It's Friday afternoon.
You just deployed a new feature to your Salesforce Org.
Something is clearly wrong with it.
Your users are complaining, your boss is calling you in panic.
You need to fix it NOW.
You open the debug log, it's a simple mistake.
You fix the error in your Apex class in just a few minutes.
You prepare the deployment package.
You wait for the package validation.
[60 minutes later], your package is ready to go.
Clearly not the Friday you wanted.
Sound familiar? I know it's happened to me more than once.
Package validation is sometimes longer than actually fixing the problem and while we have no control over components validation, we do have some control over the unit testing part of it.
Apex Testing
We do our best to cover all possible edge cases with a complete set of unit tests, but mistakes happen.
The problem is the more tests we add, the longer that package validation process takes to run. How can we make it go faster while maintaining the coverage and safety that a good set of tests provide?
My approach is to prevent as many database operations as possible in your test classes.
Step 1 - Remove logic from your triggers
Extracting business logic outside our handlers makes it easier to understand everything that is going on inside of it, but also makes it easier to design our code as reusable components.
Step 2 - Remove queries from your business logic
By writing queries directly inside your business logic, it's impossible to execute your tests without going to the database, which we want to avoid.
Step 3 - Test your querying separately
By isolating your queries from your business logic, you make your query logic reusable and testable.
Results:
Before:
Test Execution Time 705 ms Test Execution Time 660 ms Test Execution Time 535 ms Test Execution Time 593 ms
After (without insertion):
Test Execution Time 73 ms Test Execution Time 43 ms Test Execution Time 28 ms Test Execution Time 45 ms
After (with insertion):
Test Execution Time 700 ms Test Execution Time 497 ms Test Execution Time 647 ms Test Execution Time 551 ms
The test execution time when keeping everything in memory is at least 10 times faster, and the changes required to make it possible in tests has no significant hit on performance even when performing database operations.
You can try this yourself and access the code I used on Github.
Note: All my examples will be using my favorite trigger framework made by Kevin O'Hara and available on GitHub here.