JCgH4164838Gh792C124B5 commented on a change in pull request #406:
URL: https://github.com/apache/struts/pull/406#discussion_r415353423



##########
File path: plugins/json/src/main/java/org/apache/struts2/json/JSONResult.java
##########
@@ -212,12 +211,22 @@ protected Object readRootObject(ActionInvocation 
invocation) {
     }
 
     protected Object findRootObject(ActionInvocation invocation) {
+        ValueStack stack = invocation.getStack();

Review comment:
       The new logic seems to make sense, and the updated comment should make 
it clear what the behaviour will be in 2.6.
   
   There might be some users with existing implementations depending on the 
current 2.5.x behaviour (always `peek()` if root is `null`), but they could 
implement a custom `JSONResult` with the old behaviour if they really needed it.

##########
File path: plugins/json/src/main/java/org/apache/struts2/json/JSONResult.java
##########
@@ -286,7 +294,9 @@ public String getRoot() {
     }
 
     /**
-     * Sets the root object to be serialized, defaults to the Action
+     * Sets the root object to be serialized, defaults to the Action.
+     * If the Action implements {@link ModelDriven}, model will be used instead
+     * and assumptions is the Model was pushed on the top of the stack

Review comment:
       The last comment might be a bit clearer as:
   ```
        * Sets the root object to be serialized, defaults to the Action.
        * If the Action implements {@link ModelDriven}, the Model will be used 
instead,
        * with the logic assuming the Model was pushed onto the top of the 
stack.
   ```

##########
File path: plugins/json/src/main/java/org/apache/struts2/json/JSONResult.java
##########
@@ -68,8 +69,6 @@
  */
 public class JSONResult implements Result {
 
-    private static final long serialVersionUID = 8624350183189931165L;

Review comment:
       Since `JSONResult` still extends `Serializable`, maybe the 
`serialVersionUID` should be kept ?
   
   By removing it, anyone who Java-serialized instances in 2.5.x (possible) and 
then tries to Java-deserialize them in 2.6 could experience deserialization 
failures.
   
   If this change is intentional because of the logic change for 2.6, maybe a 
brand new `serialVersionUID` value should be generated instead of just dropping 
it ?

##########
File path: plugins/json/src/main/java/org/apache/struts2/json/JSONResult.java
##########
@@ -212,12 +211,22 @@ protected Object readRootObject(ActionInvocation 
invocation) {
     }
 
     protected Object findRootObject(ActionInvocation invocation) {
+        ValueStack stack = invocation.getStack();
         Object rootObject;
         if (this.root != null) {
-            ValueStack stack = invocation.getStack();
+            LOG.debug("Root was defined as [{}], searching stack for it", 
this.root);
             rootObject = stack.findValue(root);
         } else {
-            rootObject = invocation.getStack().peek(); // model overrides 
action
+            LOG.debug("Root was not defined, searching for #action");
+            rootObject = stack.findValue("#action");
+            if (rootObject instanceof ModelDriven) {
+                LOG.debug("Action is an instance of ModelDriven, assuming 
model is on the top of the stack and using it");
+                rootObject = stack.peek();
+            }
+            if (rootObject == null) {

Review comment:
       Maybe the 2nd if check should be changed to:
   ```
               else if (rootObject == null) {
   ```
   because if the rootObject is a `ModelDriven` and somehow the `stack.peek()` 
returns`null`, the PR's logic would attempt the same operation again (along 
with an extra debug output).
   




----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
[email protected]


Reply via email to