cobol-developer
/cobol-developer
Use when reading, writing, debugging, or modernizing COBOL programs — divisions and sections, data types (PIC clauses, COMP/COMP-3/DISPLAY), file handling (QSAM/VSAM/KSDS/RRDS), PERFORM and paragraph structure, COPY/REPLACE, embedded SQL (EXEC SQL), CICS programming (EXEC CICS), batch JCL integration, debugging (CEDF/Xpediter), and COBOL modernization patterns. Part of the ibm-mainframe skill family.
- Author
- joogy06 ↗
- Repository
- agent-foundry ↗
- Licence
- MIT
- Last repository change
- 31.07.2026
Assessment and security
- Files in this package
- 1
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 31.07.2026, without references to retired models or APIs.
MIT, open: use and modify while retaining author attribution.
What it does and when to use it
Use when reading, writing, debugging, or modernizing COBOL programs — divisions and sections, data types (PIC clauses, COMP/COMP-3/DISPLAY), file handling (QSAM/VSAM/KSDS/RRDS), PERFORM and paragraph structure, COPY/REPLACE, embedded SQL (EXEC SQL), CICS programming (EXEC CICS), batch JCL integration, debugging (CEDF/Xpediter), and COBOL modernization patterns. Part of the ibm-mainframe skill family.
Original author description
Use when reading, writing, debugging, or modernizing COBOL programs — divisions and sections, data types (PIC clauses, COMP/COMP-3/DISPLAY), file handling (QSAM/VSAM/KSDS/RRDS), PERFORM and paragraph structure, COPY/REPLACE, embedded SQL (EXEC SQL), CICS programming (EXEC CICS), batch JCL integration, debugging (CEDF/Xpediter), and COBOL modernization patterns. Part of the ibm-mainframe skill family.
How it works
- 1. Program Structure
- 2. Data Division Deep Dive
- 3. File Handling
- 4. Procedure Division Patterns
- 5. Embedded SQL (DB2)
- 6. CICS Programming
- 7. Batch Processing
- 8. Copybook Patterns
- 9. Debugging
- 10. Modernization
SKILL.md preview
---
name: cobol-developer
description: Use when reading, writing, debugging, or modernizing COBOL programs — divisions and sections, data types (PIC clauses, COMP/COMP-3/DISPLAY), file handling (QSAM/VSAM/KSDS/RRDS), PERFORM and paragraph structure, COPY/REPLACE, embedded SQL (EXEC SQL), CICS programming (EXEC CICS), batch JCL integration, debugging (CEDF/Xpediter), and COBOL modernization patterns. Part of the ibm-mainframe skill family.
---
# COBOL Developer Reference
Companion to `ibm-mainframe` (JCL/VSAM/utilities) and `db2-mainframe` (DB2 z/OS DBA). For Python mainframe connectivity see `python-enterprise-connectors`.
<HARD-RULE>
Always check SQLCODE after every EXEC SQL statement — unchecked SQLCODEs cause silent data corruption. At minimum check for SQLCODE = 0, +100 (not found), and negative values (errors).
</HARD-RULE>
<HARD-RULE>
Never use GO TO in new COBOL code — use structured PERFORM THRU patterns for maintainability. GO TO creates spaghetti control flow that is nearly impossible to debug or modify.
</HARD-RULE>
<HARD-RULE>
Always define numeric fields with explicit COMP-3 or COMP for calculations — DISPLAY numeric fields cause implicit conversions on every arithmetic operation, degrading performance significantly in batch loops.
</HARD-RULE>
<HARD-RULE>
Always initialize WORKING-STORAGE variables with VALUE clauses or INITIALIZE verb — uninitialized fields contain unpredictable data from previous program executions in CICS or residual memory in batch.
</HARD-RULE>
---
## 1. Program Structure
```cobol
IDENTIFICATION DIVISION.
PROGRAM-ID. ORDPROC.
AUTHOR. MAINTENANCE TEAM.
DATE-WRITTEN. 2026-03-31.
*================================================================*
* ORDER PROCESSING - BATCH PROGRAM *
*================================================================*
ENVIRONMENT DIVISION.
CONFIGURATION SECTION.
SPECIAL-NAMES.
DECIMAL-POINT IS COMMA.
INPUT-OUTPUT SECTION.
FILE-CONTROL.
SELECT ORDER-FILE
ASSIGN TO ORDIN
ORGANIZATION IS SEQUENTIAL
FILE STATUS IS WS-ORDER-FS.
SELECT REPORT-FILE
ASSIGN TO RPTOUT
ORGANIZATION IS SEQUENTIAL
FILE STATUS IS WS-REPORT-FS.
SELECT CUSTOMER-FILE
ASSIGN TO CUSTDB
ORGANIZATION IS INDEXED
ACCESS MODE IS RANDOM
RECORD KEY IS CUST-KEY
FILE STATUS IS WS-CUST-FS.
DATA DIVISION.
FILE SECTION.
FD ORDER-FILE
RECORDING MODE IS F
RECORD CONTAINS 200 CHARACTERS.
01 ORDER-RECORD.
05 ORD-ID PIC 9(10).
05 ORD-CUST-ID PIC 9(8).
05 ORD-DATE PIC X(10).
05 ORD-AMOUNT PIC S9(11)V99 COMP-3.
05 ORD-STATUS PIC X(1).
88 ORD-PENDING VALUE 'P'.
88 ORD-SHIPPED VALUE 'S'.
88 ORD-CANCELLED VALUE 'C'.
05 FILLER PIC X(163).
FD REPORT-FILE
RECORDING MODE IS F
RECORD CONTAINS 132 CHARACTERS.
01 REPORT-RECORD PIC X(132).
WORKING-STORAGE SECTION.
01 WS-FILE-STATUS.
05 WS-ORDER-FS PIC XX VALUE SPACES.
05 WS-REPORT-FS PIC XX VALUE SPACES.
05 WS-CUST-FS PIC XX VALUE SPACES.
01 WS-FLAGS.
05 WS-EOF-FLAG PIC X(1) VALUE 'N'.
88 END-OF-FILE VALUE 'Y'.
88 NOT-END-OF-FILE VALUE 'N'.
01 WS-COUNTERS.
05 WS-READ-COUNT PIC 9(9) COMP VALUE 0.
05 WS-PROC-COUNT PIC 9(9) COMP VALUE 0.
05 WS-ERROR-COUNT PIC 9(9) COMP VALUE 0.
01 WS-TOTALS.
05 WS-TOTAL-AMOUNT PIC S9(15)V99 COMP-3
VALUE ZERO.
PROCEDURE DIVISION.
0000-MAIN.
PERFORM 1000-INITIALIZE
PERFORM 2000-PROCESS UNTIL END-OF-FILE
PERFORM 9000-TERMINATE
GOBACK
.
1000-INITIALIZE.
OPEN INPUT ORDER-FILE
OPEN OUTPUT REPORT-FILE
PERFORM 1100-READ-ORDER
.
1100-READ-ORDER.
READ ORDER-FILE
AT END SET END-OF-FILE TO TRUE
NOT AT END ADD 1 TO WS-READ-COUNT
END-READ
.
2000-PROCESS.
EVALUATE TRUE
WHEN ORD-PENDING
PERFORM 2100-PROCESS-PENDING
WHEN ORD-SHIPPED
PERFORM 2200-PROCESS-SHIPPED
WHEN OTHER
ADD 1 TO WS-ERROR-COUNT
END-EVALUATE
PERFORM 1100-READ-ORDER
.
9000-TERMINATE.
CLOSE ORDER-FILE REPORT-FILE
DISPLAY 'RECORDS READ: ' WS-READ-COUNT
DISPLAY 'RECORDS PROCESSED: ' WS-PROC-COUNT
DISPLAY 'ERRORS: ' WS-ERROR-COUNT
.
```
---
## 2. Data Division Deep Dive
### PIC Clauses
| PIC | Meaning | Example Value |
|-----|---------|---------------|
| `PIC X(20)` | Alphanumeric, 20 chars | `"JOHN DOE "` |
| `PIC 9(5)` | Numeric display, 5 digits | `00123` |
| `PIC S9(7)V99` | Signed, 7 integer + 2 decimal | `+1234567.89` |
| `PIC S9(7)V99 COMP-3` | Packed decimal (BCD) | 4 bytes |
| `PIC S9(9) COMP` | Binary (fullword) | 4 bytes |
| `PIC S9(18) COMP` | Binary (doubleword) | 8 bytes |
| `PIC S9(7)V99 COMP-1` | Single-precision float | 4 bytes |
| `PIC S9(18)V99 COMP-2` | Double-precision float | 8 bytes |
| `PIC Z(5)9` | Edited numeric (leading zeros suppressed) | ` 123` |
| `PIC $ZZ,ZZ9.99` | Edited with currency/commas | ` $1,234.56` |
### Storage Formats
| Usage | Storage | When to Use |
|-------|---------|-------------|
| DISPLAY | 1 byte/digit | External files, reports |
| COMP-3 (packed) | (n+1)/2 bytes | **All calculations and DB2 columns** |
| COMP (binary) | 2/4/8 bytes | Subscripts, counters, indexes |
| COMP-1 (float) | 4 bytes | Scientific calculations (rare) |
| COMP-2 (double) | 8 bytes | Scientific calculations (rare) |
### REDEFINES
```cobol
01 WS-DATE-FIELD.
05 WS-DATE-CCYYMMDD PIC 9(8).
05 WS-DATE-PARTS REDEFINES WS-DATE-CCYYMMDD.
10 WS-DATE-CC PIC 99.
10 WS-DATE-YY PIC 99.
10 WS-DATE-MM PIC 99.
10 WS-DATE-DD PIC 99.
```
### OCCURS (Arrays)
```cobol
01 WS-MONTHLY-TOTALS.
05 WS-MONTH-ENTRY OCCURS 12 TIMES
INDEXED BY WS-MONTH-IDX.
10 WS-MONTH-NAME PIC X(10).
10 WS-MONTH-AMT PIC S9(11)V99 COMP-3.
* Variable-length array
01 WS-ORDER-LINES.
05 WS-LINE-COUNT PIC 99 COMP.
05 WS-LINE-ITEM OCCURS 1 TO 50 TIMES
DEPENDING ON WS-LINE-COUNT.
10 WS-ITEM-SKU PIC X(10).
10 WS-ITEM-QTY PIC 9(5) COMP-3.
10 WS-ITEM-PRICE PIC S9(7)V99 COMP-3.
```
### 88-Level Condition Names
```cobol
01 WS-TRANSACTION-TYPE PIC X(2).
88 TXN-SALE VALUE 'SA'.
88 TXN-RETURN VALUE 'RT'.
88 TXN-EXCHANGE VALUE 'EX'.
88 TXN-VALID VALUE 'SA' 'RT' 'EX'.
* Usage:
IF TXN-VALID
PERFORM 2000-PROCESS-TXN
END-IF
SET TXN-SALE TO TRUE *> Sets WS-TRANSACTION-TYPE to 'SA'
```
---
## 3. File Handling
### File Status Codes
| Code | Meaning |
|------|---------|
| 00 | Success |
| 10 | End of file (AT END) |
| 22 | Duplicate key on WRITE |
| 23 | Record not found (READ/START/DELETE) |
| 35 | File not found (OPEN) |
| 39 | File attribute conflict |
| 41 | File already open |
| 42 | File already closed |
| 46 | Sequential READ with no valid next |
| 47 | READ on file not opened INPUT/I-O |
| 48 | WRITE on file not opened OUTPUT/I-O/EXTEND |
### VSAM KSDS (Indexed)
```cobol
FILE-CONTROL.
SELECT CUSTOMER-FILE
ASSIGN TO CUSTFILE
ORGANIZATION IS INDEXED
ACCESS MODE IS DYNAMIC
RECORD KEY IS CUST-ID
ALTERNATE RECORD KEY IS CUST-NAME
WITH DUPLICATES
FILE STATUS IS WS-CUST-FS.
* Random read
MOVE '00012345' TO CUST-ID
READ CUSTOMER-FILE
INVALID KEY
DISPLAY 'CUSTOMER NOT FOUND: ' CUST-ID
NOT INVALID KEY
PERFORM 3000-PROCESS-CUSTOMER
END-READ
* Sequential browse (START + READ NEXT)
MOVE '00010000' TO CUST-ID
START CUSTOMER-FILE KEY >= CUST-ID
INVALID KEY DISPLAY 'START FAILED'
END-START
PERFORM UNTIL WS-CUST-FS NOT = '00'
READ CUSTOMER-FILE NEXT
AT END SET END-OF-FILE TO TRUE
END-READ
IF NOT END-OF-FILE
PERFORM 3000-PROCESS-CUSTOMER
END-IF
END-PERFORM
* Update
READ CUSTOMER-FILE
INVALID KEY DISPLAY 'NOT FOUND'
END-READ
MOVE 'ACTIVE' TO CUST-STATUS
REWRITE CUSTOMER-RECORD
INVALID KEY DISPLAY 'REWRITE FAILED'
END-REWRITE
```
### SORT/MERGE
```cobol
SD SORT-FILE.
01 SORT-RECORD.
05 SORT-KEY PIC 9(10).
05 SORT-DATA PIC X(190).
PROCEDURE DIVISION.
* Simple sort
SORT SORT-FILE
ON ASCENDING KEY SORT-KEY
USING INPUT-FILE
GIVING OUTPUT-FILE
* Sort with INPUT/OUTPUT PROCEDURE
SORT SORT-FILE
ON ASCENDING KEY SORT-KEY
INPUT PROCEDURE IS 5000-SELECT-RECORDS
OUTPUT PROCEDURE IS 6000-FORMAT-OUTPUT
5000-SELECT-RECORDS.
OPEN INPUT RAW-FILE
PERFORM UNTIL END-OF-RAW
READ RAW-FILE AT END SET END-OF-RAW TO TRUE
NOT AT END
IF RAW-STATUS = 'A'
MOVE RAW-RECORD TO SORT-RECORD
RELEASE SORT-RECORD
END-IF
END-READ
END-PERFORM
CLOSE RAW-FILE
.
6000-FORMAT-OUTPUT.
OPEN OUTPUT FINAL-FILE
PERFORM UNTIL END-OF-SORT
RETURN SORT-FILE
AT END SET END-OF-SORT TO TRUE
NOT AT END
PERFORM 6100-WRITE-FORMATTED
END-RETURN
END-PERFORM
CLOSE FINAL-FILE
.
```
---
## 4. Procedure Division Patterns
### STRING / UNSTRING
```cobol
* STRING — concatenate
STRING WS-LAST-NAME DELIMITED BY SPACES
', ' DELIMITED BY SIZE
WS-FIRST-NAME DELIMITED BY SPACES
INTO WS-FULL-NAME
WITH POINTER WS-STR-PTR
ON OVERFLOW DISPLAY 'NAME TOO LONG'
END-STRING
* UNSTRING — split
UNSTRING WS-CSV-LINE DELIMITED BY ','
INTO WS-FIELD1
WS-FIELD2
WS-FIELD3
WITH POINTER WS-UNS-PTR
TALLYING IN WS-FIELD-COUNT
ON OVERFLOW DISPLAY 'TOO MANY FIELDS'
END-UNSTRING
```
### INSPECT
```cobol
* Count occurrences
INSPECT WS-INPUT-STRING
TALLYING WS-COUNT FOR ALL 'A'
* Replace characters
INSPECT WS-INPUT-STRING
REPLACING ALL LOW-VALUES BY SPACES
* Convert (translate)
INSPECT WS-INPUT-STRING
CONVERTING 'abcdefghijklmnopqrstuvwxyz'
TO 'ABCDEFGHIJKLMNOPQRSTUVWXYZ'
```
### Reference Modification
```cobol
* Extract substring: identifier(start:length)
MOVE WS-FULL-DATE(1:4) TO WS-YEAR *> First 4 chars
MOVE WS-FULL-DATE(5:2) TO WS-MONTH *> Chars 5-6
MOVE WS-FULL-DATE(7:2) TO WS-DAY *> Chars 7-8
```
### CALL (Subprogram)
```cobol
* Static call (link-edited together)
CALL 'DATECONV' USING WS-INPUT-DATE
WS-OUTPUT-DATE
WS-RETURN-CODE
* Dynamic call (loaded at runtime — can CANCEL to free)
MOVE 'DATECONV' TO WS-PGM-NAME
CALL WS-PGM-NAME USING WS-INPUT-DATE
WS-OUTPUT-DATE
WS-RETURN-CODE
ON EXCEPTION
DISPLAY 'PROGRAM NOT FOUND: ' WS-PGM-NAME
MOVE 16 TO RETURN-CODE
END-CALL
* CANCEL (free dynamic program from memory)
CANCEL WS-PGM-NAME
```
---
## 5. Embedded SQL (DB2)
### Basic Patterns
```cobol
WORKING-STORAGE SECTION.
EXEC SQL INCLUDE SQLCA END-EXEC.
EXEC SQL INCLUDE DCLORDER END-EXEC.
01 WS-SQLCODE PIC S9(9) COMP.
01 HV-CUST-ID PIC S9(8) COMP.
01 HV-ORDER-DATE PIC X(10).
01 HV-AMOUNT PIC S9(11)V99 COMP-3.
01 HV-AMOUNT-IND PIC S9(4) COMP.
PROCEDURE DIVISION.
* Singleton SELECT
EXEC SQL
SELECT CUST_NAME, CUST_BALANCE
INTO :HV-CUST-NAME, :HV-CUST-BALANCE
FROM CUSTOMER
WHERE CUST_ID = :HV-CUST-ID
END-EXEC
EVALUATE SQLCODE
WHEN 0 CONTINUE
WHEN +100 DISPLAY 'CUSTOMER NOT FOUND'
WHEN OTHER PERFORM 9100-SQL-ERROR
END-EVALUATE
* INSERT
EXEC SQL
INSERT INTO ORDERS
(ORDER_ID, CUST_ID, ORDER_DATE, AMOUNT)
VALUES
(:HV-ORDER-ID, :HV-CUST-ID,
:HV-ORDER-DATE, :HV-AMOUNT)
END-EXEC
* NULL indicator handling
EXEC SQL
SELECT AMOUNT
INTO :HV-AMOUNT :HV-AMOUNT-IND
FROM ORDERS
WHERE ORDER_ID = :HV-ORDER-ID
END-EXEC
IF HV-AMOUNT-IND < 0
DISPLAY 'AMOUNT IS NULL'
END-IF
```
### Cursor Processing
```cobol
* Declare cursor
EXEC SQL
DECLARE CSR-ORDERS CURSOR FOR
SELECT ORDER_ID, ORDER_DATE, AMOUNT
FROM ORDERS
WHERE CUST_ID = :HV-CUST-ID
AND ORDER_DATE >= :HV-START-DATE
ORDER BY ORDER_DATE DESC
END-EXEC
* Open cursor
EXEC SQL OPEN CSR-ORDERS END-EXEC
IF SQLCODE NOT = 0
PERFORM 9100-SQL-ERROR
END-IF
* Fetch loop
PERFORM UNTIL SQLCODE NOT = 0
EXEC SQL
FETCH CSR-ORDERS
INTO :HV-ORDER-ID,
:HV-ORDER-DATE,
:HV-AMOUNT
END-EXEC
IF SQLCODE = 0
PERFORM 3100-PROCESS-ORDER
END-IF
END-PERFORM
IF SQLCODE NOT = +100
PERFORM 9100-SQL-ERROR
END-IF
* Close cursor
EXEC SQL CLOSE CSR-ORDERS END-EXEC
* Error handler
9100-SQL-ERROR.
DISPLAY 'SQL ERROR: SQLCODE=' SQLCODE
DISPLAY 'SQLERRMC=' SQLERRMC
MOVE 16 TO RETURN-CODE
EXEC SQL ROLLBACK END-EXEC
GOBACK
.
```
### Common SQLCODEs
| SQLCODE | Meaning | Action |
|---------|---------|--------|
| 0 | Success | Continue |
| +100 | Not found / end of cursor | Handle as business logic |
| -803 | Duplicate key on INSERT | Check for existing record |
| -805 | Package/plan not found | Rebind needed |
| -811 | Multiple rows from SELECT INTO | Add predicate or use cursor |
| -818 | Timestamp mismatch | Rebind with current DBRM |
| -904 | Unavailable resource | Retry or fail gracefully |
| -911 | Deadlock/timeout rollback | Retry transaction |
| -913 | Deadlock victim | Retry transaction |
---
## 6. CICS Programming
### Pseudo-Conversational Pattern
```cobol
* Main entry — pseudo-conversational
0000-MAIN.
EVALUATE EIBCALEN
WHEN 0
PERFORM 1000-FIRST-TIME
WHEN OTHER
PERFORM 2000-PROCESS-INPUT
END-EVALUATE
.
1000-FIRST-TIME.
INITIALIZE WS-COMMAREA
PERFORM 3000-SEND-MAP
EXEC CICS RETURN
TRANSID('ORDQ')
COMMAREA(WS-COMMAREA)
LENGTH(LENGTH OF WS-COMMAREA)
END-EXEC
.
2000-PROCESS-INPUT.
MOVE DFHCOMMAREA TO WS-COMMAREA
EXEC CICS RECEIVE MAP('ORDMAP')
MAPSET('ORDMAPS')
INTO(ORDMAPI)
END-EXEC
EVALUATE EIBAID
WHEN DFHENTER
PERFORM 2100-PROCESS-ORDER
WHEN DFHPF3
EXEC CICS RETURN END-EXEC
WHEN DFHPF12
EXEC CICS RETURN END-EXEC
WHEN OTHER
MOVE 'INVALID KEY' TO MSGO
PERFORM 3000-SEND-MAP
END-EVALUATE
.
3000-SEND-MAP.
EXEC CICS SEND MAP('ORDMAP')
MAPSET('ORDMAPS')
FROM(ORDMAPO)
ERASE
CURSOR
END-EXEC
EXEC CICS RETURN
TRANSID('ORDQ')
COMMAREA(WS-COMMAREA)
LENGTH(LENGTH OF WS-COMMAREA)
END-EXEC
.
```
### CICS Commands Reference
| Command | Purpose |
|---------|---------|
| `SEND MAP` / `RECEIVE MAP` | BMS screen I/O |
| `LINK PROGRAM` | Call subprogram (returns) |
| `XCTL PROGRAM` | Transfer control (no return) |
| `RETURN TRANSID` | Pseudo-conversational return |
| `READ / WRITE / REWRITE / DELETE` | VSAM file operations |
| `READQ TS / WRITEQ TS / DELETEQ TS` | Temporary Storage queue |
| `READQ TD / WRITEQ TD` | Transient Data queue |
| `START TRANSID` | Schedule async transaction |
| `RETRIEVE` | Get START data |
| `GETMAIN / FREEMAIN` | Memory allocation |
| `ASKTIME / FORMATTIME` | Date/time operations |
### Error Handling
```cobol
* RESP-based (recommended — replaces HANDLE CONDITION)
EXEC CICS READ
DATASET('CUSTFILE')
INTO(CUST-RECORD)
RIDFLD(CUST-KEY)
RESP(WS-RESP)
RESP2(WS-RESP2)
END-EXEC
EVALUATE WS-RESP
WHEN DFHRESP(NORMAL)
PERFORM 3100-DISPLAY-CUSTOMER
WHEN DFHRESP(NOTFND)
MOVE 'CUSTOMER NOT FOUND' TO MSGO
WHEN DFHRESP(DISABLED)
MOVE 'FILE DISABLED' TO MSGO
WHEN OTHER
PERFORM 9200-CICS-ERROR
END-EVALUATE
```
---
## 7. Batch Processing
### JCL Integration
```jcl
//ORDPROC JOB (ACCT),'ORDER PROCESSING',
// CLASS=A,MSGCLASS=X,MSGLEVEL=(1,1)
//STEP010 EXEC PGM=ORDPROC
//STEPLIB DD DSN=PROD.LOAD.LIBRARY,DISP=SHR
//ORDIN DD DSN=PROD.ORDERS.DAILY,DISP=SHR
//CUSTDB DD DSN=PROD.CUSTOMER.VSAM,DISP=SHR
//RPTOUT DD DSN=PROD.REPORTS.DAILY(+1),
// DISP=(NEW,CATLG),
// SPACE=(TRK,(50,10),RLSE),
// DCB=(RECFM=FB,LRECL=132,BLKSIZE=0)
//SYSOUT DD SYSOUT=*
```
### PARM Passing
```cobol
LINKAGE SECTION.
01 LS-PARM.
05 LS-PARM-LENGTH PIC S9(4) COMP.
05 LS-PARM-DATA PIC X(100).
PROCEDURE DIVISION USING LS-PARM.
IF LS-PARM-LENGTH > 0
EVALUATE LS-PARM-DATA(1:1)
WHEN 'D' SET DAILY-RUN TO TRUE
WHEN 'M' SET MONTHLY-RUN TO TRUE
END-EVALUATE
END-IF
```
### Return Codes
```cobol
* Set return code for JCL COND checking
MOVE 0 TO RETURN-CODE *> Success
MOVE 4 TO RETURN-CODE *> Warning
MOVE 8 TO RETURN-CODE *> Error
MOVE 16 TO RETURN-CODE *> Severe error
* In JCL:
* //STEP020 EXEC PGM=NEXTSTEP,COND=(8,LT,STEP010)
* (skip if STEP010 RC < 8, i.e., run only if RC >= 8)
```
---
## 8. Copybook Patterns
```cobol
* In WORKING-STORAGE:
COPY ORDCOPY. *> Include copybook as-is
COPY CUSTCOPY REPLACING
==:TAG:== BY ==WS-==. *> Replace tag prefix
* Copybook CUSTCOPY.cpy:
* 01 :TAG:-CUSTOMER-RECORD.
* 05 :TAG:-CUST-ID PIC 9(8).
* 05 :TAG:-CUST-NAME PIC X(40).
* Becomes:
* 01 WS-CUSTOMER-RECORD.
* 05 WS-CUST-ID PIC 9(8).
* 05 WS-CUST-NAME PIC X(40).
* DCLGEN (DB2 table declarations)
EXEC SQL INCLUDE DCLORDER END-EXEC
* Generates host variable declarations matching DB2 columns
```
---
## 9. Debugging
### Common ABEND Codes
| Code | Meaning | Common Cause |
|------|---------|--------------|
| **S0C7** | Data exception | Non-numeric data in numeric field, uninitialized COMP-3 |
| **S0C4** | Protection exception | Bad pointer, subscript out of range, bad LINKAGE |
| **S0C1** | Operation exception | Branching to non-executable area, bad CALL |
| **S222** | Operator cancelled job | Job exceeded time limit |
| **S322** | CPU time exceeded | Infinite loop, bad PERFORM UNTIL |
| **S0B37** | Dataset out of space | Increase SPACE in JCL |
| **S913** | RACF authorization failure | Missing permissions |
### Debugging Techniques
```cobol
* DISPLAY debugging (simplest — goes to SYSOUT)
DISPLAY 'DEBUG: CUST-ID=' WS-CUST-ID
' AMOUNT=' WS-AMOUNT
' AT PARAGRAPH 2100'
* Compile with SSRANGE to catch subscript errors
* CBL SSRANGE
* Or: PARM='SSRANGE' on IGYWCL compile step
* Compile with TEST for symbolic debugging
* PARM='TEST(ALL,SYM,SEPARATE)'
```
### CEDF (CICS Execution Diagnostic Facility)
```
CEDF ON — Enable CEDF for your terminal
Steps through each EXEC CICS command
Shows EIBRESP, EIBRESP2, before/after data
CEDF OFF — Disable
```
---
## 10. Modernization
### GnuCOBOL (Local Development)
```bash
# Install on RHEL/Ubuntu
sudo dnf install gnucobol # RHEL
sudo apt install gnucobol # Ubuntu
# Compile and run
cobc -x -o ordproc ordproc.cbl
./ordproc
# Compile with DB2 (via ODBC)
cobc -x ordproc.cbl -ldb2
```
### JSON Support (Enterprise COBOL 6.x+)
```cobol
* JSON GENERATE
JSON GENERATE WS-JSON-OUTPUT
FROM WS-ORDER-RECORD
COUNT IN WS-JSON-LENGTH
ON EXCEPTION
DISPLAY 'JSON GENERATE FAILED'
NOT ON EXCEPTION
DISPLAY WS-JSON-OUTPUT(1:WS-JSON-LENGTH)
END-JSON
* JSON PARSE
JSON PARSE WS-JSON-INPUT
INTO WS-ORDER-RECORD
WITH DETAIL
ON EXCEPTION
DISPLAY 'JSON PARSE FAILED'
END-JSON
```
### COBOL-to-Service Patterns
| Pattern | Approach | Effort |
|---------|----------|--------|
| CICS Web Service | Expose CICS program as WSDL/REST via CICS TS | Low |
| MQ Bridge | COBOL reads/writes MQ queues, external consumer | Medium |
| Zowe API | Access z/OS resources via REST (Zowe CLI/SDK) | Medium |
| Strangler Fig | Gradually replace COBOL with microservices | High |
| Side-by-Side | Run new service alongside COBOL, sync data | High |
---
## 11. Reading & Understanding Unfamiliar COBOL
The rest of this reference is authoring-focused. This section is for the opposite task:
being dropped into a large, undocumented COBOL program you did not write and having to
**understand** it — the comprehension payload for a maintenance or migration pass.
Companion comprehension references in this family: `pick-developer` (Pick/MultiValue) and
`datastage-developer` (DataStage ETL).
### Orient top-down: the four divisions are a fixed reading order
COBOL's rigid structure is an advantage when reading — the divisions always appear in the
same order and each answers a specific question:
1. **IDENTIFICATION DIVISION** — `PROGRAM-ID` (the program's callable name), author, date.
One line of real signal: the name other programs `CALL`.
2. **ENVIRONMENT DIVISION** — `CONFIGURATION SECTION` (`SPECIAL-NAMES`, e.g.
`DECIMAL-POINT IS COMMA`, `CURRENCY SIGN`) and `INPUT-OUTPUT SECTION` → `FILE-CONTROL`
with the `SELECT … ASSIGN TO …` clauses. **This is where logical file names bind to
physical DD names / datasets** — read it first to learn what the program touches.
3. **DATA DIVISION** — the data shapes: `FILE SECTION` (record layouts under each `FD`),
`WORKING-STORAGE SECTION` (the program's variables, flags, counters, tables), and
`LINKAGE SECTION` (parameters passed in from the caller / JCL `PARM`).
4. **PROCEDURE DIVISION** — the executable logic, organized into paragraphs and sections.
The first paragraph runs first; follow the `PERFORM`s from there.
A fast first pass: read `PROGRAM-ID` → the `SELECT` list (what files) → the `FD`/`01`
record layouts (what the data looks like) → the first PROCEDURE paragraph (the driver),
then follow control flow from the top.
### Follow control flow: PERFORM / PERFORM THRU / CALL / GO TO
COBOL control flow is mostly `PERFORM` (an in-program "call and return"):
```cobol
0000-MAIN.
PERFORM 1000-INITIALIZE *> run that paragraph, then return here
PERFORM 2000-PROCESS
UNTIL END-OF-FILE *> loop: repeat until the condition
PERFORM 9000-TERMINATE
GOBACK *> return to caller (or STOP RUN)
```
- **`PERFORM <para>`** — execute the named paragraph, then return to the line after the
`PERFORM`. The primary call-graph edge.
- **`PERFORM <a> THRU <b>`** (a.k.a. `PERFORM <a> THROUGH <b>`) — execute paragraphs `a`
through `b` **in source order**, then return. You must read every paragraph in the range
to know what runs — control falls from one into the next (see fall-through below). A
common convention is an empty `<b>-EXIT.` paragraph holding just `EXIT.` as the range
terminator.
- **`PERFORM … UNTIL …`** / **`PERFORM … VARYING … FROM … BY … UNTIL …`** — the loop
forms; `VARYING` is the COBOL `for` loop (note: `UNTIL` is tested **before** each
iteration by default — it is a pre-test loop unless `WITH TEST AFTER`).
- **`PERFORM n TIMES`** — fixed iteration count.
- **`CALL '<PROG>' USING …`** — invoke a **separate** program (the inter-program edge);
arguments are matched positionally to the callee's `LINKAGE SECTION`. `CALL <data-name>`
is a **dynamic** call — the target is only known at run time (treat as unresolved when
reading statically).
- **`GO TO`** — an unconditional jump that does **not** return. In legacy code it is common;
trace it as a hard branch. `GO TO … DEPENDING ON n` is a computed branch (jump to the
*n*th label).
### Decode the data: PIC, COMP-3, REDEFINES, OCCURS DEPENDING ON, 88-levels
The DATA DIVISION is where meaning lives. Key things to decode when reading a layout:
- **`PIC` (picture) clause** — the shape of a field. `9` = a digit, `X` = any character,
`S` = a sign, `V` = an **implied** decimal point (no byte stored), `A` = alphabetic.
`PIC S9(7)V99` = a signed number, 7 integer digits + 2 decimal places. **The `V` is
virtual** — `1234567` `89` is stored as 9 digits with the decimal point only implied.
- **`USAGE`** (the storage form) — `DISPLAY` (default; one byte per digit, human-readable),
`COMP`/`COMP-4`/`BINARY` (binary integer, 2/4/8 bytes), **`COMP-3`/`PACKED-DECIMAL`**
(packed BCD, `(n+1)/2` bytes, two digits per byte + a sign nibble — **not printable**;
a `COMP-3` field dumped as text is garbage), `COMP-1`/`COMP-2` (float/double). When a
numeric field looks like binary noise in a file, check its `USAGE`.
- **`REDEFINES`** — a **second name and layout over the same bytes** (COBOL's union). The
bytes are interpreted one way under the first name and another way under the
`REDEFINES` name; only one interpretation is valid at a time. Watch for a record read as
raw bytes then re-interpreted via a `REDEFINES` based on a record-type flag.
- **`OCCURS n TIMES`** — a fixed array (a table); `INDEXED BY` names its index.
- **`OCCURS 1 TO n TIMES DEPENDING ON <counter>`** (ODO) — a **variable-length** array
whose live length is the value of `<counter>` at run time. The record's length changes
with the counter — critical when reading file layouts (everything after an ODO field
shifts). Find the `DEPENDING ON` field to know how many entries are real.
- **`88`-level condition names** — a named boolean over a parent field's value(s):
```cobol
05 WS-STATUS PIC X.
88 ORD-PENDING VALUE 'P'.
88 ORD-SHIPPED VALUE 'S'.
```
`IF ORD-PENDING` means `IF WS-STATUS = 'P'`; `SET ORD-SHIPPED TO TRUE` means
`MOVE 'S' TO WS-STATUS`. 88-levels make the code read like business rules — map each
back to its parent field and value.
- **Level numbers** — `01` = a record/group; `05`/`10`/… = nested subordinate fields
(the numbers show nesting depth, not fixed steps); `77` = a standalone elementary item;
`66` = a `RENAMES` regrouping (rare).
### Connect files: FD ↔ SELECT ↔ ASSIGN (and the JCL DD)
Tracing a file end to end crosses three places:
1. `SELECT logical-name ASSIGN TO DDNAME … FILE STATUS IS WS-xx.` (ENVIRONMENT) — binds
the program's logical file name to a system DD name and a status field.
2. `FD logical-name … 01 record-layout.` (DATA / FILE SECTION) — the record structure read
or written.
3. In batch, the JCL `//DDNAME DD DSN=…` maps that DD name to the actual dataset (see §7).
So a `READ ORDER-FILE` reaches the `FD ORDER-FILE`, whose `SELECT` says `ASSIGN TO ORDIN`,
which the JCL maps to a physical dataset. **Always check the `FILE STATUS` field after I/O**
— a paragraph that reads a file and then tests `WS-ORDER-FS` is doing its error handling
there (file status `00` = OK, `10` = end of file, `23` = not found, `35` = file missing;
see §3).
### Gotchas that mislead a reader
- **Fall-through between paragraphs.** Paragraphs are not subroutines with hard
boundaries — when a paragraph ends, control **falls into the next paragraph in source
order** unless a `PERFORM` returns, a `GO TO` jumps, or the program stops. A paragraph
reached by `PERFORM A THRU B` runs A, then *everything between A and B*. Never assume a
paragraph runs in isolation; check what physically follows it.
- **`ALTER` + `GO TO`** (obsolete but present in old code) — `ALTER X TO PROCEED TO Y`
**rewrites** the target of a `GO TO` at run time, so the same `GO TO` can branch to
different places on different executions. It makes static reading treacherous; flag any
`ALTER` and trace every place that alters the same `GO TO`.
- **Implicit `MOVE` / group moves** — a `MOVE` between fields of different `PIC` does
silent conversion/truncation (alphanumeric-to-numeric, decimal alignment, space-padding).
A `MOVE` of a group item copies bytes positionally regardless of the subordinate fields'
types. The data can change shape without an obvious cast.
- **`MOVE CORRESPONDING`** — moves only the subordinate fields whose names match between
two groups; what actually moves depends on the names, not the order.
- **`REDEFINES` aliasing** — the same bytes seen under two names (above). A value "changing"
may just be the other view of the same storage.
- **`OCCURS DEPENDING ON`** — record length is not constant; offsets after the ODO field
move with the counter.
- **PERFORM range edits** — adding a paragraph *inside* an existing `PERFORM A THRU B`
range silently makes it run as part of that range. When reading a `THRU`, enumerate
every paragraph in the span.
### A reading checklist
1. `PROGRAM-ID` — the name; grep the codebase for who `CALL`s it.
2. `SELECT … ASSIGN` list — what files/datasets it touches (+ the `FILE STATUS` fields).
3. `FD` / `01` record layouts — decode `PIC`/`USAGE`/`REDEFINES`/`OCCURS … DEPENDING ON`.
4. `WORKING-STORAGE` flags and `88`-levels — the state machine and its named conditions.
5. `LINKAGE SECTION` + `PROCEDURE DIVISION USING` — the parameters from the caller/JCL.
6. The first PROCEDURE paragraph — the driver; follow `PERFORM`/`PERFORM THRU`/`GO TO`.
7. Each `CALL` — an external program (dynamic `CALL <data-name>` is unresolved statically).
8. `COPY` members — shared layouts/code pulled in (`COPY … REPLACING` renames at compile
time; the post-replace names are not in the source you see).
9. Watch the gotchas: fall-through, `ALTER`/`GO TO`, implicit/group `MOVE`, ODO lengths.
---
## Anti-Patterns
| Anti-Pattern | Why It Fails | Correct Approach |
|---|---|---|
| Using COMP-3 (packed decimal) for fields that need display output | COMP-3 is not human-readable and causes garbled output in reports and logs | Use DISPLAY for fields that appear in reports; reserve COMP-3 for calculations and storage optimization |
| Hardcoding file record lengths instead of using COPY members | Changes to record layouts require hunting through every program; missed updates cause S0C7 abends | Define record layouts in COPY members; use REPLACE for program-specific field names |
| GO TO spaghetti instead of PERFORM-based structured programming | Unreadable control flow, impossible to debug, and maintenance nightmares | Use PERFORM THRU with well-named paragraphs; eliminate GO TO except for error exit patterns |
| Ignoring SQLCODE after every EXEC SQL statement | Silent failures — a failed SELECT returns stale data, a failed INSERT loses records without alerting anyone | Check SQLCODE after every SQL call; handle -803 (duplicate), -811 (multiple rows), +100 (not found) explicitly |
| Big-bang COBOL-to-microservice rewrites | Multi-year rewrites fail 60%+ of the time; business logic is often undocumented in the COBOL itself | Use Strangler Fig pattern: wrap existing programs with APIs, migrate incrementally, validate at each step |
---
## Related Skills
| Skill | When to Use |
|-------|-------------|
| `ibm-mainframe` | z/OS JCL, VSAM, IDCAMS, DFSORT, ISPF, REXX |
| `db2-mainframe` | DB2 for z/OS (BIND, REORG, RUNSTATS, utilities) |
| `python-enterprise-connectors` | Python → z/OS via Zowe SDK, DB2 Connect |
| `datastage-developer` | DataStage ETL reading from mainframe sources |
| `cognos-admin` | Cognos reporting against DB2 z/OS data |
| `mainframe-lineage-parsers` | Lineage-grade reading: deterministic COBOL+JCL+embedded-`EXEC SQL` → OpenLineage data lineage (DD-join + host-var→column edges). When you need *where the data flows*, not just what the code does. |
| `structure-recovery` | Lineage-grade reading: reverse-engineer COBOL record layouts (PIC/USAGE/COMP-3 byte offsets, REDEFINES, OCCURS) and inferred DDL from copybooks/programs. |
| `legacy-code-intel` | Queryable SCIP-style symbol graph over COBOL (paragraph/section call graph, copybook impact). |
Versions
The database has no other skills named cobol-developer.
Database snapshot: 13.11.2026 · package version 5652ad01a31d (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 90legacy-modernization
Analyze, document, and modernize legacy systems (COBOL, Mainframe, Fortran). Use when: legacy code, cobol, mainframe, fortran, ibm z, jcl, db2, cics, modernization, rewrite, migration. - 96cobol
COBOL开发专家助手。当用户需要进行COBOL企业级应用开发、银行核心系统、遗留系统维护或大型机编程时调用。 - 90mainframe-modernization-and-data-offload
Guides agents through mainframe data modernization and offload workflows. Use when migrating or exposing data from COBOL, JCL, VSAM, IMS, DB2 for z/OS, or batch-oriented mainframe estates into modern lakes, warehouses, streaming systems, and governed data products. - 100cobol
Expert COBOL programming assistance covering enterprise batch processing, divisions, copybooks, and data formatting. Use when maintaining legacy banking, government, or mainframe systems. - 95ibm-mainframe
Use when working with IBM z/OS mainframe systems — JCL (Job Control Language) syntax and patterns, dataset types (sequential/PDS/PDSE/VSAM), IDCAMS utility, DFSORT/ICETOOL, ISPF navigation, TSO/REXX scripting, system utilities (IEBCOPY/IEBGENER/IEFBR14), RACF security basics,… - 90mainframe-planning
Create a transformation guide for replacing mainframe-specific COBOL constructs with standard COBOL equivalents. Use when preparing COBOL code for local execution or migration.