CUCO cheat sheet

COBOL source is organized into four divisions in fixed order: IDENTIFICATION, ENVIRONMENT, DATA, PROCEDURE; only IDENTIFICATION is required by the standard, but most compilers require all four for a runnable program.

In fixed-format source, columns 1-6 are the sequence area, column 7 is the indicator area (* comment, / page eject, - continuation, D debug), columns 8-11 Area A (division, section, paragraph, 01/77 level), columns 12-72 Area B (statements, other levels).

Free-format source is requested with >>SOURCE FORMAT FREE or compiler option (e.g., -free / FREE); in free format there are no column rules and *> introduces a comment to end of line.

Program names are limited to 30 characters in COBOL 85 and 31 in COBOL 2002+; the PROGRAM-ID name should match the load module name to avoid dynamic CALL failures.

COBOL is not case-sensitive for reserved words and user-defined names; literals in quotes preserve case.

A COBOL word may contain letters, digits, and hyphens, must contain at least one letter, cannot begin or end with a hyphen, and cannot be a reserved word.

Numeric literals have no sign or decimal point except optionally a leading sign; nonnumeric literals are enclosed in quotes (" or ') and the chosen quote doubles to embed itself.

Figurative constants: ZERO/ZEROS/ZEROES, SPACE/SPACES, HIGH-VALUE/HIGH-VALUES, LOW-VALUE/LOW-VALUES, QUOTE/QUOTES, NULL/NULLS; ALL literal repeats the literal to the size of the receiving item.

The sentence ends with a period; a period terminates scope for IF/PERFORM/EVALUATE, so a misplaced period changes control flow - a common exam trap.

Section order in DATA DIVISION: FILE SECTION, WORKING-STORAGE SECTION, LOCAL-STORAGE SECTION, LINKAGE SECTION; LOCAL-STORAGE is allocated and initialized on every program invocation, WORKING-STORAGE only once.

PICTURE characters: 9 numeric, S sign (not counted in size), V implied decimal (not counted), P scaling (counted), A alphabetic, X alphanumeric, Z and * editing for zero suppression, comma, period, $, +, -, CR, DB, 0 (zero insertion), B blank insertion, / slash insertion.

USAGE: DISPLAY (default, zoned), COMP/COMP-4/BINARY (native binary, halfword/fullword/doubleword by digit count), COMP-1 (single float, no PICTURE), COMP-2 (double float, no PICTURE), COMP-3/PACKED-DECIMAL, COMP-5 (native binary, full range), INDEX, POINTER, NATIONAL/UTF-8.

COMP-3 size in bytes = INT((number of digits + 1) / 2) rounded up (the +1 is the sign nibble); a PIC S9(5) COMP-3 occupies 3 bytes.

A group item has no PICTURE and no USAGE; its USAGE is DISPLAY unless all subordinate items share a usage, and elementary moves into a group are treated as alphanumeric.

The 88 level is a condition-name attached to an elementary or group item; VALUES ARE ... THRU ... define ranges; SET condition-name TO TRUE moves the associated value.

The 66 level (RENAMES) re-groups previously defined elementary items in the same record, cannot overlap other 66 entries, and cannot rename an item containing an OCCURS.

REDEFINES must immediately follow the redefined item at the same level, have the same level number, and the redefining item cannot be shorter than the original for file records; REDEFINES does not initialize storage.

The JUSTIFIED RIGHT clause applies to alphanumeric items only; BLANK WHEN ZERO applies to numeric-edited items; SIGN IS LEADING/TRAILING SEPARATE CHARACTER affects DISPLAY sign placement and size.

PERFORM ... UNTIL is test-before (may execute zero times); PERFORM ... WITH TEST AFTER executes at least once; PERFORM VARYING ... UNTIL combines initialization, increment, and test.

PERFORM ... TIMES with a count of zero or negative executes zero times; the count is evaluated once at the start.

Inline PERFORM ... END-PERFORM is preferred over out-of-line paragraphs because it avoids fall-through and simplifies scope terminators.

EVALUATE TRUE evaluates each WHEN as a condition; EVALUATE identifier WHEN value uses equality; EVALUATE ALSO supports multiple subjects with ANY as a wildcard.

GO TO ... DEPENDING ON branches by ordinal value of the identifier; values outside 1..n cause no branch and control falls to the next statement.

IF/ELSE nest with END-IF; the ELSE binds to the nearest unmatched IF unless scope terminators are used - always terminate to avoid ambiguity.

CONTINUE is a no-op statement; NEXT SENTENCE skips to the next period (deprecated), which differs from CONTINUE when inline scope terminators are used.

CALL 'literal' USING BY REFERENCE (default), BY CONTENT, BY VALUE; the subprogram must declare LINKAGE SECTION items matching the caller's BY REFERENCE items.

The INITIAL clause on a subprogram resets WORKING-STORAGE and internal state on each CALL; without it, WORKING-STORAGE retains values between calls.

File ORGANIZATION: SEQUENTIAL, LINE SEQUENTIAL, INDEXED, RELATIVE; ACCESS MODE: SEQUENTIAL, RANDOM, DYNAMIC (DYNAMIC requires a SELECT with an optional RECORD KEY).

INDEXED files require a RECORD KEY; ALTERNATE RECORD KEYs are optional and may have WITH DUPLICATES; RELATIVE files use RELATIVE KEY.

OPEN INPUT, OUTPUT, I-O, EXTEND; EXTEND is output-only and appends; I-O is required for READ ... INTO with REWRITE/DELETE.

READ ... AT END / NOT AT END; READ ... INVALID KEY for random reads; WRITE ... INVALID KEY for indexed/relative output; REWRITE and DELETE require the file opened I-O and a valid key.

START positions an indexed/relative file at a key; the next READ NEXT reads from that position.

FILE STATUS is a two-character item; '00' success, '10' end of file, '22' duplicate key, '23' record not found, '35' file not found on OPEN.

SELECT ... ASSIGN TO ... uses a name that the runtime maps via environment variables or JCL DD statements; the ASSIGN name is implementation-defined.

A file must be CLOSEd before the program ends; a CLOSE with LOCK prevents reopening in the same run unit.

Sequential files with variable-length records use RECORD IS VARYING IN SIZE with RECORD CONTAINS min TO max CHARACTERS and a RECORD DEPENDING ON item.

OCCURS n TIMES defines a table; the subscript is 1-relative, and INDEXED BY creates an index whose values are implementation-defined and only usable with SET, SEARCH, and relative subscripting.

SEARCH is for serial tables; SEARCH ALL requires the table to be sorted ascending or descending on the key and the key to be the first or only data item - it performs a binary search.

SET index-name TO 1, SET index-name UP BY 1, SET index-name DOWN BY 1 manipulate indices; PERFORM VARYING index-name FROM 1 BY 1 UNTIL ... is the classic table loop.

A subscript may be a numeric literal, data name, or arithmetic expression (COBOL 2002+); an index may be used only in subscripting, SET, SEARCH, and relative subscripting.

OCCURS DEPENDING ON (ODO) makes the table variable length; the ODO object must be an integer item defined before the table, and the table must be the last item in the record for file I/O.

Nested OCCURS requires a subscript for each dimension, listed outermost first; a two-dimensional table T(I,J) is defined by OCCURS at level 05 then OCCURS at level 10.

CALL ... USING passes parameters; the LINKAGE SECTION defines them; PROCEDURE DIVISION USING must list them in the same order and with compatible PICTUREs.

BY REFERENCE passes a pointer to the caller's item (default); BY CONTENT passes a copy that the subprogram cannot change in the caller; BY VALUE passes the value for non-COBOL interoperability.

A subprogram can be made recursive with the RECURSIVE clause; LOCAL-STORAGE is then allocated per invocation, and the program must be compiled as RECURSIVE or INITIAL as appropriate.

WITH DEBUGGING MODE and column 7 'D' lines compile only when debugging mode is active; the DEBUG-LINE and DEBUG-ITEM special registers expose the last debug line.

DISPLAY of a group item shows the raw storage, which is the fastest way to inspect COMP-3, COMP, and REDEFINES overlays.

Common abend codes: S0C7 data exception (nonnumeric data in a numeric field), S0C4 protection exception (bad address, often a bad CALL USING or subscript out of range), S0CB divide by zero.

File status '10' is normal end of file, not an error; '22' is a duplicate key on a WRITE or REWRITE with an alternate key; '23' is record not found on a random READ.

A subscript out of range in a table is undefined behavior in standard COBOL; runtime checks (e.g., SSRANGE) catch it, and production code should test the subscript before use.

Test data should include boundary values for numeric fields (PIC 9(3) max 999), empty files, duplicate keys, and truncated records.

A COPY statement pulls in copybook source at compile time; changing a copybook requires recompiling every program that copies it - a common maintenance defect.

The EVALUATE TRUE / WHEN OTHER pair is the standard way to implement a default branch; missing WHEN OTHER silently does nothing.

Paragraph and SECTION names in PERFORM must exist; a PERFORM of a paragraph that falls through into the next paragraph executes unintended code unless the paragraph ends with EXIT or a scope terminator.

COBOL 2002 added free format, user-defined functions, object-oriented COBOL (CLASS-ID, METHOD-ID), and the EVALUATE ALSO/ANY extensions.

COBOL 2014 added dynamic capacity tables, the VALIDATE facility, and UTF-8 support; COBOL 2023 added asynchronous messaging and JSON support.

GnuCOBOL (formerly OpenCOBOL) is the common open-source compiler; it supports most COBOL 85 and many 2002+ features and compiles to C.

Micro Focus (now OpenText) Enterprise Developer and IBM Enterprise COBOL are the mainframe/enterprise compilers; IBM Enterprise COBOL for z/OS supports COBOL 2002 and 2014 features.

CALL ... BY VALUE and the USAGE NATIONAL / UTF-8 items are used for interoperation with C, Java (via JNI or JVM COBOL), and .NET (Micro Focus).

JSON GENERATE and JSON PARSE (COBOL 2014+) convert between COBOL data structures and JSON text; XML GENERATE and XML PARSE provide the XML equivalent.

The CALL ... RETURNING clause and the EXTERNAL clause allow subprograms to return values and share data across programs in a run unit.

CICS and IMS are the mainframe transaction environments where COBOL programs run; EXEC CICS ... END-EXEC embeds transaction commands in the PROCEDURE DIVISION.

Containerization: COBOL programs compiled with GnuCOBOL run in Linux containers, and Micro Focus provides a native COBOL runtime for Docker/Kubernetes deployments.