Supported Opcodes (Fixed-Format C-Spec)

Supported RPG Features

Bit Manipulation (BITON / BITOFF / TESTB)

Factor 2 selects which bits BITON, BITOFF, and TESTB act on. Bits are numbered 0 - 7 from the leftmost (high-order) bit. Factor 2 accepts any of:

• A bit-number literal - one to eight digits 0 - 7 in apostrophes, e.g. '05' selects bits 0 and 5.
• A hexadecimal literal - X'nn', whose byte value is the mask directly, e.g. X'04' selects bit 5.
• A one-byte character field - the field's byte value is the mask.
• A named constant holding either a bit-number character value or a hex value. A named constant behaves exactly as the equivalent literal does.

A Factor 2 that resolves to none of these (an undefined name) is rejected with TRN1048 rather than silently treated as a no-op.

Hexadecimal Literals

Outside the bit operations above (where X'nn' is a raw mask), a hexadecimal literal has the same meaning as the corresponding character literal, exactly as on IBM i: its digits are the character data's raw bytes, and they are never CCSID-converted. Those bytes are read under the target field's own CCSID, not the job's - a hexadecimal literal names bytes, and what they mean is the page of the field they land in. X'E1' is therefore ß in a CCSID(850) field and á in a CCSID(819) one, while in a field with no CCSID keyword it is read under the job page as before. X'C1C2C3C4C5' is 'ABCDE'; a field initialized with INZ(X'04') holds the byte x'04', so supplying that field as a bit-op mask selects bit 5, a zone move from it transfers zone 0, and comparing it to a blank (x'40') is false. DSPLY of data containing a byte with no printable glyph shows a blank in that position - matching how a 5250 display renders it - while the underlying data keeps the exact byte value.

EVAL Operation Extenders

EVAL (and free-format assignment) accepts these operation extenders, which control the precision of the numeric assignment:

• (H) - half-adjust (round) the final result to the target's decimal positions instead of truncating.
• (M) - use maximum (default) intermediate precision rules.
• (R) - keep the decimal positions of every intermediate result from being reduced below the result field's decimal places. Without (R), the default precision rules can shrink an intermediate's decimals (even to zero) when the integer part is large, losing fractional precision; (R) floors each intermediate's scale at the result's decimal places. For example, with eval(r) result = a * b where result has 4 decimals, the a * b intermediate keeps at least 4 decimals. When combined with (H) (i.e. eval(rh)), intermediates keep one extra decimal beyond the result so the final half-adjust rounds on the correct digit.

Float Operands in Arithmetic Expressions

An arithmetic expression (+, -, *, /) is typed by its operands, and float is contagious: if either operand is FLOAT, the expression is a float expression, evaluated in floating point rather than fixed point. A mixed expression such as floatField * packedField is therefore float - the packed operand widens, rather than pulling the result down to a decimal type.

The result type is what a directly-nested %CHAR renders, so a float expression prints its float external representation (scientific notation), not a fixed-decimal image:

dcl-s a float(8) inz(2.5);
dcl-s b float(8) inz(4.0);
dsply %char(a * b);          // +1.000000000000000E+001

Assigning the expression to a fixed-point target is unaffected - the value converts on the store, so packedResult = a * b with two decimals gives 10.00 as usual. Exponentiation (**) stays on the fixed-point path.

CLEAR and RESET on Record Formats

When CLEAR or RESET names a record format of an externally described DISK file, the optional *NOKEY and *ALL qualifiers decide which DDS fields are affected:

DDS fields of a DISK record format are implicitly classified as input-only for CLEAR's and RESET's purposes, so plain CLEAR / RESET (and *NOKEY alone) do nothing. *ALL forces every field to be reset; *NOKEY then excludes the key columns. When both qualifiers are specified *NOKEY must precede *ALL. CLEAR resets each field to its data-type default (blanks for character, zero for numeric); RESET restores each field to its initialization value, which is the same data-type default when no *INZSR or DDS default applies.

DSPLY Numeric Rendering Contract

When DSPLY's factor 1 is a non-float numeric field (packed, zoned, integer, unsigned), the value is rendered to the workstation as its raw digit image:

• The declared precision in right-aligned digit positions - leading zeros are suppressed to blanks. The units (rightmost) position is preserved so a zero balance still shows a single 0.
• No decimal point, regardless of the field's scale - the digits are shown unedited (a PACKED(7:2) holding 12345.67 displays 1234567).
• No leading sign. A negative value appends a trailing - as the rightmost character, making its image one position wider than the declared precision.
• Unsigned fields render their full unsigned magnitude. An UNS(20) field is never negative, so it never carries a trailing sign - even a value at or above 2^63 (whose bit pattern would be a negative signed integer) shows its full unsigned decimal (UNS(20) at its maximum displays all 20 digits).

Examples:

(␣ denotes a literal space.) Character fields pass through unchanged and keep their full declared length: a fixed-length CHAR(n) field displays all n characters, including trailing blanks, matching IBM i (a CHAR(20) holding 'Hello' displays Hello followed by 15 blanks). A varying (VARCHAR) field displays only the characters it currently holds. An indicator operand (*INxx, or a named indicator) displays its one-character value - 1 when on, 0 when off - because an indicator is a one-byte character field. Wrapping a numeric in %CHAR(...) deliberately bypasses this contract and produces the minimal-width decimal string %CHAR is defined to return.

DSPLY in WORKSTN Programs

When a program that owns a WORKSTN display file runs DSPLY, rpgc suspends the display and presents a full-screen Display Program Messages panel on a 24×80 screen:

• The screen is cleared first - the previous format does not bleed through.
• Row 1 holds the centered Display Program Messages title.
• Row 3 holds a job-start line (Job .../.../... started on ... at ... in subsystem ...). Because a JVM process is not an IBM i job, its identifiers (job number, user, device, time) reflect the rpgc runtime session rather than a native job, so this line's values differ from run to run.
• Each DSPLY message renders from row 4 downward as DSPLY <value>, using the value contract described above.
• Row 19 shows Press Enter to continue. and row 23 shows F3=Exit F12=Cancel. The program resumes when the operator presses Enter.

The panel pauses for that acknowledgment only when a display format has already been shown at the workstation this session (a display file with RSTDSP(*NO) has displayed a format). A DSPLY issued before any format is shown, or in a batch/no-display program, posts its message and continues without pausing, so non-interactive runs never block waiting for an operator.

Successive DSPLY messages accumulate for the whole program call: each panel lists every message issued so far - one per line from row 4 downward - not just the latest. This holds even when an EXFMT of a display format runs between two DSPLY operations; the second panel still shows the earlier messages above the new one. The accumulated queue is reset only when the program ends or transfers control to another program (EXTPGM call/return). If the content area fills, the panel scrolls to the most recent messages.

When an EXFMT drawn with OVERLAY follows a dismissed panel, the underlying display-file image is restored and only the Display Program Messages title bleeds through at row 1 - the panel's body (job-start line, messages, prompt, and key legend) is cleared rather than left under the new format.

DUMP Opcode

The DUMP opcode emits a formatted program dump listing all program fields, indicators, and their current values. On IBM i, this output goes to the QPPGMDMP spool file; on the JVM target, it is written to System.err (standard error) so it does not interfere with DSPLY output on stdout.

Free-format syntax:

DUMP(A);            // always dump
DUMP;               // dump (always executes - DEBUG H-spec has no effect)

Fixed-format syntax:

The optional identifier (a character literal, field name, or named constant) tags the dump output so multiple DUMP calls in a program can be distinguished.

The (A) operation extender forces the dump to execute regardless of the DEBUG control-spec keyword. The DEBUG keyword has no effect in Triton RPG - DUMP always executes, whether or not (A) is specified.

SHTDN Opcode / %SHTDN BIF

The SHTDN opcode tests whether a controlled shutdown has been requested for the job, subsystem, or system. On IBM i, the HI resulting indicator (columns 71-72) is set ON when a shutdown is in progress. On the JVM target, no shutdown concept exists, so the indicator is always set to OFF.

Fixed-format syntax:

Free-format: Use the %SHTDN built-in function, which returns *ON if a shutdown has been requested, *OFF otherwise. On JVM, always *OFF.

if %shtdn;
  // perform cleanup
endif;