Semantic Analysis - Crafting a Compiler - page 357

LabeledStmt
stmtLabel stmt

Stmt

Figure 9.10: Abstract Syntax Tree for a Labeled Statement

procedure visit ( LabeledStmt ls )

newNode ← new LabelList( ls . stmtLabel ,get K ind ( ls . stmt ) , ls . stmt )

newList ← cons ( newNode ,get L abel L ist ())

call set L abel L ist ( newList )

call visit C hildren ( ls )

call set L abel L ist ( tail ( get L abel L ist ()))

end

Figure 9.11: Semantic Analysis for a Labeled Statement

If the statement is labeled, stmtLabel contains the label in string form. If the
statement is unlabeled, stmtLabel is null. LabeledStmt also contains a field
stmt which is the AST node representing the labeled statement. Unlabeled
statements need not have LabeledStmt as a parent, particularly in contexts
where a label is disallowed.
In Java, C, and C ++ (and most other programming languages), labels are

placed in a di erent name space than other identifiers. This means that an

identifier used as a label may also be used for other purposes (a variable name,
a type name, a method name, and so on) without confusion. This is because
labels are used in very limited contexts (in continues, breaks, and perhaps
gotos). Labels cannot be assigned to variables, returned by functions, read
from files, etc.
We will maintain a list of labels that are currently visible to the AST
being analyzed. The function

ff

get L abel L ist will return the current label list

(possibly null). The procedure will set the current label list via its

parameter. The label list is set to null when we begin the analysis of a method,
constructor body, or a static initializer.
Each LabelList node contains three fields. The first is label (a string that
contains the name of the label). Next is kind (one of iterative , switch ,or other
that indicates the kind of statement that is labeled). The last is AST (a link to

set L abel L ist