Fixed-Length Records

What Is a Fixed-Length Record?

A fixed-length record is one where every record in the dataset has the same number of bytes. That length is called the logical record length (LRECL). For example, if LRECL=80, every record is exactly 80 bytes. There is no prefix (no RDW); the first byte of the record is the first data byte. The system knows how long each record is from the dataset's DCB (Data Control Block), so it can read one record after another without a length field in each record.

RECFM=FB and RECFM=F

RECFM=FB means Fixed Block: fixed-length records, blocked (multiple records per physical block). RECFM=F means Fixed, unblocked (one record per block; less common). For DFSORT, you typically use RECFM=FB and LRECL=n when allocating SORTIN and SORTOUT for fixed-length data. The "B" allows efficient I/O by grouping records into blocks.

Position Numbering

For fixed-length records, position 1 is the first byte of the record, position 2 is the second byte, and so on up to position LRECL. So:

1
  SORT FIELDS=(1,10,CH,A,11,4,PD,D)

sorts by the first 10 bytes (character ascending), then by bytes 11–14 (packed decimal descending). All positions refer to the same 1-based byte positions in each record. There is no offset for an RDW because fixed-length records do not have one.

LRECL on SORTIN and SORTOUT

For fixed-length input, allocate SORTIN with RECFM=FB and LRECL equal to the length of each input record. For fixed-length output, allocate SORTOUT with RECFM=FB and LRECL equal to the length of each output record. If you use INREC to shorten or lengthen the record, the record seen by the sort phase has the new length; if you use OUTREC to build the final record, the length of that built record must match SORTOUT's LRECL. So:

• Input 80 bytes, no INREC/OUTREC → SORTIN LRECL=80, SORTOUT LRECL=80.
• Input 80 bytes, OUTREC builds 60 bytes → SORTOUT LRECL=60.
• Input 80 bytes, INREC builds 50 bytes for sort, OUTREC builds 60 for output → SORTOUT LRECL=60 (the sort works on 50-byte records; output is 60).

Fixed vs Variable: Comparison

RECORD Statement for Fixed-Length

When DFSORT cannot infer the record format from the DD or dataset (e.g. tape or dynamic allocation without DCB), use the RECORD statement:

1
  RECORD TYPE=F,LENGTH=80

TYPE=F indicates fixed-length. LENGTH=80 is the logical record length in bytes. For normal disk datasets with DCB=(RECFM=FB,LRECL=80), RECORD is usually not needed.

When to Use Fixed-Length

• Downstream programs or reports expect a constant record length (e.g. 80-byte lines).
• All logical records have the same layout and length; no need for variable length.
• Legacy file formats (card image, fixed-format tape) are fixed-length.
• You want simple, predictable positioning: byte 1 is always the first byte of the record.

When to Use Variable-Length Instead

Use variable-length (VB) when records naturally vary in length (e.g. free-text, variable-length comments) and you want to avoid padding short records to a fixed size. VB uses an RDW and saves space when many records are shorter than the maximum.

Explain It Like I'm Five

Imagine a stack of index cards where every card has exactly 80 spaces—no more, no less. When we look at "position 1," we mean the first space on the card. We don't need a label saying how long the card is because every card is the same length. That's fixed-length. The sort program reads one card (80 bytes), then the next 80 bytes, and so on. When we write the sorted cards to the outbox, each card we write is also 80 spaces. So the inbox and outbox are both made for 80-space cards.

Exercises

1. Your input has LRECL=100. You use OUTREC to build a 70-byte record. What LRECL should you use for SORTOUT?
2. What is the difference between RECFM=F and RECFM=FB? Which do you typically use for DFSORT?
3. For a fixed-length 80-byte record, does position 80 refer to the last byte of the record? What about position 81?
4. When would you code RECORD TYPE=F,LENGTH=80 in SYSIN?