Categorizing GHC extensions

Question

I was wondering if GHC's extensions can be divided into basically two different categories

• those that provide "syntactic suggar" or convenience
• and those that introduce something new, a new paradigm for instance.

Now provided one could divide the existing extensions into the above categories which extension would fit into which category?

Answer 1

I think a more appropriate categorization would be to divide it up by the compiler pipeline:

Syntactic extensions

• -XMagicHash
• -XUnicodeSyntax
• -XNewQualifiedOperators
• -XViewPatterns
• -XNPlusKPatterns
• -XDoRec
• -XTransformListComp
• -XNoImplicitPrelude
• -XPostfixOperators
• -XTupleSections
• -XDisambiguateRecordFields
• -XNamedFieldPuns
• -XRecordWildCards
• -XPackageImports
• -XExplicitForAll
• -XKindSignatures
• ...

Type System Extensions

• -XUnboxedTuples
• -XLiberalTypeSynonyms
• -XGADTs
• -XMultiParamTypeClasses
• -XFlexibleContexts
• -XConstrainedClassMethods
• -XOverlappingInstances and -XIncoherentInstances
• -XTypeFamilies
• -XImplicitParams

Cross-cutting extensions

• -XTemplateHaskell
• -XForeignFunctionInterface

Optimizatsions

• -fenable-rewrite-rules
• -fspec-constr
• -O2

Code Generation Extensions

• -fllvm
• -fasm
• -fvia-C

Runtime Extensions

• -threaded

What do you think? Not every flag is either (a) definable in terms of existing constructions, or (b) a new part of the compiler. It's more subtle.

There are many other extensions too, see if you can classify them in this form.

Answer 2

The flags are already categorized in the flag reference in the GHC's users guide, and the language extensions are broken down into various categories in the section on language features.