Way in the @future - Problem of using Salesforce as Real-time Relay System

Problem

For one of the project I was on, we had following as one of the requirements.  In this requirement, the System 1 would send a record to Salesforce, which would store the record, transform it based on the data stored in Salesforce and transmit the transformed record to System 2.  The whole process from System 1 to System 2 need to take less than 5 minute to complete.

Originally, we had implemented a trigger on the creation of the original record - Record A - that would transform it and send the transformation - Record A" - to System 2 by calling a web service through a future method.

This has been working as expected for quite sometime.  However, the production org was on NA14, and when the NA14 performance degradation occurred last year, we were witnessing that this request was taking more than half hours, and sometimes more than several hours.

Our investigation revealed that the problem was that the request to the future method gets stuck in the Apex Job Queue due to the degradation issue.  We had investigated both the batch apex and outbound-message workflow, but the problem was that both batch Apex and outbound-message jobs are queued the same way as the future method was.


Workaround

The way we addressed this problem was by implementing a custom web service.  This web service stored the Record A into Salesforce as well as transform the record the same way the trigger was doing.  We had proposed two solutions on how to transmit the transformed record to system 2.  One was for the custom web service make a callout; the second was to return the transformed record to System 1 as the response to the web service call, and System 1 would transmit it to System 2.  Our test revealed that both methods eliminated the queueing issue; however, the client went with the second option given the importance of this process.

Takeaway

The takeaway for me was that Salesforce really isn't a ideal tool for implementing a real-time processing of data, especially if you are using one of asynchronous methods (@future method, batch apex, etc.) as Salesforce do not provide any SLA on the timeliness of these processes.

Curious Case of Disappearing ID

Problem

Last week, I was deploying some changes to the recently deployed Apex classes into the production, but somehow it was failing on previously migrated test classes that were not related to the changes being deployed. To confirm that it had something to do with production environment, rather than the new code, I ran those pre-existing test classes in the production environment, and they were indeed failing even without the changes (I also had them run in Sandbox and confirm that they passed).  My initial suspicion was that someone had added something like a validation rule on the sObjects for which I was generating test data.  Although it was related to the test records being generated, the problem, it turned out, was not related to the validation rule, but something rather more strange.

The test classes had a structure similar to the following where I had a static method to generate test data for all the test methods and a test method that queries for the test data.

@isTest
private class TestCustomController {

  static void createBackground() {

    List<Account> testAccts = new List<Account> {
      new Account(Name = 'Test Account 1'),
      new Account(Name = 'Test Account 2')};
    Database.insert(testAccts);

    List<Account_Child__c> children = new List<Account_Child__c> {
      new Account_Child__c(Name = 'Child 1', Account__c = testAccts[0].id)};
   Database.insert(children);

  }

  static testMethod void manipulateCustomController() {

    TestCustomController.createBackground();

    Account queriedAcct = [SELECT Id FROM Account WHERE Name = 'Test Account 1'];

    // Initialize custom controller, pass on the account ID, 
    //and call method to query for the children of the accounts.

  }

}

To debug the problem in the production, I had put breakpoints in the createBackground() method after the insertion of the account and in the manipulateCustomController() method after query for the account through Developer Console and ran this test class.  What I found out was that the ID of the Test Account 1 in the createBackground() didn't match the ID in the manipulateCustomController()!

Workaround

I tried moving the code in the createBackground() method into the manipulateCustomController() method, and that did resulted in the ID being consistent through out the run of that test method.  However, this did not completely solve the test failure as I had another weird issue where the symbol table not showing some properties of an in-memory instance of a sObject, while it showed in the heap.

Takeaway

This does demonstrate that there are major differences in the internal implementation of the Test and "Production" environments of the Salesforce, mainly that the Test environment is somewhat unstable and has weird kink in it.  Also, it does demonstrate how valuable breakpoints can be when debugging problem in the Production org. 

Incrementing Instance Variable of Virtual Class

Problem

I had following piece of code that increments the instance variable in the parent class within the for-loop of the child class:

public virtual class ParentClass {
  public Integer mapIndex{get; set;}
}
public class ClassWithError extends ParentClass {
  


  public ClassWithError {
    Integer mapSize;
    ...
    for (mapIndex = 0; mapIndex < mapSize; mapIndex++) {
      ...
    }
    ... 
  }
  ...

}

which resulted in following error:

Unknown error: virtual class data member of the parent in for-loop

Solution

The fix is to simply create a temporary variable to do the increment and assign the value of the temporary variable to the instance variable inside the for-loop:

public virtual class ParentClass {
  public Integer mapIndex{get; set;}
}
public class ClassWithError extends ParentClass {
  


  public ClassWithError {
    Integer mapSize;
    ...
    mapIndex = 0;
    for (Integer tempIndex = 0; tempIndex < mapSize; tempIndex ++) {
      mapIndex = tempIndex;
      ...
    }
    ... 
  }
  ...

}

Do Not Expect - Trouble with HTTP

This is a bit of old issue, but I making this post as it does drive home an important lesson regarding Salesforce.com custom code development.  For one of our customer, there was this legacy code that used HTTP Callouts, which has been working for several years.  However, around the time of Winter '14 Release, the users started reporting that that piece of code was not working.

After digging through it, I was able to narrow it down the problem to following error on response from the callout to the external web service - a standard non-descriptive error message that appears for pretty much all error with HTTP callout:

Read Timeout Error
Unexpected End of Element...

Because the way Apex parses the response,  I couldn't find out much more from the debug log.  As a next step, I tried to find out if the issue was related to how Apex handles the Callout or if it was something wrong with the web service itself by duplicating the request that the Apex code is suppose to generate within SoapUI.  However, from SoapUI, we were getting the response that we expect to get, leading us to go back to the Salesforce side of things.

This time around, what I did to find the issue in the Callout was to strip down the Callout message to bare bones to see if it would work and then adding the elements back to the message one at a time.  The bare bone message did indeed work from Salesforce, and it turned out that problem was caused by following header:

Expect: 100-Continue

The fix for this problem was simply removing that header.  I don't know for sure if this issue was introduced with Winter '14 Release or through the changes to the external web service (btw, this limitation is now documented here).  However, this does provide a reminder on the dangers of using custom code in terms of its fragility when it comes to upgrade either to Salesforce or external service it interfaces.

Introducing Lightning Components - Salesforce Front-end Framework: First Thoughts and Gotchas

One of the biggest announcements from Dreamforce '15 was the new Lightning user interface. For us developers, the part that would be of most interest to us is Lightning Framework – the new component-based Javascript Front-end web framework for Salesforce.com. From going through the Developer's Guide for the Lightning Components, here are my first thoughts and some gotchas I found.

The Lightning Components is intended to allow developing responsive-design single-page app.  This should allow the developers to develop not only a Salesforce app that is more responsive than with Visualforce, but it should also allow us build one app that would optimize based on the client, instead of building separate app for Desktop and Mobile devices.

In terms of the structure, I found it to be quite similar MV*-based front-end frameworks such as AngularJS, Backbone.js or Ionic Framework, and if you are familiar with any of those frameworks, you shouldn't have much difficulty with it. The basic structure is that you create an app, which consists of one or more components with each component representing a section of the page. Each component then can have its own stylesheet, controller/helper scripts, and events.

Some of the main benefits I see is that because it is based on the Javascript, it should make it easier to implement custom Javascript components, such as auto-complete input box.  If you have had experience of taking over the input field with custom code in Visualforce page, you know how nightmarish of task it can become, dealing with Visualforce-specific convention regarding IDs and trying to resolve conflict between Visualforce tag and your Javascript code. Also, the UI testing and testing automation should be easier with this framework as you don't have to deal with auto-generated mark-ups from compiled Visualforce code and aforementioned ID conventions.  I would also like to mention Live Preview support in regards to testing, which you can open from Developer Console.  Finally, this will make managing your CSS stylesheet much easier as you have a separate CSS file associated with each component, instead of having to deal with either static resources or embedded CSS.

As for drawbacks, I did find a few potential issues.  One such issue may be that you have to customize the look-and-feel of every component with your own stylesheet.  One advantage of Visualforce is that it gives you a consistent look-and-feel to the Salesforce platform by default, which is lost with custom stylesheet.  Another issue I found was the file structure.  With the component-based structure, it appears to have completely different file-system structure to the rest of the Apex and Visualforce code directory structure we are familiar with.  As far as I know, you can only edit Lightning Components code from Developer Console, but I would assume it will become available in Force.com IDE and Mavensmate, and I wonder how those IDE/editros will be able to handle this rather diverging structure.

Finally, I would like to close this post with some gotchas I found while going through the Developer's Guide:

•  You can’t use Force.com Canvas apps in Salesforce1 if you enable Lightning components.
•  I think this may be the biggest issue for people looking into enabling Lightning components.
• In order to use Lightning Components, you need to register separate namespace for it.
• Some HTML tags are not supported with Lightning Components.
• Please see https://developer.salesforce.com/docs/atlas.en-us.lightning.meta/lightning/ref_supported_html_tags.htm for the list of the supported components.
• You need to create a custom tab in order to enable your component in the Salesforce1 Navigation menu.  
• Also, accessing your Lightning Component from the full Salesforce site is not supported - another major issue that I would think would be resolved before going into GA.

A Developer's Guide to Dreamforce

Another Dreamforce is upon us, and, although I won't be attending this year, I thought this is an opportune time to discuss how to use your time while at Dreamforce based on my experience from last year.

The first advice I would give to any developer attending Dreamforce for the first time would be don't try to attend all the sessions.  Rather, I would minimize the time spend on attending sessions, but try to maximize your time at the Developer Zone.  At the Developer Zone, you will have a chance to get some hands-on experience with new products and emerging technologies.  If it is anything like last year, you will also be able to drop in and listen to a number sessions that are held at the open spaces, and you can, of course, participate on the mini hackathons, even if you are not in the $1 million main hackathon.

As for the actual sessions, I would focus on attend sessions that are presented by either a developer from Salesforce or developers well-known within the community.  Googling the presenters prior to attending sessions could help you decide whether you would want to attend the given session.  I would avoid any session that is specific to a product or a panel discussion as I found most of those were more of promotional rather than educational.  One exception to this is the open developer Q&A session at the end of the conference.  Some of the sessions that I found to be on the better presented include:

• Testing and Testability
• Javascript Debugging

Another fun aspect of Dreamforce is the numerous After-party.  These are great opportunities to social and get some great free food and drinks.  However, it's pretty exclusive, so try to do some research and sign up for those as soon as possible.

Finally, if this is your first time in the Bay area, I would take the weekend and do some sightseeing.  I heard the tour to Alcatraz and Segway tour are really fun, although I didn't get to do those last year.  Also, renting bike and crossing the Golden Gate bridge is great as well.

Anyway,   if this your first time at Dreamforce, I'm sure you will have a great time, and you'll learn thing or two about Salesforce as well.

Validation, Approval Process and Testing

Two of the common sources of headache for Salesforce developer is approval process and validation.  Because these are declarative aspect of Salesforce (i.e. configured through the Salesforce user interface) with some restriction, these can be difficult to manager and migrate

With it being now migratable, many of the pain points for developing approval process in the sandbox has been removed.  However, once you activate an approval process, Salesforce removes some of the functionalities.  You can find more about these limitations here, but I found the following to be particularly difficult to deal with when you have changes in business logic and need to modify the approval process:

• You cannot delete active approval process, and you can't deactivate an approval process if there is any records that are submitted for approval (you have to hard delete those records).
• Once you activate an approval process, you can't add, delete or change the orders of the approval step even if you deactivate them.

With the validation rule, the difficulty I sometimes encounter is when trying to migrate some code and there is additional validation rule in the destination org (e.g. production or QA sandbox).  These validation rule can cause your test methods to fail in the destination org even though it has passed in the development org.  You can either create those validation rule in your development org and refactor your test methods to meet the validation rule, or if you are confident that the additional validation rule will not cause the business logic error when the new code is deployed, you can temporarily disable to the validation rule in the destination org during the migration and re-activate it afterwards.