Symbol Tables and Declaration Processing - Crafting a Compiler - page 283

procedure build S ymbol T able ()

call process N ode ( ASTroot )

end

procedure process N ode ( node )

switch ( ( node ) )

case Block

kind

call symtab . openScope()
case Dcl
call symtab . enterSymbol( node . name , node . type )
case Re f
sym ← symtab . retrieveSymbol( node . name )
if sym =

1

null
then call

, sym )
foreach c ∈ node . getChildren() do call processNode( c )
if

error (” Undeclared symbol :”

kind ( node ) = Block

then

call symtab . closeScope()

2

end

Figure 8.2: Building the symbol table

enter ( name , type ) enters name in the symbol table's current scope.

The parameter type conveys the data type and access attributes of name 's
declaration.

S ymbol

retrieve S ymbol ( name ) returns the symbol table's currently validdeclaration

for name . If no declaration for name is currently in e ff ect, then a null

pointer is returned.

declared ( name ) tests whether name is present in the symbol table's

current (innermost) scope. If it is, true is returned. If name is in an outer
scope, or is not in the symbol table at all, false is returned.

L ocally

To illustrate the use of this interface, Figure 8.2 contains code to build the
symbol table for the AST shown in Figure 8.1. The code is specialized to the
type of the AST node. Actions may be performed both before and after a
given node's children are visited. Prior to visiting the children of a Block node,
code at Marker 1 increases opens a new scope. After the subtree of the Block
is processed, Marker 2 abandons the scope. The code for Re f retrieves the
symbol's current definition in the symbol table. If none exists, then an error
message is issued.