tree-sitter-haskell

Haskell grammar for tree-sitter.

Note This grammar needs at least tree-sitter 0.19.4 and C++-14.

References

• Haskell 2010 Language Report – Syntax References
• GHC Language Extensions

Building with nvim-treesitter

When installing the grammar from source, be sure to include the scanner in the source files:

lua <<EOF
local parser_config = require "nvim-treesitter.parsers".get_parser_configs()
parser_config.haskell = {
  install_info = {
    url = "~/path/to/tree-sitter-haskell",
    files = {"src/parser.c", "src/scanner.cc"}
  }
}
EOF

Depending on what compilers are installed in your system, it may be necessary to force nvim-treesitter to use a specific one to satisfy the C++-14 requirement (see this issue for more info):

lua require'nvim-treesitter.install'.compilers = { "clang" }
" or
lua require'nvim-treesitter.install'.compilers = { "clang++" }

or

lua require'nvim-treesitter.install'.compilers = { "gcc" }

Building on MacOS

If you get loads of C++ errors when building the parser, a workaround might be to install GCC with homebrew, in addition to selecting gcc as the compiler used by tree-sitter, as described for Neovim in the previous section.

First, run this command in a terminal:

$ brew install gcc

This may not be enough, since clang might still be used by default. You can find out whether the gcc executable is linked to something else with:

$ ls -l $(which gcc)

If this doesn't point to something like gcc-11, you can remove it (if it is a symlink) and/or link the real gcc-N binary to a bin directory that's at the head of your $PATH, something like:

ln -sf /opt/homebrew/bin/gcc-11 /usr/local/bin/gcc

or this for earlier versions of homebrew.

ln -sf /usr/local/bin/gcc-11 /usr/local/bin/gcc

Supported Language Extensions

These extensions are supported ✅, unsupported ❌ or not applicable because they don't involve parsing ➖️:

• AllowAmbiguousTypes ➖️
• ApplicativeDo ➖️
• Arrows ❌
• BangPatterns ✅
• BinaryLiterals ✅
• BlockArguments ✅
• CApiFFI ✅
• ConstrainedClassMethods ✅
• ConstraintKinds ✅
• CPP ✅
• CUSKs ✅
• DataKinds ✅
• DatatypeContexts ✅
• DefaultSignatures ✅
• DeriveAnyClass ➖️
• DeriveDataTypeable ➖️
• DeriveFoldable ➖️
• DeriveFunctor ➖️
• DeriveGeneric ➖️
• DeriveLift ➖️
• DeriveTraversable ➖️
• DerivingStrategies ✅
• DerivingVia ✅
• DisambiguateRecordFields ➖️
• DuplicateRecordFields ➖️
• EmptyCase ✅
• EmptyDataDecls ✅
• EmptyDataDeriving ✅
• ExistentialQuantification ✅
• ExplicitForAll ✅
• ExplicitNamespaces ✅
• ExtendedDefaultRules ➖️
• FlexibleContexts ✅
• FlexibleInstances ✅
• ForeignFunctionInterface ✅
• FunctionalDependencies ✅
• GADTs ✅
• GADTSyntax ✅
• GeneralisedNewtypeDeriving ➖️
• GHCForeignImportPrim ✅
• Haskell2010 ➖️
• Haskell98 ➖️
• HexFloatLiterals ✅
• ImplicitParams ✅
• ImplicitPrelude ➖️
• ImportQualifiedPost ✅
• ImpredicativeTypes ➖️
• IncoherentInstances ➖️
• InstanceSigs ✅
• InterruptibleFFI ✅
• KindSignatures ✅
• LambdaCase ✅
• LexicalNegation ❌
• LiberalTypeSynonyms ✅
• LinearTypes ✅
• MagicHash ✅
• Modifiers ❌
• MonadComprehensions ➖️
• MonadFailDesugaring ➖️
• MonoLocalBinds ➖️
• MonomorphismRestriction ➖️
• MultiParamTypeClasses ✅
• MultiWayIf ✅
• NamedFieldPuns ✅
• NamedWildCards ✅
• NegativeLiterals ➖️
• NondecreasingIndentation ❌
• NPlusKPatterns ➖️
• NullaryTypeClasses ✅
• NumDecimals ➖️
• NumericUnderscores ✅
• OverlappingInstances ➖️
• OverloadedLabels ✅
• OverloadedLists ➖️
• OverloadedStrings ➖️
• PackageImports ✅
• ParallelListComp ✅
• PartialTypeSignatures ✅
• PatternGuards ✅
• PatternSynonyms ✅
• PolyKinds ➖️
• PostfixOperators ➖️
• QualifiedDo ✅
• QuantifiedConstraints ✅
• QuasiQuotes ✅
• Rank2Types ✅
• RankNTypes ✅
• RebindableSyntax ➖️
• RecordWildCards ➖️
• RecursiveDo ✅
• RoleAnnotations ✅
• Safe ➖️
• ScopedTypeVariables ✅
• StandaloneDeriving ✅
• StandaloneKindSignatures ✅
• StarIsType ✅
• StaticPointers ❌
• Strict ➖️
• StrictData ➖️
• TemplateHaskell ✅
• TemplateHaskellQuotes ✅
• TraditionalRecordSyntax ➖️
• TransformListComp ✅
• Trustworthy ➖️
• TupleSections ✅
• TypeApplications ✅
• TypeFamilies ✅
• TypeFamilyDependencies ✅
• TypeInType ✅
• TypeOperators ✅
• TypeSynonymInstances ➖️
• UnboxedSums ✅
• UnboxedTuples ✅
• UndecidableInstances ➖️
• UndecidableSuperClasses ➖️
• UnicodeSyntax ✅
• UnliftedFFITypes ➖️
• UnliftedNewtypes ✅
• Unsafe ➖️
• ViewPatterns ✅

Bugs

CPP

Preprocessor #elif and #else directives cannot be handled correctly, since the parser state would have to be manually reset to what it was at the #if. As a workaround, the code blocks in the alternative branches are parsed as part of the directives.

Layout

NondecreasingIndentation is not supported (yet?).

Operators on newlines in do

A strange edge case is when an infix operator follows an expression statement of a do block with an indent of less or equal the do's layout column:

f = do
  readSomething
  >>= doSomething

The >>= causes the do's layout to be terminated, resulting in an AST similar to

f = (do readSomething) >>= doSomething

This is checked heuristically, probably unreliably.