In our final step, we connect application programs to the new service programs via binding.

It's funny sometimes how the computer world goes in cycles. It seems ages ago that we wrote source code, compiled it into object code, and then linked the pieces together. We left that behind for all-in-one compilers, but in today's brave new world of interoperable ILE languages, we've revived the whole concept and now we have to once again think about linking. In current parlance, the concept is known as "binding," and this last article in our series will show you how to manage it.

Quick Recap

This series wasn't a how-to on creating service programs, but more of an application-level view of how to add service programs to your infrastructure. Part 1 addressed how to group your business functions into service programs, while Part 2 talked about using prototypes to tell your programs what the service program procedures look like. We finish off now with a discussion about binding your service programs to your applications.

Creating a Service Program

Remember that a service program is a collection of procedures. If you had only one procedure, you could theoretically just use an old-fashioned bound RPGLE program with a traditional parameter list (PLIST) and be done with it. And in fact, for many parts of your architecture, that old technique is still the best. I find it especially useful in the overall application architecture when I'm separating user interface from business logic. If the business logic is pretty much standalone, there isn't much use for a service program.

Creating a service program has two steps: compiling the module and creating the service program. Creating the service program is only a little more complicated than creating a bound program. You create a module rather than a program. In ILE RPG, that means using the CRTRPGMOD command rather than CRTBNDRPG (option 15 rather than option 14 on PDM). This step creates an object of type of *MODULE rather than*PGM. The extra step is the CRTSRVPGM command, which takes the *MODULE object and turns it into a *SRVPGM.

I hear you out there shouting at your screen, "But we can include multiple modules in a single service program!" Yes, you can. But the focus of this article is ease of programming, and for that I find that starting out with a single module per service program works easiest. Remember, that single module can have dozens of procedures, so it's not like you're creating one service program per procedure. But no less authoritative a source than Jon Paris and Susan Gantner recommend combining multiple modules into a single service program, so in another conversation we can talk about that.

As I said, I want to use the simplest technique possible. That's why I use a mostly vanilla CRTSRVPGM command; I only override the TEXT parameter to grab the text from the module. Although there's no magic parameter setting to use the module text, I set up a custom option in both Rational and PDM to execute the following command:

CRTSRVPGM SRVPGM(&N) TEXT(&T)

That grabs the text from the source and puts it into the service program. That's one of the reasons I like the approach of one module per service program. Let me now expand the command to show you the implied values for a couple of the other keywords:

CRTSRVPGM SRVPGM(&N) MODULE(*SRVPGM) EXPORT(*SRCFILE)
         SRCFILE(QSRVSRC) SRCMBR(*SRVPGM) TEXT(&T)

The MODULE(*SRVPGM) keyword tells the binder to use only the one module with the same name as the service program. EXPORT (*SRCFILE) says we're going to use binder source rather than just exporting all the procedures. The SRCFILE and SRCMBR parameters say where to get the binder source. The alternative is just to export all the procedures and not use any source. You actually do that when you're first creating a service program, but that's a different discussion for a different day. Right now, we assume you have a binder source because that allows you to control the signature, which is critical to the proper management of service programs. By default, signatures change only when a procedure is added or removed, or the order of the exports changes. In reality, adding a procedure really shouldn't change the signature, while at times you may not add a procedure but you still want the calling programs to be recompiled. Binder source allows you to control that process.

Binding to the Application Program

Now that the service program is properly created, it's time to bind it to the application. Remember that the program included a copybook to define the prototype for the service program procedures. However, this doesn't tell the compiler where those procedures can be found, and normally, if we attempt to compile the program with nothing but the copybook, we will get an error that some references could not be resolved. We address that by pointing to a service program.

Pointing to the service program is easy, but it takes a couple of steps. First is to create the object that the compiler uses to find the service program. That's the binding directory, and creating it is as easy as executing the CRTBNDDIR command. Next, the WRKBNDDIRE command allows you to interactively manage your binding directory, adding and removing service programs as needed. Finally, add the binding directory to your RPG program using the BNDDIR keyword on the H-spec (or ctl-opt, for you free-form folks!).

It's that easy! However, like many things about ILE, getting something to work isn't the hard part; creating a framework that allows you to easily manage your development and expand your application infrastructure is much more difficult. And the difficulty isn't because it's hard to do; the difficulty is that it's so easy to do in so many different ways.

In particular, with binding directories you run into a corollary of the question of how many modules to put into a service program. You can have as many modules as you want in a service program, and you can have as many service programs as you want in a binding directory. It's entirely up to you how to arrange and manage your environment. Personally, I like to have one primary binding directory for all applications, and have specific binding directories only for those procedures that I don't want regular applications to access. An example might be an internal security interface; only internal functions should access those procedures, so I keep them in a completely separate binding directory.

Another issue comes into play when compiling service programs that reference other service programs. You may find yourself creating binding directories specific to compiling in order to prevent duplicate references. It's even more difficult when you have circular references in which two service programs call procedures from each other (this is an indirect argument in favor of the "one big service program" design, but that's not for everyone).

How you group your service programs and binding directories is ultimately up to you, but however you choose to arrange them, it's important that you have a plan ahead of time. One of the hardest things in the world is to change your framework in midstream. Hopefully, this series of articles has given you a foundation to work from in building your own service program infrastructure.