This post stems from recent conversation on the subject of prototyped parameters. I always pass parameters as read only (using the const keyword) in order to ensure that no unexpected changes in the called procedure can impact the calling procedure.
With this approach, the only value you need to worry about is the value explicitly returned by the procedure and this enforced simplicity makes life much, much easier for whoever ends up maintaining the application.
But, my colleague wanted to know, what do you do if you want your procedure to return multiple values? It doesn’t make much sense, for example, to call a separate procedure for each line of a customer address.
The answer is to use a data structure.
As an example, here’s a main procedure that, among other things, retrieves a customer address. As should be clear, I am using a data structure (ADR) to hold the address and using the procedure getAddress to populate this data structure.
p Main b
d Main pi
d ADR ds qualified
d Street 50a inz
d House 5s 0 inz
d Postcode 4s 0 inz
d City 50a inz
/free
// Some loop reading customers
// ...
// Retrieve the customer address
ADR = GetAddress(CustomerNumber);
// Continue
// ...
return;
/end-free
p Main e
And this is what the getAddress procedure looks like:
p getAddress b
d getAddress pi likeds(ADR)
d Customer 10p 0
d ADR ds qualified
d Street 50a inz
d House 5s 0 inz
d Postcode 4s 0 inz
d City 50a inz
/free
// Populate the ADR fields
ADR.Street = 'SomeStreet';
...
// And return the data structure
return ADR;
/end-free
p getHistory e
As you can see, as long as the ADR data structure is locally defined in both procedures you can easily pass the contents of the data structure from one procedure to the other.
Footnote
I know I am using fixed format definitions here, but the box we were looking at while discussing this is only on 7.1.