Star [email protected]

Two tasks are performed here. As in the calc example, AST nodes are constructed. At the same time, attributes of the node, such as the type, are computed. Why is this possible?

The semantic analyzer can fall back on two different sets of attributes. The scope is inherited from the caller. The type declaration can be computed (or synthesized) by evaluating the name of the type declaration. The language is designed in such a way that these two sets of attributes are sufficient to compute all attributes of the AST node.

An important aspect is the declare-before-use model. If a language allows the use of names before declaration, such as members inside a class in C++, then it is not possible to compute all attributes of an AST node at once. In such a case, the AST node must be constructed with only partially computed attributes or just with plain information (such as in the calc example).