I could be mistaken, but I think there's a fundamental distinction between the computing power required to recognize a language and the kinds of functions a language can compute. AFAIK (which isn't much), usually the distinction between regular, context-free, context-sensitive and recursively-enumerable languages turns on the power of the recognizer rather than the power of the language to express functions. In other words, the languages are limited in their syntactic expressivity but not their semantic expressivity. I've always understood the latter to be the key to determining the computational power. Take lambda calculus as an example. It is Turing complete, but it is amenable to a context-free syntactic implementation (see http://www.soe.ucsc.edu/classes/cmps112/Spring03/readings/la...). For C++, I always thought it was that not only is it capable of expressing semantically any computable function, but that a [EDIT: Turing Complete] parser was required to parse C++ strings.
You are right that the power necessary to parse a language is completely different from the power of the language itself.
There is however a different correspondence between recognizing things and computing functions. If we encode a function as taking a bit string as input and generating a bit string as output, we can turn it into a series of recognition problems as follows: "recognize given S as input, whether bit i of f(S) is 1".
Given all these recognizers, we can compute the function, and given a method to compute the function, we can implement the recognizers.
That is, you can reformulate the question "is f computable" to a series of language recognition problems.
