Branescript syntax does not support array indexing on identifiers.

[DanielVoogsgerd]

The documentation says Branescript allows array indexing in C style manner: foo[0].

However, looking at the formal definition, that seems to be not the case.

Expressions are defined as:

 Expr ::= LPAREN Expr RPAREN
        | Expr BinOp Expr
        | UnaOp Expr
        | Array | ArrayIndex | Call | IDENT | Instance | Literal
        | Projection

This does contain Array and ArrayIndex, but array is a definition of an array, not the usage of an identifier (variable) that points to / contains an array.

ArrayIndex suffers the same fate as it is prefixed with: Array itself: ArrayIndex ::= Array LBRACKET Expr RBRACKET

Proposed solutions:

We either have to change identifiers such that identifiers can be indexed by default, or we have to change ArrayIndex to be:

ArrayIndex ::= Array LBRACKET Expr RBRACKET | IDENT LBRACKET Expr RBRACKET\

The latter seems better to me in this case. @Lut99 Let me know what you think.

[Lut99]

Good catch! You're really diving into the BraneScript grammar with no holds barred xD that's amazing 😁

Regarding this fix; it's every more complicated as projections are also allowed on the lefthand-side of array indices (e.g., foo.data[0]). So I think that, grammar-wise, any expression is allowed; it just depends on types whether the array index actually makes sense on that particular expression. But let me double-check that in the compiler and get back to you.

[DanielVoogsgerd]

This is all part of a personal side project. I really wanted to implement a treesitter grammar at some point, and I don't come across unimplemented languages often 😅. The advantage for Brane is that if the BNF is incorrect, my syntax highlighting is broken, so the issues lit up like a Christmas tree.

Yeah, that sounds reasonable. We should only be careful with precedence in "open" expressions like 1+1 [2] which should be parsed as 1+(1[2]) I think.

[Lut99]

Indeed. After lunch I'll see what the compiler currently does and whether that needs fixing (precedence should be a really easy fix)

[Lut99]

Precedence checks out (note it's higher than any binary expression). Further, while quite messy, the BraneScript parser considers array indices to follow any other expression.

I'll push a PR with the correct syntax in a sec. Indeed, we can remove ArrayIndex and instead treat the index as a special kind of unary operator.