A while ago I read an answer on stackoverflow about error handling. I can’t seem to find that question now, so you’ll have to take my word for it. The question was asking about error handling best practices in VBA and I found one of the answers interesting. The answerer said that you could use the Terminate event of a custom class module as your error logger. I had never thought of that.
I’ve been using the Professional Excel Development error handling method with great success for many years. This method controls the error as it moves back up the call stack, either by returning a Boolean to the calling procedure or by rethrowing the error. Without error handling, VBA sends the error back up the call stack automatically until it is handled. So, for instance, if you only handled errors in your entry point procedures, you would still have an error handler. You just wouldn’t have the information about the stack that would be critical to finding out where the error occurred.
The class method of error handling takes advantage of the built-in ability of VBA to pull an error back up the stack. It uses the fact that local variables go out of scope when the procedure is complete, such as when an unhandled error occurs and the cursor is sent to the calling procedure. If you had a local variable pointing to a class, that class’s Terminate
event would fire when an unhandled error occurred.
Borrowing the setup from PED2, let’s see how this would work.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 |
Sub EntryPoint() Dim clsError As CError On Error GoTo ErrorHandler Set clsError = New CError: clsError.SetLoc "Module1", "EntryPoint" SubProc1 ErrorExit: Exit Sub ErrorHandler: If gbDebugMode Then Stop: Resume Else Set clsError = Nothing MsgBox Err.Description Resume ErrorExit End If End Sub Sub SubProc1() Dim clsError As CError Set clsError = New CError: clsError.SetLoc "Module1", "SubProc1" SubProc2 End Sub Sub SubProc2() Dim clsError As CError Set clsError = New CError: clsError.SetLoc "Module1", "SubProc2" Debug.Print 1 / 0 End Sub |
EntryPoint
calls SubProc1
. SubProc1
calls SubProc2
. An error occurs in SubProc2
. Only EntryPoint
has error handling. It uses On Error Goto ErrorHandler
to route program flow to the error handling section. Neither SubProc1
nor SubProc2
have any error handling. We’re going to let VBA kick the error back up to the call stack until it gets to EntryPoint
.
Each procedure has a local variable that points to an instance of CError
. CError
is a custom class whose Terminate
event I’ll be using to log the error as it moves back up the stack. When the error occurs in SubProc2
, the clsError
variable in SubProc2
goes out of scope and its Terminate
event fires. The error is passed up to SubProc1
by VBA by design. Because there is no error handling in SubProc1
, that error causes the instance of clsError
in SubProc1
to go out of scope and its Terminate
event fires.
Once again, VBA does it’s thing by passing control back up the stack, error in tow. EntryPoint
does have error handling, so when program control reaches it, the ErrorHandler section goes into action. Assuming we’re not in debug mode, the first thing to do is terminate clsError
by setting it to nothing. By the time we exit this procedure, the built-in Err
object will have been reset and we won’t have anything to log. By setting clsError
in EntryPoint
to Nothing
, we get the last entry in our log. After that, the error is displayed and program control is sent back to ErrorExit
for any clean up (no clean up in this example, just the Exit Sub
).
The log looks like this:
01 Jan 14 21:40:40 [errorclass2.xlsm]Module1.SubProc2, Error 11: Division by zero
01 Jan 14 21:40:40 [errorclass2.xlsm]Module1.SubProc1, Error 11: Division by zero
01 Jan 14 21:40:40 [errorclass2.xlsm]Module1.EntryPoint, Error 11: Division by zero
Of course I made it virtually identical to PED’s log entry.
Instead of putting error handling in all of the downstream procedures, I just put a local variable that will terminate when an error occurs. The class looks like this:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 |
Private mlErrorID As Long Private msProcedureName As String Private msModuleName As String Private Sub Class_Terminate() If Err.Number > 0 Then Debug.Print Format(Now, "dd mmm yy hh:mm:ss") & Space(1) & Me.Location & ", " & Me.ErrDescription End If End Sub Public Property Let ModuleName(ByVal sModuleName As String): msModuleName = sModuleName: End Property Public Property Get ModuleName() As String: ModuleName = msModuleName: End Property Public Property Let ErrorID(ByVal lErrorID As Long): mlErrorID = lErrorID: End Property Public Property Get ErrorID() As Long: ErrorID = mlErrorID: End Property Public Property Let ProcedureName(ByVal sProcedureName As String): msProcedureName = sProcedureName: End Property Public Property Get ProcedureName() As String: ProcedureName = msProcedureName: End Property Public Sub SetLoc(ByVal sModule As String, ByVal sProc As String) Me.ModuleName = sModule Me.ProcedureName = sProc End Sub Public Property Get Location() As String Location = "[" & ThisWorkbook.Name & "]" & Me.ModuleName & "." & Me.ProcedureName End Property Public Property Get ErrDescription() As String ErrDescription = "Error " & Err.Number & ": " & Err.Description End Property |
I’ve kept the logging pretty simple for this example. In the Class_Terminate
event, I first check to see if Err.Number
is zero. This method relies on the fact that the Terminate
event will fire when an error occurs. But in reality, the Terminate
event will fire when the subprocedure finishes without error too. It fires whenever my local variable goes out of scope and that happens when there’s an error or when the subprocedure completes. We only want to log when an error occurs, so we have to check that.
The logging is a simple Debug.Print
statement. To replicate the PED method, that would need to be expanded to write to a log file.
This is a very simple example that I put together to see how this might be setup. There might be some problems with this method that I haven’t encountered. I’m not advocating that you use this method, but I am intrigued by its simplicity. If you have any thoughts on this method of error handling or on error handling in general, leave a comment below.
You can download errorclass2.zip
Hi Dick,
Happy new year, and thanks for your very good article.
I’m using PED error handler too and i’m a big fan.
Your method seems really great too, but for example, I’m using lot of functions in my code and don’t you think that calling for each of them a new clserror object won’t be too much for memory (even if your class is pretty small) ?
I’ll try your way instead of PED for my next development and give you a feedback.
Thanks for sharing.
Cyril
Cyril: Good point about excessive memory. They’re pretty small classes, as you say, but if you get too many levels deep into the code that could add up. It would be interesting to see what kind of effect that has.
Fascinating – and thanks for writing it up.
I see a couple of issues.
First I like to get the line number of the vba statement that experienced the error – currently with PED that is passed in as Erl – can the class approach support it?
Second, as per PED, there are usually clean up statements before exiting any procedure – and those clean up statements are preceded by “On Error Resume Next” just in case there is an error that we didn’t bother to handle. What if an unhandled but resumed error happens just before the subroutine terminates the class – will the err number appear in the class termination code and hence get processed like it actually happened?
And finally, I wonder about the overhead as mentioned earlier – with the non-class approach the error handling does not impact memory and performance until an error occurs – except I guess the first “On Error Goto ErrHandler” statement. Which is how it should be – we don’t want to slow down normal processing to handle unexpected and rarely occurring errors. It would be great to hear back if there are any memory and performance issues.
Thanks for posting – it is always worthwhile challenging the standard way of handling things just in case there is a better way. I particularly liked the concept of SystemState that does use the class concept – it cleaned up my code immensely and now I use it all the time (where appropriate).
–Charlie
Yes, Erl will work here too.
Yes, errors in the clean up procedure will be logged, but not displayed.
What do you guys think is a good test of the memory cost? How many levels deep do I have to go before it’s even significant? There is a cost to the PED way too. All of your code gets compiled and loaded into memory and the PED way uses far more lines per procedure than this way. I’m guessing that neither is significant, but I think it’s time for a test. I’m on it.
Fifteen subprocedures under an entry point procedure (16 in the call stack) and the memory shown by Task Manager went up by … wait for it … zero.
Thanks to SC for finding the Stack Overflow question: http://stackoverflow.com/questions/3792134/get-name-of-current-vba-function
And here’s a Wikipedia entry about it: http://en.wikipedia.org/wiki/RAII
Interesting approach!
I like the fact that you don’t need an error handler in each routine ( i.e. -3 lines when using : ) but instead only add 2 lines of code (or actually 1 with : )!
However, the huge advantage of the PED error handler is the Stop: Resume statement if in debug mode in each routine! This makes debugging so much easier. So I guess I’d end up up with the error handler anyway – and then the class approach doesn’t add any additional value. :-(
Or am I missing something here?
Wouldn’t this code produce the same functionality/result as the PED approach ?
Dim x01
Sub EntryPoint()
On Error Resume Next
x01 = "Module1.Entrypoint"
SubProc1
If Err.Number <> 0 Then MsgBox x01
End Sub
Sub SubProc1()
x01 = "Module1.SubProc1"
' Debug.Print 1 / 0
SubProc2
End Sub
Sub SubProc2()
x01 = "Module1.SubProc2"
Debug.Print 1 / 0
End Sub
@snb: It would, but only in this special case. Imagine the error would be caused in SubProc1 _after_ calling subProc2. Then x01 would hold the wrong information…
I knew there was going to be a fatal flaw this method and the Stop:Resume piece is it. Even if you put a Stop in the Terminate event, the offending procedure has already popped off the stack, so you can’t even get to it that way.
Hi Dick
I’ve just returned to vba programming after several years away and just wonder at the speed you can be up and running with something functional in no time at all. Anyway, I’ve been going over the site in the last few weeks and have to congratulate you guys on the choice of topics and excellent examples. I’m almost ‘classed’ out but having read your error handling classes I feel I have room for more.
I love the idea of this approach and wonder about yours and Peter’s comments about the advantages of debug stop. Without sounding like a total fool I feel like I’m missing something too – surely the error log should provide all the information you need. What is the most important thing here? debugging or elegant, by design error trapping?
Jase: The problem with the error log is that it doesn’t give you a line number. So it’s very difficult to determine which line caused the error.
Okay, but what if I have a debug flag set. If I’m in debug mode, the class never gets created and the code errors right where it errors. I’ll have to think about this.
I know is over a year old but for anyone who might stumble on this: I came up with a work-around to the seeming limitation of the class for the Stop:Resume that Dick Kusleika pointed out.
If you feed the VBE project name (via Global Constant gsVBEProj eg) and module name (via module constant msMODULE eg) to the class upon initialization (just as is already done for the location), then you can emulate the Stop:Resume behaviour with this piece of code:
Stop: Application.Run VBEProj & “.” & ModuleName & “.” & ProcedureName
Effectively this will return to the procedure where the error occurred, BUT not at the offending line.
Sadly I think this is as close as you can get but not quite it.
Great approach, and very usefull for debugging.
IMO, there is no need for a resume or stop mecanic, as this class is (as stated) an “Error Logger”, its job is to log errors and that’s what it does.
Any other suggested mecanics belong to a something else.