248: Document -X extensions
• #9737: Document ANN in pragmas chapter

Tasks

• Search for email with subject "Branched and unbranched" in my GHC folder -- about refactoring some CoAxiom stuff
• Simplification for axioms: they can be applied to types, not coercions. See email with subject "checkAxInstCo"
• Refactor the type-checking algorithm, à la Typechecking.