Lookahead woes and more about packcc

[dolik-rce]

I don't think it is a problem related to lookahead...

The a[777].foo correctly matches FunctionCall rule, but it doesn't match Indexing. That is partially because Indexing tries to first match only index access, and it never gets to FunctionCall. Switching the order will help: Indexing <- FunctionCall / Symbol _ "[" _ Expr _ "]". But this is not the only problem, which makes it much harder to debug, because even if you fix one problem, it still doesn't work 🙂

I'd suggest to rewrite the grammar slightly to separate the index access from field access. Currently they're kind of intertwined (HasField contains Indexing). I can't tell you how exactly, I'd have to give it much more time than I have available, sorry.

[joagre]

Thanks for the input. I can see now that my grammar snippet is ambiguous/faulty and need to be rewritten. From a completelly naively point of view I would like to have an even more ambiguous grammar though:

Program            <- _ TopLevelExpr (_ "," _ TopLevelExpr)* EOF
TopLevelExpr       <- Expr / Binding

# Precedence ordered
Expr               <- Add
Add                <- Multiplicate (_ "+" _  Multiplicate)*
Multiplicate       <- Indexing (_ "*" _  Indexing)*
Indexing           <- Expr _ "[" _ Expr _ "]"  / FunctionCall
FunctionCall       <- FunctionName _ "(" _ ExprSequence? _ ")" / FieldAccess
FieldAccess        <- Expr (_ "." _ Expr)* / Primary

FunctionName       <- Symbol
Symbol             <- [a-zA-Z_][a-zA-Z_0-9_]*
ExprSequence       <- Expr (_ "," _ Expr)*

Primary            <- Literal / Symbol
Literal            <- NumberLiteral
NumberLiteral      <- [0-9]+

Binding            <- MatchPattern _ "=" _ Expr
MatchPattern       <- Literal / FieldAccess / Symbol

_                  <- WS*
WS                 <- [ \t\r\n]
EOF                <- _ !.

:-)

It would require a parser with indefinite backtracking. Maybe not even that would suffice. Life is hard.

Thanks