#5144: Pattern synonyms -------------------------------------+------------------------------------ Reporter: simonpj | Owner: cactus Type: feature request | Status: new Priority: normal | Milestone: _|_ Component: Compiler | Version: Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Unknown/Multiple Type of failure: None/Unknown | Difficulty: Unknown Test Case: | Blocked By: Blocking: | Related Tickets: -------------------------------------+------------------------------------ Comment (by cactus): Progress update: Code is in the `pattern-synonyms` branch at https://github.com/gergoerdi/ghc.git Until I am ready to submit for initial review, I will do all kinds of funky rebasing and history rewriting, so use it for read-only snapshots as of now. Using the names from the wiki link, pattern-only pattern synonyms and simple pattern synonyms now (mostly) work. There's a bug in the typing of pattern-synonyms-as-expressions, so e.g. with the following program: {{{ {-# LANGUAGE PatternSynonyms #-} pattern One x = [x] singleton :: a -> [a] singleton x = One x }}} I get an error that the type of `One` is too rigid; I expect this to be easy to fix: {{{ Couldn't match type ‛t0’ with ‛a’ because type variable ‛a’ would escape its scope This (rigid, skolem) type variable is bound by the type signature for singleton :: a -> [a] at /tmp/PatSyn.hs:18:14-21 Expected type: [a] Actual type: [t0] }}} The one big missing feature is exporting pattern synonyms. First of all, since pattern synonyms are in the namespace of constructors, not types, we need special syntax to export them; I propose the following: {{{ module M (pattern P) where pattern P x = _:_:x:_ }}} (note that you can have a type called `P` as well, but you'd get a clash if you had a constructor with that name). To be honest, I have no idea yet how to actually implement exporting, since I haven't looked at that part of GHC yet -- any pointers are appreciated. After I've fixed the bug mentioned and implemented exporting, I'll do a formal submission of my (clean-up) patch for review. After that baseline, the 'bidirectional pattern synonyms' proposal from the wiki page should fit right into it as well. I have no opinion yet how difficult adding associated patsyns will be, I'll need to do more research first. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/5144#comment:15 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler