FWD » Core Development » Database

Ovidiu, while recreating a separate issue, I am getting many messages like this in the terminal:

Failed to detect the DYNAMIC-FUNCTION calls.

These are going to stderr when they really should be logged if they're important (I think they are), but aside from that...they are caused by this code in RuntimeJastInterpreter.prepare(JavaAst, boolean):

      if (dynamicEvaluation)
      {
         dynamicCalls = new HashMap<>();
         // scan for dynamic calls
         Iterator<Aast> it = jast.iterator();
         while (it.hasNext())
         {
            Aast next = it.next();
            if (DYNAMIC_CALL.equals(next.getText()))
            {
               dynamicCalls.put(next, NOT_EVALUATED);
            }
         }

         if (dynamicCalls.isEmpty())
         {
            System.err.println("Failed to detect the DYNAMIC-FUNCTION calls.");
         }
      }

DYNAMIC_CALL is defined as "ControlFlowOps.invokeDynamicFunction". This causes us to miss all of the other ControlFlowOps dynamic function method variants:

• ControlFlowOps.invokeDynamicFunctionIn
• ControlFlowOps.invokeDynamicFunctionWithMode
• ControlFlowOps.invokeDynamicFunctionInWithMode

Each of these variants is itself overloaded.

What I am actually seeing in the Java AST in this case is:

compilation unit [COMPILE_UNIT]:38654705665 @0:0
   com.goldencode.p2j.util.* [KW_IMPORT]:38654705667 @0:0
   com.goldencode.p2j.util.BlockManager.* [STATIC_IMPORT]:38654705668 @0:0
   DynGenQuery [KW_CLASS]:38654705669 @0:0
       [CS_INSTANCE_VARS]:38654705673 @0:0
         = [ASSIGN]:38654705717 @0:0
            AdaptiveQuery [REFERENCE_DEF]:38654705718 @0:0
            AdaptiveQuery [CONSTRUCTOR]:38654705719 @0:0
       [CS_INSTANCE_METHODS]:38654705676 @0:0
         execute [METHOD_DEF]:38654705678 @0:0
             [BLOCK]:38654705679 @0:0
               initialize [METHOD_CALL]:38654705722 @0:0
                  query0 [REFERENCE]:38654705723 @0:0
                  tt [REFERENCE]:38654705724 @0:0
                  ? [STRING]:38654705725 @0:0
                   [NULL_LITERAL]:38654705726 @0:0
                  tt.recid asc [STRING]:38654705727 @0:0
                   [EXPRESSION]:38654705728 @0:0
                     new Object [REFERENCE_DEF]:38654705729 @0:0
                      [INITIALIZER]:38654705730 @0:0
                        P2JQuery.Parameter [CAST]:38654705731 @0:0
                            [LAMBDA]:38654705732 @0:0
                               [LPARENS]:38654705733 @0:0
                              isEqual [STATIC_METHOD_CALL]:38654705735 @0:0
                                 logical [CONSTRUCTOR]:38654705736 @0:0
                                    ControlFlowOps.invokeDynamicFunctionIn [STATIC_METHOD_CALL]:38654705737 @0:0
                                       hqlParam1dq [REFERENCE]:38654705738 @0:0
                                       thisProcedure [STATIC_METHOD_CALL]:38654705739 @0:0
                                       hqlParam2dq [REFERENCE]:38654705740 @0:0
                                 hqlParam3dq [REFERENCE]:38654705741 @0:0
   com.goldencode.p2j.persist.* [KW_IMPORT]:38654705705 @0:0
   com.goldencode.testcases.dmo._temp.* [KW_IMPORT]:38654705706 @0:0
   com.goldencode.p2j.util.CompareOps.* [STATIC_IMPORT]:38654705742 @0:0
   com.goldencode.p2j.util.ProcedureManager.* [STATIC_IMPORT]:38654705743 @0:0
   com.goldencode.p2j.persist.lock.* [KW_IMPORT]:38654705744 @0:0

Note the use of invokeDynamicFunctionIn instead of invokeDynamicFunction.

Is it safe for me to modify the DYNAMIC_CALL condition to check for any of these variants? My inclination is yes, since it appears all the parameters in the interpreter are managed dynamically. But I wanted to get your advice before making this assumption and modifying the code.

Eric Faulhaber wrote:

Is it safe for me to modify the DYNAMIC_CALL condition to check for any of these variants? My inclination is yes, since it appears all the parameters in the interpreter are managed dynamically. But I wanted to get your advice before making this assumption and modifying the code.

Yes, it should be. I do not see any reason it would not work.
You may want to replace the if (DYNAMIC_CALL.equals(next.getText())) line with if (next.getText().startsWith(DYNAMIC_CALL)) in RuntimeJastInterpreter:332 and boolean dynamicCall = (dynamicCalls != null) && DYNAMIC_CALL.equals(methodName); line with boolean dynamicCall = (dynamicCalls != null) && methodName.startsWith(DYNAMIC_CALL); at RuntimeJastInterpreter:1141. And since DYNAMIC_CALL is now a prefix marker instead of the actual marker, we probably should rename it accordingly.