First VyperNode output

vyper_ast.py

class VyperNode:

    def __init__(source_code, pos):
        self.source_code = source_code
        self.lineno, self.col_offset = pos


class Name(VyperNode):
    pass


class Subscript(VyperNode):
    pass


class Index(VyperNode):
    pass


class arg(VyperNode):
    pass


class Tuple(VyperNode):
    pass


class FunctionDef(VyperNode):
    pass


class arguments(VyperNode):
    pass


class Import(VyperNode):
    pass


class Call(VyperNode):
    pass


class Str(VyperNode):
    pass


class Compare(VyperNode):
    pass


class Num(VyperNode):
    pass


class NameConstant(VyperNode):
    pass


class Attribute(VyperNode):
    pass


class BinOp(VyperNode):
    pass


class BoolOp(VyperNode):
    pass


class UnaryOp(VyperNode):
    pass


class List(VyperNode):
    pass


class Dict(VyperNode):
    pass


class Bytes(VyperNode):
    pass


class Add(VyperNode):
    pass


class Sub(VyperNode):
    pass


class Mult(VyperNode):
    pass


class Div(VyperNode):
    pass


class Mod(VyperNode):
    pass


class Pow(VyperNode):
    pass


class In(VyperNode):
    pass


class Gt(VyperNode):
    pass


class GtE(VyperNode):
    pass


class LtE(VyperNode):
    pass


class Lt(VyperNode):
    pass


class Eq(VyperNode):
    pass


class NotEq(VyperNode):
    pass


class And(VyperNode):
    pass


class Or(VyperNode):
    pass


class Not(VyperNode):
    pass


class USub(VyperNode):
    pass


class Expr(VyperNode):
    pass


class Pass(VyperNode):
    pass


class AnnAssign(VyperNode):
    pass


class Assign(VyperNode):
    pass


class If(VyperNode):
    pass


class Assert(VyperNode):
    pass


class For(VyperNode):
    pass


class AugAssign(VyperNode):
    pass


class Break(VyperNode):
    pass


class Continue(VyperNode):
    pass


class Return(VyperNode):
    pass


class Delete(VyperNode):
    pass


class stmt(VyperNode):
    pass


class ClassDef(VyperNode):
    pass


class ImportFrom(VyperNode):
    pass

Suggestions:

Add an id field so that a program can refer to an AST node by its id.

If you define a string srcmap analogous to how solc does it, that would help interoperability of tools (e.g. static analysis tools, linters, tracers). It is documented here.

There is an implicit compression that can be generalized: if a length is 1 it can be omitted. If there is only one "file" or (container) it would be 0 and that would be omitted. So then the only change is from line/col to byte offset, a single (string) value.

Also note that when you get down to bytecode or ewasm. The sourcemap idea can be used again and it allows for a construct to refer to mulitiple non-contiguous regions of llvm/evm bytecode/ewasm.

On second thought, where the id field is most important is in the JSON representation. In addition to the id field, there would be the class names would appear as a field in "nodeType" (as it does in solc's ast) And there would be a "nodes" field which contains all of the children nodes.