Unstored calculation fields and performance

[TJ53]

Let’s say we have two related tables: “Invoice” and “Invoice_Item”. We could create a calculation field in the “Invoice” table called “total_amount” with this formula:

total_amount = Sum (Invoice_Item::amount)

This field would have a negative impact in performance when appearing in the layout, since it would have to be defined as unstored, because it’s referencing a field from a related table.

Now let’s suppose this field is not used for any scripts, tooltips, conditional format, etc … would the performance of the database be negatively affected ONLY when this field appeared in a layout? 

In other words, would adding an unstored calculation field to a table involve a performance penalty, even in the “unreal” case where this field didn’t appear in any layout, script, conditional format, etc.? thanks in advance!

[comment]

Unstored calculation are evaluated as needed. If the field is not shown on the layout, and not referenced in any way, it will not be evaluated.

[TJ53]

8 minutes ago, comment said:

Unstored calculation are evaluated as needed. If the field is not shown on the layout, and not referenced in any way, it will not be evaluated.

Thank you! now let's say we have a list layout where the unstored calculation field "total_amount_uc" is showing for each invoice. Therefore, it would have to be evaluated for each invoice record, and therefore there would be a performance penalty.

Now let's say that in order to improve performance, we create a number field called "total_amount_script". Then we create a script called "Update invoice total amount", that executes when an related invoice item amount is entered or modified (or an invoice item is deleted).

Would there be a difference in performance of setting the "total_amount_script" in these two different ways in the script created for updating the number field?

Option 1) Set Field [ Invoice::total_amount_script ; Sum(Invoice_Item::amount) ]

Option 2) Set Field [ Invoice::total_amount_script ; Invoice::total_amount_uc ]

Thanks in advance!

 

 

[comment]

38 minutes ago, TJ53 said:

Would there be a difference in performance of setting the "total_amount_script" in these two different ways

I don't think so - but all performance questions are best answered by performing an actual test. Sometimes the results can be surprising.

Performance aside, if you're going to use a script to populate a stored field with the aggregate value, then you don't need the unstored calculation field - and there is no good reason for the script to depend on its existence.

Note that there may be additional factors to consider here, for example record locking.

 

 

 

 

[TJ53]

Thank you very comment for you insight. The main reason why I would consider going for option 2 (as long as there are no performance implications), even having to define an extra field, is because the unstored calculation field definition would be more easily accessible in case I needed to adjust the calculation.

Otherwise I would have to go to the script workspace and search for the “Update invoice total amount” subscript in order to make a change in the calculation.

On the other hand, I see going for option 2 looks a bit "strange" compared to how standard programming languages work, since in this case the logic would be part of the interface. Also could be confusing to have both fields in the table, and the unstored calculation field not showing in any layout, but only used in the update field script.

It looks like the best practice is to go for option 1, and that’s how I’ve done it in the past. But I have to say that lately I’m tempted to go for option 2, just for having the calculation more quickly accessible in case I need to review!

Edited February 18, 2017 by TJ53

[comment]

One could turn your argument around: suppose you are editing the script, and you want to make a change to the Set Field[] step; you would have to leave the script workspace and look for the calculation field in order to make the change there.

I believe that if an action is scripted, it's good practice to place all the necessary logic within the script. And if you divide your scripts into logical parts and take care to comment each part, you should not have to look very hard in order to find what you're looking for.

[bruceR]

+1 for comment's comments.

As for a field definition being more easily accessible. Do you have any awareness of the risks of doing that?

I suspect not; or you would not have made the statement. Though maybe you just don't have anybody using this system.

What happens when you're defining a field while users are entering data? What can happen to the integrity of the file?

You want users out of the system when doing this. Thus your timeslot for accessibility is extremely limited.

[TJ53]

Thank you very much both. I assumed best practice was option 1, and that's how I've been working so far, but was just wondering about option 2.

20 minutes ago, BruceR said:

+1 for comment's comments.

As for a field definition being more easily accessible. Do you have any awareness of the risks of doing that?

I suspect not; or you would not have made the statement. Though maybe you just don't have anybody using this system.

What happens when you're defining a field while users are entering data? What can happen to the integrity of the file?

You want users out of the system when doing this. Thus your timeslot for accessibility is extremely limited.

Yes, I am aware. That's why I rarely do development in live production systems, specially when doing scripting. In any case that's very valid point, thank you!

[Wim Decorte]

Performance issues aside (and I try to avoid unstored calcs if I can, I prefer my logic in scripts), there is another factor to consider.  If part of the business logic changes and you are using unstored calcs then by changing the calc you risk changing all the historic records.  Which can be a huge problem if you don't think through that ahead of time.

[TJ53]

Thank you very much for the responses.