diff options
Diffstat (limited to 'compiler/Language/Haskell/Syntax/Expr.hs')
| -rw-r--r-- | compiler/Language/Haskell/Syntax/Expr.hs | 64 |
1 files changed, 56 insertions, 8 deletions
diff --git a/compiler/Language/Haskell/Syntax/Expr.hs b/compiler/Language/Haskell/Syntax/Expr.hs index a9592304e6..6f5150a1b4 100644 --- a/compiler/Language/Haskell/Syntax/Expr.hs +++ b/compiler/Language/Haskell/Syntax/Expr.hs @@ -267,6 +267,55 @@ is Less Cool because typecheck do-notation with (>>=) :: m1 a -> (a -> m2 b) -> m2 b.) -} +{- +Note [Record selectors in the AST] +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ +Here is how record selectors are expressed in GHC's AST: + +Example data type + data T = MkT { size :: Int } + +Record selectors: + | GhcPs | GhcRn | GhcTc | +----------------------------------------------------------------------------------| +size (assuming one | HsVar | HsRecSel | HsRecSel | + 'size' in scope) | | | | +----------------------|--------------|----------------------|---------------------| +.size (assuming | HsProjection | getField @"size" | getField @"size" | + OverloadedRecordDot) | | | | +----------------------|--------------|----------------------|---------------------| +e.size (assuming | HsGetField | getField @"size" e | getField @"size" e | + OverloadedRecordDot) | | | | + +NB 1: DuplicateRecordFields makes no difference to the first row of +this table, except that if 'size' is a field of more than one data +type, then a naked use of the record selector 'size' may well be +ambiguous. You have to use a qualified name. And there is no way to do +this if both data types are declared in the same module. + +NB 2: The notation getField @"size" e is short for +HsApp (HsAppType (HsVar "getField") (HsWC (HsTyLit (HsStrTy "size")) [])) e. +We track the original parsed syntax via HsExpanded. + +-} + +{- +Note [Non-overloaded record field selectors] +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ +Consider + data T = MkT { x,y :: Int } + f r x = x + y r + +This parses with HsVar for x, y, r on the RHS of f. Later, the renamer +recognises that y in the RHS of f is really a record selector, and +changes it to a HsRecSel. In contrast x is locally bound, shadowing +the record selector, and stays as an HsVar. + +The renamer adds the Name of the record selector into the XCFieldOcc +extension field, The typechecker keeps HsRecSel as HsRecSel, and +transforms the record-selector Name to an Id. +-} + -- | A Haskell expression. data HsExpr p = HsVar (XVar p) @@ -285,11 +334,10 @@ data HsExpr p -- solving. See Note [Holes] in GHC.Tc.Types.Constraint. - | HsRecFld (XRecFld p) - (AmbiguousFieldOcc p) -- ^ Variable pointing to record selector - -- The parser produces HsVars - -- The renamer renames record-field selectors to HsRecFld - -- The typechecker preserves HsRecFld + | HsRecSel (XRecSel p) + (FieldOcc p) -- ^ Variable pointing to record selector + -- See Note [Non-overloaded record field selectors] and + -- Note [Record selectors in the AST] | HsOverLabel (XOverLabel p) FastString -- ^ Overloaded label (Note [Overloaded labels] in GHC.OverloadedLabels) @@ -334,7 +382,7 @@ data HsExpr p -- NB Bracketed ops such as (+) come out as Vars. -- NB Sadly, we need an expr for the operator in an OpApp/Section since - -- the renamer may turn a HsVar into HsRecFld or HsUnboundVar + -- the renamer may turn a HsVar into HsRecSel or HsUnboundVar | OpApp (XOpApp p) (LHsExpr p) -- left operand @@ -486,7 +534,7 @@ data HsExpr p -- - 'GHC.Parser.Annotation.AnnKeywordId' : 'GHC.Parser.Annotation.AnnDot' -- -- This case only arises when the OverloadedRecordDot langauge - -- extension is enabled. + -- extension is enabled. See Note [Record Selectors in the AST]. | HsGetField { gf_ext :: XGetField p @@ -500,7 +548,7 @@ data HsExpr p -- 'GHC.Parser.Annotation.AnnDot', 'GHC.Parser.Annotation.AnnCloseP' -- -- This case only arises when the OverloadedRecordDot langauge - -- extensions is enabled. + -- extensions is enabled. See Note [Record Selectors in the AST]. | HsProjection { proj_ext :: XProjection p |