How to avoid recursive trigger other than the classic 'class w/ static variable' pattern?
Another good way to avoid recursion is highlighted in Chapter 6 of Dan Applemans's excellent book, Advanced Apex Programming. This particular example is around recursion when looking for a specific field change which is very common in many triggers.
As mentioned in @BarCotters answer, you generally run your trigger on very specific criteria. A change in a field can trigger certain logic in your trigger. If you look at the Order of Operations for triggers, you can see that any workflow rules and field updates based on those rules are run after the both before and after triggers have run. It runs however prior to the changes actually being committed to the DB, which is where this can cause problems when detecting field changes.
So lets say we are using a very simple example where we have a trigger on the Opportunity object that looks for opportunities that have been Won, and based on that it creates a new custom object record.
trigger OpportunityTrigger on Opportunity (after update) {
map<Id,Opportunity> justWonOppsMap = new map<Id,Opportunity>();
for (Opportunity o : Trigger.new) {
if (o.isWon != Trigger.oldMap.get(o.Id).isWon) {
justWonOppsMap.put(o.Id, o);
}
}
List<Some_Custom_Object__c> objs = new list<Some_Custom_Object__c>();
for(Opportunity o : justWonOppsMap.values()){
objs.add(new Some_Custom_Object__c(
Name = 'New object based off ' + o.Name,
Opportunity__c = o.Id
));
}
insert objs;
}
This trigger will work fine, and will create a new custom object when you close an Opportunity. It will work fine that is, assuming you have no WFR's that have field updates on the Opportunity Object.
Let's say now I had a workflow rule on the Opportunity that When an Opp is closed, it changes the CloseDate of the Opportunity to Today. (A fairly simple and common WFR that many admins may add).
That simple change now Breaks my trigger. This same trigger will now actually create 2 custom objects when my opportunity is closed. This is due to the fact that the WFR rule is now firing the triggers one final time.
This is how you would think the logic works
- Trigger Run 1 (Old Value = Not Won, New Value = Won)
- Workflow Trigger
- Run 2 (Old Value = Won, New Value = Won)
This is not how it works though, even though the opp has been updated, it has not yet been committed to the Database, so in the second run of the trigger it still see's the opp and just being closed like below
This is actually how it works
- Trigger Run 1 (Old Value = Not Won, New Value = Won)
- Workflow
- Trigger Run 2 (Old Value = Not Won, New Value = Won) Identical as first run
This is where Dan Appleman's solution comes in. He advises to use a mechanism that actually checks for the 'correct old value'. This would allow the second run of the trigger to detect the value that was set in the first run of the trigger.
trigger OpportunityTrigger on Opportunity (after update) {
OpportunityTriggerHelper.OppAfterUpdate(trigger.ew, trigger.old, trigger.newMap, trigger.oldMap);
}
public class OpportunityTriggerHelper {
Private static Map <Id,boolean> oldIsWonMap = null;
public static void OppAfterUpdate(list<Opportunity> newOpps, list<Opportunity> oldOpps, map<Id,Opportunity> newMap, map<Id,Opportunity> oldMap) {
if(oldIsWonMap == null) {
oldIsWonMap = new map<Id,boolean >();
}
map<Id,Opportunity> justWonOppsMap = new map<Id,Opportunity>();
for (Opportunity o : Trigger.new) {
//This checks to see if there was a value set in a previous trigger run
boolean oldIsWon = (oldIsWonMap.containsKey(o.id)) ? oldIsWonMap.get(o.id) : oldmap.get(o.id).isWon;
//this checks the current opp value with the 'correct' old value
if(o.isWon && !oldIsWon){
justWonOppsMap.put(o.Id, o);
}
//this puts in the 'correct' old value in case the trigger is run again in the same context
if(oldIsWon != o.isWon) {
oldIsWonMap.put(o.id,o.isWon);
}
}
List<Some_Custom_Object__c> objs = new list<Some_Custom_Object__c>();
for(Opportunity o : justWonOppsMap.values()){
objs.add(new Some_Custom_Object__c(
Name = 'New object based off ' + o.Name,
Opportunity__c = o.Id
));
}
insert objs;
}
}
With this change, now the WFR rule does not break our trigger. Only 1 custom object is created when the opp is closed.
I know this was longwinded, but I hope it helps. I would highly getting this book as this was only a few pages worth of goodness, and this book is jam packed with great knowledge that I think all SFDC developers should have.
Heres the link again.
http://advancedapex.com/
Your trigger should only be firing on very specific criteria. Make them as specific as possible by comparing old and new field values to make sure your criteria is met.
For example if you wanted to only fire when a contacts BirthDate changed then you could do the following:
trigger ContactBeforeTrigger on Contact (before insert, before update, before delete) {
if (Trigger.isUpdate) {
for (Contact contact : Trigger.new) {
if (contact.Birthdate != Trigger.oldMap.get(contact.Id).Birthdate) {
contact.SomeField__c = 'DOB Changed';
}
}
}
}
Even though the trigger is updating the current Contact, we don't need to worry about recursion because:
- The trigger will only fire on updates due to the
Trigger.isUpdatecheck - The trigger will only fire when the Birthdate is changed.
If you want to limit each trigger to fire only once then what you described would be the solution. SalesForce document it here but it may be a valid scenario that triggers fire more than once in the same transaction, the issue is if it is an infinite loop. You will hit a Governor Limit if there is more than 16 recursions of the same trigger.
Keeping a list of the last value seen by the trigger as described by Chris Duncombe is not quite sufficient. In addition to the documented case of a workflow field update causing a trigger to be fired a second time with the same old/new versions of a record, there are a couple other cases you may need to handle.
- Update events firing before insert events
- Seeing the new version of a record in
Trigger.oldbefore seeing the old version
Both of these can occur when there are multiple triggers on an object. (The second one is relevant to Chris Duncombe's answer.) Consider the following triggers:
trigger MyTrigger on Account (after insert, after update) {
for (Account a: Trigger.new) {
if (Trigger.isInsert || a.My_Field__c != Trigger.oldMap.get(a.Id).My_Field__c) {
...
}
}
}
MyTrigger is the trigger you wrote, in which you are trying to solve the double firing problem.
trigger TheirTrigger on Account (after insert, after update) {
List<Account> updates = new List<Account>();
for (Account a: Trigger.new) {
// Assume there's a good reason for doing this in an after trigger
updates.add(new Account(Id = a.Id, Their_Field__c = ...));
}
update updates;
}
TheirTrigger is another trigger which updates the object used by MyTrigger.
When TheirTrigger fires before MyTrigger, both of the scenarios listed above can occur. After an insert, TheirTrigger updates the Account causing the MyTrigger to fire in an after update context. If you have different after insert and after update logic, you need to make sure that MyTrigger still handles the after insert event when it occurs.
Now suppose My_Field__c is updated from "No" to "Yes". When TheirTrigger fires first, MyTrigger will first get called in an after update context in which My_Field__c already contains "Yes" in Trigger.oldMap. Then after the callstack for TheirTrigger unwinds, MyTrigger fires again with the previous value of "No" in Trigger.oldMap.
The best solution I've found is to keep all of the old versions seen by each trigger. When the trigger gets a version of an object in Trigger.oldMap that it has already processed due to a workflow field update, it can ignore it; otherwise, it should check the trigger-specific criteria to decide whether to process it.