IMS- Segments

A segment is the smallest unit of data DL/I can retrieve: roughly the IMS equivalent of a record. Segments are typed (STUDENT, COURSE, GRADE), and each type has a fixed layout. Understanding segment types, keys and parent-child rules is the foundation for writing correct DL/I calls.

Segment types: root, dependent, twin

• Root segment: the top of the hierarchy. There is exactly one root segment type per database, and one root occurrence per database record.
• Dependent segment: any segment below the root. A dependent has exactly one parent segment type, but a parent can have many dependent types beneath it.
• Twin segments: multiple occurrences of the same segment type under one parent occurrence. Three COURSE segments under one STUDENT are twins.
• DL/I always walks the tree top-down, left-to-right: root first, then each dependent subtree in order.

Keys: sequence fields and search fields

• A sequence field (SEQ) orders the occurrences of a segment type under their parent, e.g. COURSEID orders COURSE twins.
• SEQ,U means the sequence field must be unique among twins; SEQ,M allows duplicates.
• A segment type can also have search fields: non-key fields you can still qualify in an SSA, though access is slower than by sequence field.
• The concatenated key of a segment is the chain of sequence fields from the root down to it (STUDID + COURSEID + GRADE...). IMS can return it in the key feedback area of the PCB.

Parent-child rules

• Every dependent occurrence must have a parent occurrence; IMS will not store an orphan.
• Insert rule decides where a new twin goes: FIRST (at the front), LAST (at the end) or HERE (after current position).
• Delete rule decides what happens to dependents when a parent is deleted: PHYSICAL (dependents deleted too), LOGICAL (dependents stay but are unreachable), VIRTUAL (both rules combined in logical databases).
• Insert and delete rules are coded in the DBD as RULES=(insert-rule,delete-rule), for example RULES=(L,L).

DBDGEN: defining the database

• The DBD is coded with assembler-like macros and assembled by the DBDGEN utility. Below is a real DBD for a student database with STUDENT root, COURSE dependent and GRADE dependent.

• DBD: names the database (STUDDBD) and its access method (HIDAM on VSAM).
• DATASET: ties the DBD to a DD name used in execution JCL.
• SEGM: defines a segment type. PARENT= names the parent type (omitted for the root). BYTES= is the segment length. PTR=T builds twin forward pointers.
• FIELD: defines fields inside the segment. NAME=(STUDID,SEQ,U) makes STUDID the unique sequence field; START= is the byte offset; TYPE=C is character data.
• DBDGEN / FINISH: generate the control blocks and end the assembly.

Segment layout vs COBOL copybook

• The COBOL I/O area for a segment must match the DBD field layout byte for byte. For STUDENT above:

• If the copybook is shorter than BYTES=, IMS pads; if longer, data is truncated or the call fails. Keep them in sync.
• Generate or hand-maintain one copybook per segment and share it across programs to avoid drift.