BNFC generates non-usable Haskell code (TestFoo.hs)

[dramforever]

The following LBNF generates bad haskell code with a few weird p[Expr]s.

comment "//";
comment "/*" "*/";

separator Expr ",";
coercions Expr 3;

ExprAdd.        Expr            ::= Expr  "+" Expr1;
ExprSub.        Expr            ::= Expr  "-" Expr1;
ExprMul.        Expr1           ::= Expr1 "*" Expr2;
ExprDiv.        Expr1           ::= Expr1 "/" Expr2;
ExprPow.        Expr2           ::= Expr3 "^" Expr2;
ExprInt.        Expr3           ::= Integer;
ExprDouble.     Expr3           ::= Double;
ExprVarRef.     Expr3           ::= Ident;
ExprFunCall.    Expr3           ::= Ident "(" [Expr] ")";

separator Statement ";";

StmtExpr.       Statement       ::= Expr;

separator VarMaybeInit ",";

VarInit.        VarMaybeInit    ::= Ident "=" Expr;
VarNoInit.      VarMaybeInit    ::= Ident;

HasGlobals.     Globals         ::= "globals" [VarMaybeInit] ";";
NoGlobals.      Globals         ::= ;

separator Ident ",";

HasExterns.     Externs         ::= "externs" [Ident];
NoExterns.      Externs         ::= ;

separator Function "";

FuncWithLocals. Function        ::=
    "function" Ident "(" [VarMaybeInit] ")"
        "locals" [VarMaybeInit]
    "do"
        [Statement]
    "end";

FuncNoLocals.   Function        ::=
    "function" Ident "(" [VarMaybeInit] ")" "do"
        [Statement]
    "end";

MkMain.         Main            ::=
    "main"
        [Statement]
    "end";

MkProgram.      Program         ::= Globals Externs [Function] Main

main in generated TestFoo.hs

main :: IO ()
main = do args <- getArgs
          case args of
            [] -> hGetContents stdin >>= run 2 p[Expr]
            "-s":fs -> mapM_ (runFile 0 p[Expr]) fs
            fs -> mapM_ (runFile 2 p[Expr]) fs

As you can see, the p[Expr] is breaking the program.

[dramforever]

I'm new to bnfc, so tell me if I'm using it wrong

[dramforever]

I'm beginning to understand it. BNFC is trying to start parsing Foo by calling pFoo. However, the Foo here is [Expr], so it's invalid. I still think this is a bug, nevertheless.

[gdetrez]

Unless told otherwise (using the entrypoints pragma) the test program will try to parse using the first category defined in your grammar (in your case, the first category is [Expr]).

Using a list category as an entry point is not supported in the current version. But I don't really see an immediate reason why that is, so it should probably be fixed.

A quick workaround is to create a dummy category to be the root of your abstract syntax tree. I.e. put this at the top of your grammar:

Foo. Bar ::= [Expr];

[dramforever]

Thanks for that. :smiley_cat:

[gdetrez]

Varsågod!