Python is not context free.
The parser is context free-- LL(1), actually.
The lexer maintains one extra stack to track indentations, meaning it's more of a PDA than a DFA, but the actual complication is mild.
That alone does not make it a nice language, of course.
I can only assume that it's lying its pants off, as I noted in an other subthread `(a) - (b)` can't be parsed (to an AST) without knowing the identity of `a` in the current scope: if it's a type the expression is a cast of `-(b)` to `a`, if it's a value it's a subtraction of `(b)` from `(a)`.
By "it can't be parsed to an AST" I meant "it can't be parsed to a useful AST used to run or generate code" on its own, it needs to be disambiguated through contextual information before anything can be done with the prospective subtree. Sorry for the lack of clarity.
> It's called an ambiguous grammar, which can still be context free.
The only way to disambiguate is to provide expression context (namely visible name bindings at this point), it's essentially the same problem Eli Bendersky described when talking about C grammar's context-sensitivity: http://eli.thegreenplace.net/2007/11/24/the-context-sensitiv...