Free-format RPG is the future, but getting there is still an adventure.
Even the best conversion tools only get you part of the way to free-format RPG; here is the first step of the rest of the journey.
I no longer write programs with fixed-format specifications. In fact, when I do occasionally have to work on an older machine, I confess that I find myself stumbling around a bit with F-specs for files. The other day I had to remind myself how to define a data area. It took a while to get to the point where I was comfortable writing all my code in free-form RPG, especially when it came to display files. But the longest journey starts with a single step, and this article is intended to start you on that path.
Use Your Tools
This set of articles is intended to provide you with the finishing work: those things that require manual intervention even after an initial conversion. But you do have to get to that initial point of working with free-form code (note that we now use the term free-form as opposed to /free, since the /free directive is no longer required). To start any conversion, you have to remember to use your tools. There are two basic paths you can use: IBM or HelpSystems.
The IBM path consists of using CVTRPGSRC to convert your RPG III code to fixed-format RPG IV and then using the Convert Selection to Free-Form option in Rational Developer for the i. At this point, that will only get your C-specs into free-form syntax; IBM's tools don't handle converting H-, D-, or F-specs.
Another option is to use a third-party utility. Linoma has such a conversion tool. Recently acquired by HelpSystems, Linoma has been providing various tools in the IBM midrange space for a very long time, and their RPGWIZ conversion tool can convert RPG III code almost completely into reasonably well-structured free-form RPG. Over the years, I've fine-tuned my use of the many options the tool provides to get me to a near total conversion. The result doesn't follow exactly the same standards I use for my brand-new programs, but it's close enough to coexist (you can't always say that about IBM's conversion).
IBM also partners with another company, Arcad, that has a conversion tool. I haven't used it, so I can't tell you how well it does the conversion, but I do know that it is priced by the number of conversions.
What's Left: The CALL Opcode
Whether you use a procedural approach or a conversion utility, you may still have some remaining pieces to convert. One of the most common is a program call. There are two basic flavors of the CALL opcode in RPG: one that uses PARMs after the CALL opcode and the other that uses a PLIST. I'm not telling you anything you don't already know when it comes to PLISTs, but for completeness let's just say that a PLIST is a way of encapsulating the PARMs for a program call. It's a relative of the KLIST, which we'll be dealing with in a subequent article. The same PLIST can be used for different programs, and you can use different PLISTs for the same program. I think this flexibility is the big reason that tools have problems with program calls.
What this article will do is provide you with a technique for handling the simplest conversion along with guidance on how to address the more complicated situations. Let's start with the simple case: a CALL opcode followed by one or more PARM opcodes.
C CALL 'PROGRAM'
C PARM PARM1
C PARM PARM2
This is the most pared-down syntax. Converting it to free-form is a two-step process. First, you must use a procedure declaration to identify this as a call to an external program. Then you replace the fixed-format CALL opcode with a free-form CALLP. The procedure declaration in this case is very simple:
dcl-pr PROGRAM extpgm;
Thanks to RDi's syntax coloring, you can see that, other than keywords, all of the changeable names come directly from the original code. Now this is a very abridged version of the procedure declaration, and it only works because of a few assumptions I've made:
- The name of the procedure is the same as the name of the program.
- I don't use name placeholders for the parameters.
- Parameters are bidirectional.
For the first assumption, I personally favor naming my prototypes the same as the name of the program. This of course means that you're restricted to the IBM i limit of 10 characters, and you have to use whatever naming convention applies to the program object. Some people prefer a longer, more descriptive name and that's fine, but that allows you to call the same program by different names in different calling programs. I get confused when the same program is called different things, but that's just me; I'm a simple guy.
The second thing is the use of *N for the parameter names. This simply means that I'm not naming the parameters, and once again, I'm opting for simplicity. You can name those parameters whatever you want, but it's a little strange to me, because those names aren't actually defined in the program. Names in procedures are really no more than comments, and I prefer my comments to be preceded by slashes (which also turns them turqoise). You may also notice that I'm using the LIKE parameter, which in turn means that I'm assuming that PARM1 and PARM2 are defined. Well, if they're in PARM opcodes, they must be defined. The only wrinkle would be if they were defined in the PARM opcode itself, which you can do:
C PARM PARM1 10
You'd have to change this code to instead define PARM1 in a D-spec somewhere in order for my technique to work. The better conversion utilities will do that for you.
Finally, as I noted, this assumes bidirectional parameters. That really comes into play more on the call, so let's see what the converted call looks like:
PROGRAM( PARM1: PARM2);
It doesn't get much simpler than that. No matter how many parameters in your original PARMs, just keep adding them to the list here. Since a free-form specification can span multiple lines, just keep adding until you're done.
I said there were potential complications. For example, what about when you have a constant? You may have something like this:
C CALL 'IOCUST'
C PARM 'G' OPCODE
C PARM CUSTNO
C PARM RECORD
This isn't intended to be a real example, but hopefully it will suffice to show the fundamental change that can best handle this case. The prototype would be more like this:
dcl-pr IOCUST extpgm;
*n like(OPCODE) const;
The parameters are defined the same way, the difference being that first parameter, which has the keyword CONST added. CONST says that the value being passed to the program will not be changed in the calling program, which in turn allows you to specify a literal and have the compiler create a temporary variable with that literal to pass to the program. The call then looks like this:
IOCUST( 'G': CUSTNO: RECORD);
A temporary variable of the same type as OPCODE will be created and initialized to the value ‘G’ before being passed to IOCUST. This won't handle every possible nuance of the PARM opcode. If you expected to get a value back in the OPCODE variable, then obviously it won't happen here.
Other issues come into play. What if you call the same program with different lists of parameters or different PLISTs? Then you have to start making some decisions on how you want to set up your procedure declaration. Though daunting, it's not as impossible as it might seem at first glance. For example, even if you don't always pass the same paramters (or even the same number), in all but the most unusual program designs you always pass the same kind of paramters. In that case, using *NOPASS on your parameters can handle different numbers of parameters, and you can pass any variable you'd like by just naming it in the procedure call.
If you're using the same PLIST for multiple programs, you'll probably want to create separate procedural declarations for each. There are some slick ways around that, which we'll cover in another article.
Just the Beginning
The topics in this article should get you started on your way to complete free-form RPG programming. Look for other topics, including KLISTs and BIFs, in subsequent articles.