Error Handling in COBOL

Unlike languages with built-in try/catch exception frameworks, classic COBOL requires the programmer to explicitly check for error conditions after nearly every I/O or arithmetic statement, because without that checking the runtime either silently continues with corrupted results or abends the entire job step. A batch program that reads a file without checking FILE STATUS after each READ, for instance, will not automatically detect end-of-file or a record-not-found condition unless the programmer has coded an AT END or INVALID KEY clause, or wired up a FILE STATUS check, to catch it. This explicit, statement-by-statement discipline is a defining characteristic of COBOL error handling, and mainframe shops that skip it routinely suffer abends deep into multi-hour batch runs that then require an operator to diagnose from a dump.

Declaring a two-character FILE STATUS field in the SELECT clause, such as SELECT CUSTFILE ASSIGN TO CUSTDD FILE STATUS IS WS-CUST-STATUS, causes the runtime to populate that field after every OPEN, READ, WRITE, REWRITE, DELETE, and CLOSE against the file, letting the program branch on the two-digit code immediately afterward. A status of '00' means the operation succeeded, '10' signals end-of-file on a sequential READ, '23' means record-not-found on an indexed READ by key, and '35' means the OPEN failed because the file does not exist, and disciplined programs check WS-CUST-STATUS after every single I/O verb rather than assuming success. Many shops standardize this pattern by writing a single shared paragraph, such as 9000-CHECK-FILE-STATUS, invoked via PERFORM after each I/O operation, so that unexpected status codes are logged and the program terminates cleanly with a meaningful message instead of an unexplained abend.

Arithmetic statements such as COMPUTE, ADD, SUBTRACT, MULTIPLY, and DIVIDE support an optional ON SIZE ERROR imperative-statement clause that executes only when the result would overflow the receiving field's PICTURE clause or when a division by zero is attempted, letting the program substitute a default, log a warning, or branch to cleanup logic instead of silently truncating a financial total. STRING and UNSTRING statements have an analogous ON OVERFLOW clause that fires when the receiving field is too small to hold the concatenated or split result. For broader, file-level exception handling across many paragraphs, COBOL's DECLARATIVES section, placed at the start of the PROCEDURE DIVISION and paired with USE AFTER STANDARD ERROR PROCEDURE ON file-name, lets a program centralize I/O error handling for a specific file so every READ or WRITE against it automatically triggers the declarative paragraph on any non-successful FILE STATUS, without a manual check after every single statement.