Why (seealldata=true) results into more code coverage?

Question

I wrote a trigger to prevent more than n incidents for a client(lookup to user) at any point of time.

  trigger CountLimit on Incident__c (after insert,after update,after undelete) {
  Set<Id> clientids = new Set<Id>();
    Map<Id,Integer> clientmap = new Map<Id,Integer>();
    Max_Incidents__c cs = Max_Incidents__c.getInstance('Platinum');
    Integer maxincs = (Integer)cs.MaxIncidentsPerClient__c;

    for(Incident__c inc:Trigger.new)
    {
        clientids.add(inc.Client__c);
    }

    Map<Id,User> usermap = new Map<Id,User>([select id,name from user where id in:clientids]);

    for(AggregateResult result:[select count(id) sum, client__c client from Incident__c where client__C in:clientids group by client__C])
    {
        clientmap.put((Id)result.get('client'),(Integer)result.get('sum'));
    }

    for(Incident__c inc:Trigger.new)
    {
        if(clientmap.get(inc.Client__c)>maxincs)
        {
           inc.addError('You can\'t open more than '+maxincs+' incidents for '+ usermap.get(inc.Client__c));
        }
    } 

And I wrote this test class-

@isTest
public class IncidentLimit_Test {
    public static testmethod void testcountincident(){

        User usr = CreateUser('Ben','Bentaleb');
          List<Incident__c> inclist = new List<Incident__c>();  
        try{
        System.runAs(usr){
            Test.startTest();
            for(integer i=1;i<=5;i++)
            {
              inclist.add(new Incident__c(client__c=usr.id,Description__c='incident'+i));
            }
           insert inclist;
        Test.stopTest();
        }
       }

    catch(exception e){
      e.getmessage().contains('You can\'t open more than');
      }
        finally{
             System.assertEquals(5, inclist.size());}
               }

public static UserRole CreateUserRole(String name) {
    UserRole ur = new UserRole();
    ur.Name = name;
   insert ur;
    return ur;
}

    public static User CreateUser(String firstname, String lastname) {
    UserRole ur = CreateUserRole('Testing');
         profile pr = [SELECT Id FROM Profile WHERE Name='Standard User'];
    User usr = new User(FirstName = firstname,
                        LastName = lastname,
                        Username = firstname + '.' + lastname +      '@testing.com',
                        IsActive = true,
                        UserRoleId = ur.Id,
                        Email = '[email protected]', 
                        Alias = 'tusr', 
                        TimeZoneSidKey = 'Europe/Berlin' , 
                        LocaleSidKey = 'en_US', 
                        EmailEncodingKey = 'UTF-8', 
                        ProfileId = pr.id, 
                        LanguageLocaleKey = 'en_US'
                        );
        insert usr;
    return usr;
  }
}

The issue I'm here facing is until I add the parameter (seealldata=true) into the annotation the code coverage remains 26% otherwise it's 93%. Please advise.

Answer 1

Max_Incidents__c.MaxIncidentsPerClient__c will not be set when using seeAllData=false, when it's true you will be using whatever value is set in your org.

Your test should be creating that custom setting as well.

Answer 2

In your trigger not all of the code will execute every time. You have if-statements and for-loops. Which lines of your code are executed depend actually much on the specific data you are processing.

With seealldata=false the tests act on a subset of data, usually test-data you create programmatically just for the test.

However if you use seealldata=true, the test can operate over the content of you entire database. Because greater variety of records, more lines will be executed in the trigger and your codecoverage is higher. At the same time seealldata=true comes with serious tradeoffs, since test-results are depending on variable data, you might loose portability and huge data volume can cause severe performance issues. One example I had a couple of years ago was that code deployment form Sandbox to Production took longer and longer. It started with about 30 seconds and soon reached > 10 minutes. Reason was test-execution with seealldata=true. And 10 minutes is just the start: deployments might cost hours after your amount of data has grown big.

To improve the situation without using seealldata=true, you need to create better test data so that more lines of your code will be executed. You can use Developer Console to highlight the uncovered lines of code and adjust or extend you test data accordingly.

Unfortunately there is some data you can't create, like the Standard Pricebook. To get related code covered, seealldata might by the only option in some scenarios.

As best practice you should avoid seealldata=true when possible and use it only in rare cases where really required. At the same time you should try to generate a wide range of test-data programmatically to ensure good coverage.

See also:

• http://www.salesforce.com/us/developer/docs/apexcode/Content/apex_testing_data_access.htm
• http://www.salesforce.com/us/developer/docs/apexcode/Content/apex_testing_best_practices.htm