Application Package Weirdness

I’m in an environment that is at PTools level 8.53.  I’ve got a App Package with three classes, and I’m using PSUnit to test the various class methods before I use them in as part of a project I’m working on.

I’ve got a method that simply finds the next Saturday from the current date.  It’s very simple:

/**
* Sets the amend effective date to the
* next Saturday following the current system date.
*/
method SetNextSaturday
Local string &sql = “SELECT NEXT_DAY(sysdate,’SATURDAY’) from dual”;

SQLExec(&sql, &amend_effdt);

end-method;

The output bind variable &amend_effdt is defined as a private instance for the class.  The method SetNextSaturday is invoked within the class constructor, meaning that as soon as the class object is instantiated the method gets called.

My test is also very simple:

method Run_DateUtils
%This.Msg(“TestAmendUtils.Run_DateUtils”);
Local ZG_NEG_AMEND:AmendNegotiation:DateUtils &util = create ZG_NEG_AMEND:AmendNegotiation:DateUtils();
Local date &amend_dt = &util.amendEffDt;
Local string &sql = “SELECT NEXT_DAY(sysdate,’SATURDAY’) from dual”;
Local date &next_sat;
SQLExec(&sql, &next_sat);
%This.AssertDatesEqual(&amend_dt, &next_sat, “Method DateUtils not deriving amend date correctly!”)
end-method;

When tested this error is thrown:

Test [ZG_NEG_AMEND_TEST:TestDateUtils] failed!
SqlExec: parameter 2 is neither a record object nor a name. (2,284)

Odd.  The SQL runs perfectly in SQL Navigator.

And then I looked at the error explanation.  Parameter 2 is a problem.  Hmmmm.

So I refactored by declaring a local date variable in the method, then assign it’s value to the instance variable.  That worked.  That’s also ugly.

This has also been a problem since 2008 PTools version 8.44 based on this post.  It seems that a property cannot be an output argument of a function, and that an instance variable is treated as a property.

What makes this more confusing is apparently an instance variable is fine for input.  It can be read from, just not written to.

More fun with Application Packages – Instances

In my last post I had made this statement:

Instance variables are like Static variables in Java.

Hat tip to Helena who provided insight:

One comment though – instance variables are not static variables (in terms of scoping rules, one the same private instance variable is not shared by multiple instances of the class); they are in fact instantiated for each instance/object of the class.

So if I have a class definition ClassDef, with 1 private instance variable &InstanceVar, and I instantiate 2 objects of type ClassDef, I will have 2 instances of &InstanceVar in memory, not just 1.

This contrasts with static member variables in Java, where the variable is shared and in the example above, only 1 instance of &InstanceVar would have been allocated in memory.

Good point – so lets dig into this a bit – as others have here and here.

Since I made the statement like Static variables in Java let’s look at what those are.  From Oracle’s Java Tutorials on Understanding Instance and Class Members:

When a number of objects are created from the same class blueprint, they each have their own distinct copies of instance variables. In the case of the Bicycle class, the instance variables are cadence, gear, and speed. Each Bicycle object has its own values for these variables, stored in different memory locations.

Sometimes, you want to have variables that are common to all objects. This is accomplished with the static modifier. Fields that have the static modifier in their declaration are called static fields or class variables. They are associated with the class, rather than with any object. Every instance of the class shares a class variable, which is in one fixed location in memory. Any object can change the value of a class variable, but class variables can also be manipulated without creating an instance of the class.

I added the bolding and underlining in the above.  So as Helena pointed out, in Java the same memory location is used by every instance of the class having a variable defined as static.  Each instance of a class gets a reference (having been weaned with C I would say pointer but a Java purist would start throwing acorns at me) to the location in memory where the value actually resides.  Which falls in line with how memory allocation works in Java.  Java stores objects on the heap, variables sit on the stack.  But variables are ‘merely’ pointers/references to the objects sitting on the stack.  Enough teasing – I also know there are differences between a pointer and a reference – but that is a C++ convention, not C.  In any event, it’s efficient for Java to have Static variables work this way.

PeopleBooks provides this tidbit in Declaring Private Instance Variables:

A private instance variable is private to the class, not just to the object instance. For example, consider a linked-list class where one instance needs to update the pointer in another instance. Another example is the following class declaration:

class Example private instance number &Num; end-class;

A method of Example could reference another instance of the Example &Num instance variable as:

&X = &SomeOtherExample.Num;

Avoid making every instance-scoped variable a public property. You should consider each variable individually and decide if it belongs in the public interface or not. If it does, decide if the variable warrants get or set modifiers and therefore should be a public property. If the variable only serves internal purposes to the class, it should be declared as a private instance variable.

Again, I added bolding and underline.  But note the word pointer – a double plus unJava word.  So lets look at a language that does use pointers.

I’m a pack rat.  I don’t throw books away.  So I looked in my collection and grabbed my copy of Deitel & Deitel C++ How To Program:

Each object of a class has its own copy of all the data members in the class.  In certain cases only one copy of a variable should be shared by all objects of a class…  A static class variable represents “class-wide” information. 

Let us motivate the need for static class-wide data with a video game example.  Suppose we have a video game with Martians and other space creatures.  Each Martian needs to be brave and willing to attack other space creatures when the Martian is aware that there are at least 5 Martians present.  If there are fewer than 5 Martians present, each Martian becomes cowardly.  So each Martian needs to know the martianCount.  We could endow class Martian with martianCount as a data member.  If we do this, then every Martian will have a separate copy of the data member and every time we create a new Martian we will have to update the data member martianCount in every Martian.  This wastes space with the redundant copies and wastes time in updating the separate copies.  Instead, we declare martianCount to be static.  This makes martianCount class-wide data.  Every Martian can see the martianCount as if it were a data member of the Martian, but only one copy of the static martianCount is maintained by C++  This saves space.  We save time by having the Martian constructor increment the static martianCount.  Because there is only one copy, we do not have to increment separate copies of martianCount for each Martian object.

Hmmm.  Something fishy here.  Both C++ and Java are using static variables the same way.  Let’s look at PeopleSoft some more.  There is some code here that bolsters what Helena had pointed out.  The instance in an App Package class gets instantiated as its own distinct block of memory – so it breaks the notion and value of having a C++/Java type static variable.  However it’s able to be referenced and changed by another instance of the same class.   So take the less efficient part of the Deitel explanation and there you have it.

To my mind this is bordering on an architectural bad smell of connector-envy, and can lead to some code smells on the part of development.

Defining and Using Constants with PS Application Package

I’ve used Application Packages more and more after reading about them in Jim Marion’s PeopleSoft PeopleTools Tips & Techniques.  And I like using them far more than creating functions in a FUNCLIB_ library.

I’ve got an App Engine program I’m developing, and I’m using an App Package class for error handling.  I’ve a small list of error codes that get sent by the exception handler routine and those codes are used to get a description.  I could use message catalog for the texts of course, but in this case I’m going to be lazy and keep the values in the code.  This is for a batch process that will be scheduled to run nightly; it isn’t customer facing.

To avoid having magic numbers appear in my code I wanted to define a set of constant integer values in the App Package class.  PeopleBooks gets a bit obtuse on this subject though.

So, let’s go thru the exercise of creating them, then how to use them.

In the class definition I create a private constant:

private
Constant &INVALID_EMPLID = 1;

So far, so good.  any methods inside the class can use that value:

Evaluate &errCode
When = &INVALID_EMPLID
&msg = “Invalid emplid provided.”;
Break;

But I want the code throwing the exception access to the same constant.  PS App Package definitions are – well – eccentric.

Java, C++ and C# class structures look similar.  You define the class, then add fields if needed, and the methods of the class.

App Packages have methods, but then bring in the idea of properties and instances as well as Constants.

A property is exactly that, it’s a property of the instantiated object from the class – it’s the Has A of the object.  As in my class House Has A property of RoomColor.  PeopleBooks provides an explanation for a property as an attribute of an object.

Instance variables are like Static variables in Java.  An instance is private to the class, not just to the instantiated object.

Which gets us nowhere in terms of having our constant value used in my exception handling – yet.

Here is the key part – PeopleSoft set up properties to take the place of creating get and set methods.  They consider it more efficient to declare a property (which has no parameters) than to use a method to do get/set routines.

So instead of having a method:

method getSomething() as integer;

You would instead declare a property:

property integer mySomething get;

This was the long way around to explain how I get to use my Constants definition.  You saw above where I created a constant for &INVALID_EMPLID.

I declare a public (this has to be public – PS will bark at you if you try to do anything else with it) property – making sure it has a get as part of the definition:

property integer invalidEmplid get;

Then you write a get routine:

get invalidEmplid
/+ Returns Integer +/
Return &INVALID_EMPLID;
end-get;

Now my App Engine code can access that constant value – assuming that I’ve declared the App Package Class inside the code:

&errnum = &err.invalidEmplid;

If I find later on that I want to change the value &INVALID_EMPLID points to, I only have to do it in one place.  Less code maintenance, which is the real value of not using magic number coding.

%d bloggers like this: