1.3.3 Reusability enhancement

From time to time we want to migrate from an existing language to another and reuse as much as possible. In case when the two languages share exactly the same semantic features but differ syntactically, the only thing we have to do is modify the SDF description. This is because in all the modules which defines the original language, only the SDF description depends on the concrete syntax. Other modules, of course, must conform to a certain syntax, but that syntax is defined in action semantics and is thus universal.

For instance, by changing the action notation example in the last section to:

ActionPrefix "(" Actions ")" $\to$ Action {left}

Action $\to$ Actions

Action "," Actions $\to$ Actions

"actlist" $\to$ ActionPrefix

we completely change the action syntax. actlist(A1), actlist(A1, A2, $\cdots$, An), actlist(A1, actlist(A2, A3), A4, actlist(A5)) are now acceptable. (Supposing A1, A2, $\cdots$, An are actions.)

An other example is to change the data notation in the last section to the following:

Data Data DataOpPostfix $\to$ Data

"+"|"-"|"*" $\to$ DataOpPostfix

to let the language accept postfix expressions. Note that because a postfix expression has explicit evaluation ordering and is never ambiguous, we can specify {left} for the first statement, or {right}, or just ignore this option.

In reality, when the language is modified, it is impossible to keep all the other modules untouched and just modify the SDF description. Fortunately, our reason for introducing the SDF level is not to avoid changes in the language description, but to minimize them.

Without this additional level, in practice people often find rewriting the whole language description more reasonable than modifying the existing one, because it is too difficult to find out all the lexical symbols scattering in the whole description and modify them correctly. Maintainability and readability must also be considered.