Statement list conventions

Every statement contains some basic information about it (which really lives in the Statement class). Specific types of Statements are specialized by deriving them from the Statement class. The Statement class contains the type of the statement, pointers to the lexically next and previous statements, the original line number, a pointer to the immediately outer loop (if any) in which this statement is found, the set of expressions read, the set of expressions written, the set of predecessor and successor statements, the assertions associated with the statement, etc.

The Polaris base keeps the next, prev, predecessors, successors, follow, lead, outer, in_refs, out_refs, and act_refs incrementally correct throughout the compilation process. Th only exception to this is for the in_refs, out_refs, and act_refs when using the assign/pull operations. After those are done, the user must re-build the _refs explicitly.

• FLOWENTRY/FLOWEXIT stmts There is a single FLOWENTRY statement at the beginning of the StmtList in every ProgramUnit. Likewise, there is a single FLOWEXIT statement at the end of the StmtList in each ProgramUnit. This is even true for BLOCKDATA ProgramUnits (and they are the only statements in such a StmtList).

  The FLOWENTRY statement has all ENTRY statements in the ProgramUnit in its successor set. The FLOWEXIT has all STOP and RETURN statements in the ProgramUnit in its predecessor set.

• IF block structure (logical IF, arithmetic IF The IF statement always has an ENDIF statement associated with it. The IF statement contains a follow() member function which points at the associated ENDIF statement. Likewise, the lead() of the ENDIF is always pointed at the IF. The follow() of the ENDIF is NULL.

  There is always a then-part associated with the IF statement. The then-part always contains an IMPLIED GOTO statement just prior to the ENDIF statement. If an ELSE statement exists within the IF construct, then there will also be an IMPLIED GOTO statement placed just prior to the ELSE statement.

  The IMPLIED GOTO statements exist to represent the implicit branch which interrupts the control flow at those two statements when the program executes. They are merely markers and contain no useful information. The flow of control change is reflected in the next_ref() and prev_ref() member functions for those statements.

• DO loop structure The DO statement always has a matching ENDDO. The follow() of the DO points at the ENDDO and the follow() of the ENDDO likewise points at the DO.