13.3 Evaluating nulled symbols

In theory, the semantics of nulled symbols, like any semantics, can be arbitrarily complex. In practice, we are dealing with the semantics of the empty string, which is literally the “semantics of nothing”. If what we are dealing with truly is primarily a parsing problem, we can usually expect that the semantics of nothing will be simple.

The possible subtrees below a nulled symbol can be seen as a set, and that set is a constant that depends on the grammar. Since the input corresponding to the nulled symbol is also a constant (the empty string), the semantics of a nulled symbol will also be constant, unless one of the following is true:

• The semantics is non-deterministic.
• The semantics takes into account, not just the input, but its context. For example, the semantics might depend on parse location, in which case the nulled symbol instance of x at location 5 might mean something different from an nulled instance of x at location 50.

Both of these exceptions are unusual. When they do occur, the upper layer can implement the semantics of the nulled symbols with a function or a closure.