Force.com Labs Round Robin Record Assignment - Case Owner issue

There's an existing defect in the Salesforce package for Round Robin Record Assignment.  (http://appexchange.salesforce.com/reviews?listingId=a0N3000000178fsEAA).  The Force.com lab unmanaged package assigns Cases to an owner/queue when available.  The package contains two triggers which assign the case to an owner; caseRoundRobin.trigger and caseOwnerUpdate.trigger.

The two triggers inadvertently fight with eachother.  The fighting causes a Case to be re-assigned to it's original owner when it is first reassigned ownership. Say you have a case w/ owner John, and you want to reassign it to owner Mary.  In the Salesforce UI, after you save the ownership, it reverts back to John on the first attempt.  A second attempt will get you to the right owner, however.

The defect arises in caseOwnerUpdateTrigger.  After the code loops through the effected cases in the Trigger, via Trigger.new, it adds the cases to a temp list and then does a SOQL query to retrieve them once more via Id. The second loop does a query against uncommitted data and returns the yet unmodified owner of the Case, which would be John from the example above.  The case is then updated w/ the former Owner and appears the Case never changes hands.

The following change to the trigger should clear things up, consolidating everything into one loop, with no 'dirty', or uncommitted reads.

Two notes here: 1) Since the case in the loop is in process of being edited by the Trigger, it is locked and in read-only mode.  Therefore, a temp case is created and added to the update list.
2) The update call is made on the list of cases rather than on individual cases in the loop.  This is a Salesforce best practice, in order to mitigate SOQL queries and DML.

trigger caseOwnerUpdate on Case (after update) {
    List updateCS = new List();
    Map cases = new Map();
   
    for (Case cs : Trigger.new)
    {
        if(Trigger.isUpdate) { 
            System.debug('>>>>> Owner ID: '+cs.ownerId+' Temp Owner ID: '+cs.TempOwnerId__c);
            if(cs.TempOwnerId__c <> null && cs.TempOwnerId__c <> '') {
                if(cs.OwnerId <> cs.TempOwnerId__c) {
                    Case temp = new Case();
                    temp.OwnerId = cs.TempOwnerId__c;
                    temp.TempOwnerID__c = 'SKIP';
                    updateCS.add(temp);
                }
            }          
        }  
    }
 
    System.debug('>>>>>Update Cases: '+updateCS);
   
    //
    //Update last assignment for Assignment Group in batch
    //
    if (updateCS.size()>0) {
        try {
            update updateCS;
        } catch (Exception e){

        }
    }
}

Inheritance and Visualforce page controllers

Salesforce Apex is pretty finicky with inheritance in comparison to Java.
Even though it is Java based, there are many caveats, especially in inheritance and OO related principles.

In Visualforce, you have the option of denoting either a StandardController or Controller, in addition to controller extensions.  The extensions may or may not have code in common with the controller.  If they do, then you'll want to take advantage of inheritance and creating an abstract controller.

Below is a simple example of a Visualforce page utilizing an abstract controller.

Test.page:

<apex:page controller="Test" extensions="Test1,Test2,Test3">
<apex:form>
<apex:commandButton value="Go" action="{!go}" rerender="out" />
<apex:outputPanel id="out">{!output}</apex:form>
</apex:page>

Test.cls:

public abstract class Test {
  public String output {get; set;}
 
  private Test controller = null;
 
  public Test(){}
 
  public Test(Test controller){
      this.controller = controller;
  }
 
  public virtual void go(){
      output = 'Test';
  }
}

Test1.cls:

public class Test1 extends Test{
  public override void go(){
      this.output = 'Test1';
  }
  //Visual force needs this constructor to perform the controlling logic
  public Test1(Test test){ super(test); }
}

Test2.cls:

public class Test2 extends Test{
  public override void go(){
      this.output = 'Test2';
  }
  public Test2(Test test){super(test);}
}

Test3.cls:

public class Test3 extends Test{
  public override void go(){
      this.output = 'Test3';
  }
  public Test3(Test test){super(test);}
}

FYI, if you're reading through the Advanced developer study guide, 'Test1' is printed after clicking  the go command button.

ORA-12519

I created a polling adapter in my BPEL process.  Once I threw it on the server, it caused the Oracle XE (10.1.3?) database to crash.  After the database crashed, I could not view the BPEL console nor execute any commands on the database through SQL Developer.  The error I received was 'ORA-12519'.

What happens is that the XE database by default only allows a small number of connections.  BPEL consumes those connections rather quickly. If you encounter this problem, restart the XE database:

1.   Perform the following SQL as SYS user: ALTER SYSTEM SET PROCESSES=150 SCOPE=SPFILE
2. Restart the database
3. Restart BPEL / SOA

BPEL and TopLink - Method Accessors have not been selected

This error occurs when adding a DbAdapter in BPEL, having a master-detail or 1 to many relationship.  There's more information on this defect and a solution at this link: http://iadviseblog.wordpress.com/category/toplink/.  Search for 'Method accessors'.

BPEL - Importing schemas for Polling orchestration

By default, BPEL orchestrations are started with a client partner link with two activities, one for input and one for output. When creating a polling orchestration, or an orchestration that polls from a database rather then waits for an http request, the first task is to remove the partner link and the two activities. Unfortunately, this renders the client WSDL useless.

Without the client WSDL, there's no other way to import additional schemas into the BPEL orchestration. Or is there? When you create the polling activity in BPEL, it generates an XSD and a WSDL file for the Polling partner link. When you need to add additional schemas to your BPEL process, then add the following into your Polling WSDL. (If the types element already exists, then you can add your schema import there).



After saving the WSDL, be sure to add the declaration of the schema and namespace to your bpel source. Then you should be able to reference the schema in your orchestration.

Passing null values for xml elements

I've been working on a Salesforce webservice class to create cases. After generating the wsdl, I ran into a couple issues.

First issue: The order of the child elements in the method's parent need to match the webservice method you're calling. Found this out after changing some elements around and getting 'invalid format for xsd:dateTime' when it was a string or an Integer instead. This also means you need to specify all the elements as well. Which leaves me to my next issue.

Second issue: I kept receiving 'invalid type '' for xsd:dateTime when attempting to pass nulls into the element. ,, all return the same error. Instead, you have to declare the schema xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance and include xsi:nil="true" in your element. After I changed all my optional dateTime elements to , I was able to invoke my web service.

Now, why I have to include a separate schema just to include null values in my xml is beyond me.

- J

Oracle remote access to procedure

For the past few days, I've literally been stumped at how to allow a trigger to call a remote stored procedure. I created a trigger on a 9i server, calling a procedure on an 11g server. The trigger compilation failed due to an error regarding 'Cannot execute DDL statements on a remote server'.

There were already several triggers set up on the same 9i system calling the same procedure on the 11g server. I was in stumped mode.

Turns out, a public database link had already been created. This is a must in oracle for calling remote databases. The public database was created to connect as the 'aq_admin' user, meaning that the aq_admin user on the remote, 11g server, had to have permissions to execute the procedure.

The aq_admin user had priveleges, in fact it was the owner of the procedure. This stumped me even more.

The aha moment came when I took a look at the synonyms. The other triggers were in the SYSADM schema, and the trigger I was trying to compile was in the PRODADM schema. A synonym to the procedure package was created under the SYSADM schema, but not the PRODADM schema.

After creating the schema, wa-la everything clicked into place.
So here are the instructions (pseudo code) for connecting and executing a remote stored procedure in Oracle.

Create the remote procedure.
Create a (public) database link from the local server to the remote server.
On the remote system, make sure that the user the link is listed under, has privileges to the procedure. (Grant execute on procpackage to user).
Create a synonym under the schema where you need the trigger.
Create synonym "SCHEMA"."SYNONYM" on "SCHEMA"."TABLE"@"NAME_OF_DATABASE_LINK"
Compile the trigger.

And that should do it.