RPG/400

More Efficient Parameter Passing

This tip will provide cleaner and shorter code. When calling a program, the result field can contain the name of a parameter list to pass to the calling program. Instead of getting a program that resembles:

 ....1....+....2....+....3....+....4....+....5....+....6....+ C CALL 'PROG1' C PARM P1 C PARM P2 C PARM P3 C PARM P4 C PARM P5 C* C CALL 'PROG2' C PARM P3 C PARM P4 

You will get a program that resembles:

 C PLIST1 PLIST C PARM P1 C PARM P2 C PARM P3 C PARM P4 C PARM P5 C* C PLIST2 PLIST C PARM P3 C PARM P4 C* C CALL 'PROG1' PLIST1 C CALL 'PROG2' PLIST2 ....1....+....2....+....3....+....4....+....5....+....6....+ 

- Michael Catalani

Default RPG Header Specifications

For RPG/400 programs, do not include an H-spec in the source member when you need special entries like an alternate collating sequence. Instead, create a character data area, DFTHSPEC, in library QRPG which is 80 characters long containing your installation's standard values for the currency symbol, date format and edit, decimal notation, etc. The "H" in column 6 is not required. This allows global changes without changing a source member.

The DFTHSPEC data area can be overridden by including an H-spec in the source. It will also be overridden if a data area, RPGHSPEC, is found anywhere in the library list. This allows manipulating the library list to select an H-spec.

To activate the DEBUG option automatically for testing: leave column 15 blank in DFTHSPEC or production library versions of RPGHSPEC, but place a '1' in column 15 in your test library versions of RPGHSPEC. When you are through testing the program in the test library, simply recompile the program to the production library.

- Douglas Handy

Limitation of RPG DO Loop

When using DO loops in RPG/400 with an increment value specified in Factor 2 of the ENDDO statement, make sure the increment value is always positive. The RPG/400 Reference manual description of the ENDDO opcode increment simply says:

"It can be positive or negative, requires no decimal positions, and can be one of the following: an array element, table name, data structure, field, named constant, or numeric literal. If Factor 2 is not specified on the ENDDO, then by default, the increment increases by 1."

While that is true, it is misleading at best. Most languages would let you code the following to decrement through an array:

 ....1....+....2....+....3....+....4....+....5....+....6....+ C 100 DO 1 X C [code] C END -1 

But RPG would simply set X to 100, decide 100 is greater than the limit value, 1, and never execute the loop even once.

- Douglas Handy

Quick Removal of RPG Serialization Numbers

To remove obsolete serialization from columns 1-5 in RPG source members, key LLT5 in the line command area of the top line in SEU. On the last statement prior to compile time tables or arrays, enter a matching LLT. This will shift the entire block left five bytes while allowing truncation. The screen will light up like a Christmas tree with syntax errors. Repeat the process with RRT5 and RRT to restore the proper column positions. The T suffix is optional since no truncation will occur, and the 5 may be specified at either the top or bottom of the block.

- Douglas Handy

Create Your Own Edit Codes

One of the better features of OS/400 is the ability to create your own edit codes. Three of the ones that I use are one for a modified "Y" edit code, a social security edit code and a telephone number edit code. The three edit code commands are Create Edit Description (CRTEDTD), Delete Edit Description (DLTEDTD), and Display Edit Description (DSPEDTD). There is no command to change an edit code description.

To create an edit code of your own, you must first delete one of the edit codes, numbered 5-9, that IBM provides as samples (DLTEDTD 5). Then you can replace that code with one of your own. I will show you the three that I use, but I will use a lowercase "x" for the edit code. You can choose your own. The character 'b' in the following commands represents a blank.

 CRTEDTD EDTD(x) INTMASK('b0/bb/bb') DECPNT(N) ZEROBAL(*NO) + TEXT('Similar to Y, but no print if date = 000000') CRTEDTD EDTD(x) INTMASK(' - - ') DECPNT(N) + FILLCHAR(0) + ZEROBAL(*YES) TEXT('Social Security Edit Code') CRTEDTD EDTD(x) INTMASK('bbb)&bbb-bbbb') DECPNT(N) + ZEROBAL(*NO) LFTCNS('(') TEXT('Telephone edit code') 

- Tim Johnston

Take Care of Decimal Data Errors

If you are plagued with decimal data errors, especially if you have migrated from a S/36, and you need a quick fix, try out this technique. You can compile your program with the compiler option, Ignore Decimal Data Error (IGN-DECERR), set to *YES. Not only will it ignore the error, but if you are updating files it will initalize the numeric fields to 0. Therefore, you should make sure your data errors are being caused by blank characters in a numeric field, and not because of some sort of data corruption. If you do have a file with blanks in some numeric fields that you would like to clean up quickly, a quick program to do this would be:

 ....1....+....2....+....3....+....4....+....5....+....6....+ FFILENAMEUP E DISK C UPDATFILE10 FILENAME=Name of the File FILE10=Format Name Make sure to compile with IGNDECERR(*YES). 

- Michael Catalani

Use the CLEAR Operation to Reduce Program Statements

The CLEAR opcode can save a lot of time and headaches. When writing a program that adds records to a file, don't code a series of MOVE *BLANK or *ZERO lines; instead, code a single CLEAR operation with record name in Factor 2. The CLEAR opcode automatically sets all alpha fields to blanks and all numeric fields to zero. In the future, you will only need to recompile the program if you add fields to the record. Using the old method, you would have to initialize each of the new numeric fields to 0, or risk getting decimal data errors.

Old Method:

 ....1....+....2....+....3....+....4....+....5....+....6....+C @INIT BEGSR CUCST# C Z-ADDO CUNAME C MOVEL*BLANKS CUADR1 C MOVEL*BLANKS CUADR2 C MOVEL*BLANKS CUCITY C MOVEL*BLANKS CUSTTE C MOVEL*BLANKS CUZIP C ENDSR 

New Method:

 C @INIT BEGSR C CLEARCMCUST10 C ENDSR ....1....+....2....+....3....+....4....+....5....+....6....+ 

- Micheal Catalani

Reduce I/O During Record Validation

For the I/O conscious programmer: In an interactive program where you would take the user's input as a key and chain out to a file for validation, I suggest using the SETLL operation with an indicator in position 58. If the indicator is turned on, an exact match was found. This technique works well in V1R3M0 machines. It also works well in V2R1Mx machines if the file is not a multi-format logical. If it is, use CHAIN instead.

- Charlie A. Shanks

Keep SETLL in Auxiliary Programs

When creating RPG programs that only need a SETxx opcode to position a pointer in the file(s) you can first use a CL program to override the database file with SHARE(*YES), call your RPG program from the CL, create a second RPG program which is called from the first RPG passing the key values as parameters which will position the pointer in the file, then return. By eliminating any opcode that changes the position of the pointer (other than a READ, READP) from within the main program, the RPG program will be able to block records during input and therefore reduce I/O count on the input files. This can reduce time and overhead considerably.

 CL Main RPG Aux RPG OVRDBF CALL... SETLL SHARE(*YES) READ SETGT CALL... (Blocking allowed) 

- Greg Leister

Avoid Record Lock Errors

To prevent RPG programs running in batch from blowing up on a record lock, use the error indicator on the CHAIN operation (pos. 56-57) as follows:

 ... 1 ...+... 2 ...+... 3 ...+... 4 ...+... 5 ...+... 6 ...+ C *IN59 DOUEQ'0' C Key CHAINFormat 5059 C ENDDO 

Indicator 59 will only be set on if there is an error with retrieving the format-it does not replace or interfere with the no record found indicator (50). In interactive programs you would not want to use the do-until loop; however, using the error indicator is a good way to let a user know that someone else is using the record.

The error indicator is also available with any of the read operations, but the file has to be repositioned if it is set on.

- William MacKenzie Picou

Use CAT Instead of MOVEL

The RPG operation code MOVEL is restricted to moving eight characters in Factor 2 to the Result field. Use the CAT opcode to move left up to 16 characters in one operation, as in:

 ... 1 ...+... 2 ...+... 3 ...+... 4 ...+... 5 ...+... 6 ...+ C '12345678'CAT '90ABCDEF'RESULT 

- Frank Kolmann

RTNDTA Keyword Trick

The DDS keyword Return Data (RTNDTA) can eliminate the need to create separate variable names in display files used for database file updates. The display file can reference the database fields instead. No need to worry about field attributes matching between the database file and the display file. I don't have to code any MOVEs or Z-ADDs, and my program size is smaller as there are not as many variables.

Here's how it's done.

DDS coding:

Place the RTNDTA keyword at the record level on the appropriate record of the display file. Use the database field names and reference the database file. This will bring the database fields into the display file automatically when you retrieve the database record.

RPG coding:

1. CHAIN and release the database record.

2. Display the screen.

3. Read the screen looking for changes and, if changes have been made, CHAIN again to the database record.

4. Reread the display format using the READ operation. This is what makes this technique work. Since I CHAINed again to the database record, the internal representation of the display file record is overlaid and any changes to the display record are lost. But by reading the display record again (with the RTNDTA keyword), you can retrieve the original values of the screen.

5. Update the database record.

The DDS manual is a good reference where this keyword is concerned.

- Andrew Krueger