IBM RPG


RPG is a high-level programming language for business applications. RPG is an IBM proprietary programming language and its later versions are available only on IBM i- or OS/400-based systems.
It has a long history, having been developed by IBM in 1959 as the Report Program Generator — a tool to replicate punched card processing on the IBM 1401 — then updated to RPG II for the IBM System/3 in the late 1960s, and since evolved into an HLL equivalent to COBOL and PL/I.
It remains a popular programming language on the IBM i operating system, which runs on IBM Power platform hardware. The current version, RPG IV, a.k.a. ILE RPG, provides a modern programming environment.

Overview

An RPG program once typically started off with File Specifications, listing all files being written to, read from or updated, followed by Data Definition Specifications containing program elements such as Data Structures and dimensional arrays, much like a "Working-Storage" section of a COBOL program. This is followed by Calculation Specifications, which contain the executable instructions. Output Specifications can follow which can be used to determine the layout of other files or reports. Alternatively files, some data structures and reports can be defined externally, mostly eliminating the need to hand code input and output specifications.
In the early days of RPG, its major strength was known as the program cycle: every RPG program executes within an implied loop, which can apply the program to every record of a file. At that time each record would be compared to each line in the program, which would act upon the record, or not, based upon whether that line had an "indicator" turned "on" or "off" — from a set of logical variables numbered 01–99 for user-defined purposes, or other smaller sets based upon record, field, or report processing functions. The concept of level breaks and matching records is unique to the RPG II language, and was originally developed with card readers in mind.
Since the introduction of the IBM System/38 in 1979 most RPG programmers discontinued use of the cycle in favor of controlling program flow with standard looping constructs, although IBM has continued to provide backward compatibility for the cycle.

History

RPG is one of the few languages created for punched card machines which are still in common use today, because it has evolved considerably over time. It was originally developed by IBM in 1959. The name Report Program Generator was descriptive of the purpose of the language: generation of reports from data files, including matching record and sub-total reports.
FOLDOC accredits Wilf Hey with work at IBM that resulted in the development of RPG.
FARGO was the predecessor to RPG on the IBM 1401. Both languages were intended to facilitate ease of transition for IBM tabulating machine unit record equipment technicians to the then-new computers.
Tab machine technicians were accustomed to plugging wires into control panels to implement input, output, control and counter operations. Tab machines programs were executed by impulses emitted in a machine cycle; hence, FARGO and RPG emulated the notion of the machine cycle with the program cycle. RPG was superior to and rapidly replaced FARGO as the report generator program of choice.
The alternative languages generally available at the time were Assembler, COBOL or FORTRAN. COBOL was a natural language-like business oriented language and FORTRAN was a language that facilitated mathematical applications. Other languages of the era included ALGOL and Autocoder and a few years later PL/I. Assembler and COBOL were more common in mainframe business operations and RPG more commonly used by customers who were in transition from tabulating equipment.

Language evolution

RPG II was introduced with the System/3 series of computers. It was later used on System/32, System/34, and System/36, with an improved version of the language. RPG II was also available for larger systems, including the IBM System/370 mainframe running DOS/VSE. ICL also produced a version on its VME/K operating system.
RPG III was created for the System/38 and its successor the AS/400. RPG III significantly departed from the original language, providing modern structured constructs like IF-ENDIF blocks, DO loops, and subroutines. RPG III was also available for larger systems including the IBM System/370 mainframe running OS/VS1. It was also available from Unisys for the VS/9 operating system running on the UNIVAC Series 90 mainframes.
DE/RPG or Data Entry RPG was exclusively available on the IBM 5280 series of data-entry workstations in the early '80s. It was similar to RPG III but lacking external Data Descriptions to describe data like on the System/38 and its successors. Instead, the DDS part had to be included into the RPG source itself.
RPG/400 was effectively RPG III running on AS/400. IBM renamed the RPG compiler as "RPG/400" but at the time of its introduction it was identical to the RPG III compiler on System/38. Virtually all IBM products were rebranded as xxx/400 and the RPG compiler was no exception. RPG III compiled with the RPG/400 compiler offered nothing new to the RPG III language until IBM began development of new operation codes, such as SCAN, CAT and XLATE after several years of AS/400 availability. These enhancements to RPG III were not available in the System/38 version of RPG III.

RPG IV

RPG IV, a.k.a. RPGLE or ILE RPG,) was released in 1994 as part of the v3r2 release of OS/400. With the release of RPG IV, the RPG name was officially no longer an initialism. RPG IV offered a greater variety of expressions within its Extended Factor-2 Calculation Specification and, later in life, its free-format Calculation Specifications and Procedure syntax. RPG IV is the only version of RPG supported by IBM in its long line of successors to the AS/400 system.
In 2001, with the release of OS/400 V5R1, RPG IV offered greater freedom for calculations than offered by the Extended Factor-2 Calculation Specification: a free-format text-capable source entry, as an alternative to the original column-dependent source format. The "/FREE" calculation did not require the operation code to be placed in a particular column; the operation code is optional for the EVAL and CALLP operations; and syntax generally more closely resembles that of mainstream, general-purpose programming languages. Until November 2013, the free format applied exclusively to the calculation specifications. With the IBM i V7R1 TR7 upgrade to the language, the "/free" and "/end-free" calculations are no longer necessary, and the language has finally broken the ties to punched cards.
Today, RPG IV is a more robust language. While editing can still be done via SEU, the simple green screen editor, a long progression of tools has been developed over time. Some of these have included Code/400 and Visual Age for RPG, which were developed by IBM. Currently the preferred editing platform is IBM's Websphere Development Studio client, now named RDi, which is a customized implementation of Eclipse. Eclipse, and therefore RDi, runs primarily on personal computers and other devices. IBM is continually extending its capabilities and adding more built-in functions. It has the ability to link to Java objects, and i5/OS APIs; it can be used to write CGI programs with the help of IBM's Cgidev2 Web toolkit, the Cozzi , , the RPG Toolbox, and other commercial Web-enabled packages. Even with the changes, it retains a great deal of backward compatibility, so an RPG program written 37 years ago could run today with little or no modification.
The SQL precompiler allows current RPG developers to take advantage of IBM's cost-based SQE. With the traditional F-Spec approach a developer had to identify a specific access path to a data set, now they can implement standard embedded SQL statements directly in the program. When compiled, the SQL precompiler transforms SQL statements into RPG statements which call the database manager programs that ultimately implement the query request.
In 2006, OS/400 was renamed i5/OS to correspond with the new IBM System i5 branding initiative; the 5 was later dropped in favor of just System i. In March 2008 i5/OS was renamed IBM i as part of the Power Systems consolidation of System i and System p product lines. The new Power Systems also adopt more mainstream version numbers, substituting 6.1 for the twenty-year-old V1R1M0 notation. The latest release is now referred to as IBM i 7.3 and fully supports the RPG IV language, as well as many others.
The RPG IV language is based on the EBCDIC character set, but also supports UTF-8, UTF-16 and many other character sets. The threadsafe aspects of the language are considered idiosyncratic by some as the compiler team has addressed threads by giving each thread its own static storage, rather than make the RPG run-time environment re-entrant. This has been noted to muddle the distinction between a thread and a process.

RPG Open Access

RPG Edition was launched in 2010 with the full name of Rational Open Access.
Upon its launch, IBM i chief architect Steve Will described it as an "enhancement that's really an architectural change that opens RPG up to whole new uses, and it’s being brought to the community by IBM i, our friends in IBM Rational, and several ISVs who’ve been working with IBM to hit the ground running. The idea behind the new capability is to remove the necessity for all output from RPG programs to be forced into the limited 5250 data stream. Instead, it can be passed to handlers, which can transform it in any way the handler wants. Wrap it in XML; put it in the format required by a pervasive device; make a Web page that doesn’t look like a 5250 screen; whatever you want."
While many companies were skeptical about the usefulness of Open Access, it gained immediate acceptance from some others, including Profound Logic corporation, which rapidly incorporated it into their products aimed at opening the i, and hired Scott Klement, a product developer well known for creating open-source tools to access the technology.

Data types

RPG supports the following data types.
Note:
The character in the data type column is the character that is encoded on the Definition Specification in the column designated for data type. To compare, in a language like C where definitions of variables are free-format and would use a keyword such as int to declare an integer variable, in RPG, a variable is defined with a fixed-format Definition Specification. In the Definition Specification, denoted by a letter D in column 6 of a source line, the data type character would be encoded in column 40. Also, if the data type character is omitted, that is, left blank, the default is A if no decimal positions are specified, P when decimal positions are specified for stand-along fields, and S when decimal positions are specified within a data structure.
Data typeNameLengthDescription
AAlphanumeric character1 to 16,773,104 bytes
1 to 16,773,100 bytes
Alphanumeric character
BBinary numeric1 byte
2 byte
4 bytes
8 bytes
Signed binary integer
CUCS-2 character1 to 8,386,552 characters
1 to 8,386,550 characters
16-bit UCS-2 character
DDate10 bytesDate: year, month, day
FFloating point numeric4 bytes
8 bytes
Signed binary floating-point real
GGraphic character1 to 8,386,552 characters
1 to 8,386,550 characters
16-bit graphic character
IInteger numeric1 byte
2 bytes
4 bytes
8 bytes
Signed binary integer
NCharacter indicator1 byte'1' = TRUE
'0' = FALSE
OObjectSize undisclosedObject reference
PPacked decimal numeric1 to 63 digits,
2 digits per byte plus sign
Signed fixed-point decimal number with integer and fraction digits
SZoned decimal numeric1 to 63 digits,
1 digit per byte
Signed fixed-point decimal number with integer and fraction digits
TTime8 bytesTime: hour, minute, second
UInteger numeric1 byte
2 bytes
4 bytes
8 bytes
Unsigned binary integer
ZTimestamp26 bytesDate and time:
year, month, day, hour, minute, second, microseconds
*Basing-Pointer
Procedure-Pointer
System-Pointer
16 bytesAddress to Data
Address to Activated Procedure
Address to Object

Example code

The following program receives a customer number as an input parameter and returns the name and address as output parameters.
This is the most primitive version of RPG IV syntax. The same program is shown later with gradually more modern versions of the syntax and gradually more relaxed rules.

* Historically RPG was columnar in nature, though free-formatting
* was allowed under particular circumstances.
* The purpose of various lines code are determined by a
* letter code in column 6.
* An asterisk in column 7 denotes a comment line
* "F" specs define files and other i/o devices
F ARMstF1 IF E K Disk Rename
* "D" specs are used to define variables
D pCusNo S 6p
D pName S 30a
D pAddr1 S 30a
D pAddr2 S 30a
D pCity S 25a
D pState S 2a
D pZip S 10a
* "C" specs are used for executable statements
* Parameters are defined using plist and parm opcodes
C *entry plist
C parm pCusNo
C parm pName
C parm pAddr1
C parm pAddr2
C parm pCity
C parm pState
C parm pZip
* The "chain" command is used for random access of a keyed file
C pCusNo chain ARMstF1
* If a record is found, move fields from the file into parameters
C if %found
C eval pName = ARNm01
C eval pAddr1 = ARAd01
C eval pAddr2 = ARAd02
C eval pCity = ARCy01
C eval pState = ARSt01
C eval pZip = ARZp15
C endif
* RPG makes use of switches. One switch "LR" originally stood for "last record"
* LR flags the program and its dataspace as removable from memory
C eval *InLR = *On

The same program using free calculations available starting in V5R1:

* "F" specs define files and other i/o devices
FARMstF1 IF E K Disk Rename
* "D" specs are used to define variables and parameters
* The "prototype" for the program is in a separate file
* allowing other programs to call it
/copy cust_pr
* The "procedure interface" describes the *ENTRY parameters
D getCustInf PI
D pCusNo 6p 0 const
D pName 30a
D pAddr1 30a
D pAddr2 30a
D pCity 25a
D pState 2a
D pZip 10a
/free
// The "chain" command is used for random access of a keyed file
chain pCusNo ARMstF1;
// If a record is found, move fields from the file into parameters
if %found;
pName = ARNm01;
pAddr1 = ARAd01;
pAddr2 = ARAd02;
pCity = ARCy01;
pState = ARSt01;
pZip = ARZp15;
endif;
// RPG makes use of switches. One switch "LR" originally stood for "last record"
// LR actually flags the program and its dataspace as removable from memory.
*InLR = *On;
/end-free

Assume the ARMSTF1 example table was created using the following SQL Statement:

create table armstf1
,
arname char,
aradd1 char,
aradd2 char,
arcity char,
arstte char,
arzip char)

The same program using free calculations and embedded SQL:

* RPG IV no longer requires the use of the *INLR indicator to terminate a program.
* by using the MAIN keyword on the "H" spec, and identifying the "main" or
* entry procedure name, the program will begin and end normally without using the
* decades-old RPG Cycle and instead a more "C like" begin and end logic.
H MAIN
* "D" specs are used to define variables and parameters
* The "prototype" for the program is in a separate file
* allowing other programs to call it
/copy cust_pr
* The "procedure interface" describes the *ENTRY parameters
P getCustInf B
D getCustInf PI
D pCusNo 6p 0 const
D pName 30a
D pAddr1 30a
D pAddr2 30a
D pCity 25a
D pState 2a
D pZip 10a
/free
exec sql select arName, arAddr1, arAdd2, arCity, arStte, arZip
into :pName, :pAddr1, :pAddr2, :pCity, :pState, :pZip
from ARMstF1
where arCNum = :pCusNo
for fetch only
fetch first 1 row only
optimize for 1 row
with CS;
/end-free
P GetCustInf E

As of V7R1 of the operating system, the above program would not necessarily need the prototype in a separate file, so it could be completely written as:

H main
D ARMSTF1 E DS
P GetCustInf B
D GetCustInf PI extpgm
D inCusNo like const
D outName like
D outAddr1 like
D outAddr2 like
D outCity like
D outState like
D outZip like
/free
exec sql select arName, arAdd1, arAdd2, arCity, arStte, arZip
into :outName, :outAddr1, :outAddr2, :outCity, :outState,
:outZip
from ARMSTF1
where arCNum = :inCusNo
fetch first 1 row only
with CS
use currently committed;
/end-free
P GetCustInf E

Lastly, if you apply the compiler PTFs related Technology Refresh 7 to your 7.1 operating system, then the above program can be coded completely in free-form, as follows:

ctl-opt main;
dcl-ds ARMSTF1 ext end-ds;
dcl-proc GetCustInf;
dcl-pi *n extpgm;
inCusNo like const;
outName like;
outAddr1 like;
outAddr2 like;
outCity like;
outState like;
outZip like;
end-pi;
exec sql select arName, arAdd1, arAdd2, arCity, arStte, arZip
into :outName, :outAddr1, :outAddr2, :outCity, :outState,
:outZip
from ARMSTF1
where arCNum = :inCusNo
fetch first 1 row only
with CS
use currently committed;
return;
end-proc;

Platforms

As stated above, the RPG programming language originally was introduced by IBM for their proprietary 1401, /360, /3, /32, /34, /36, /38 AS/400 and System i systems. There have also been implementations for the Digital VAX, Sperry Univac BC/7, Univac system 80, Siemens BS2000, Burroughs B700, B1700, Hewlett Packard HP 3000, ICL 2900 series, Honeywell 6220 and 2020, Four-Phase IV/70 and IV/90 series, Singer System 10 and WANG VS, as well as miscellaneous compilers and runtime environments for Unix-based systems, such as Infinite36, and PCs.
The latest platform to receive an RPG compiler is Windows.Net through the WINRPG compiler which was purchased by I-Net System Solutions in March 2013. This version contains extensions to RPG IV beyond that of the base IBM compiler. These extensions provide Microsoft Windows and.NET Framework hooks in the Native and System/36 environment, as well as the ability to port DB/2 files to Microsoft Access and Microsoft SQL Server databases via Open Database Connectivity.
Open-source compiler implementations, targeting platforms such as Microsoft Dot-NET and in various stages of completion, can also
be found on various repository sites such as "github.com."
RPG II applications are still supported under the IBM z/VSE and z/OS operating systems, HP MPE operating system on HP3000 and the OpenVMS operating system on VAX, Alpha, and Unisys MCP and now Windows Servers and Desktops under I-Net RPG.