A TYPECASE statement has the form:
TYPECASE Expr OF
T_1 (v_1) => S_1
| ...
| T_n (v_n) => S_n
ELSE S_0
END
where Expr is an expression whose type is a reference type,
the S's are statements, the T's are reference types,
and the v's are identifiers. It is a static error if
Expr has type ADDRESS or if any T is not a subtype of
the type of Expr. The "ELSE S_0"
and each "(v)" are optional.
The statement evaluates Expr. If the resulting reference value is a member of any listed type T_i, then S_i is executed, for the minimum such i. (Thus a NULL case is useful only if it comes first.) If the value is a member of no listed type and "ELSE S_0" is present, then it is executed. If the value is a member of no listed type and "ELSE S_0" is absent, a checked runtime error occurs.
Each (v_i) declares a variable whose type is T_i and whose scope is S_i. If v_i is present, it is initialized to the value of Expr before S_i is executed.
If (v_i) is absent, then T_i can be a list of type expressions separated by commas, as shorthand for a list in which the rest of the branch is repeated for each type expression. That is:
T_1, ..., T_n => S
is shorthand for:
T_1 => S | ... | T_n => S
For example:
PROCEDURE ToText(r: REFANY): TEXT =
(* Assume r = NIL or r^ is a BOOLEAN or INTEGER. *)
BEGIN
TYPECASE r OF
NULL => RETURN "NIL"
| REF BOOLEAN (rb) => RETURN Fmt.Bool(rb^)
| REF INTEGER (ri) => RETURN Fmt.Int(ri^)
END
END ToText;