Module 5: REXX EXECIO File I/O

REXX- EXECIO File I/O

A REXX exec becomes truly useful when it can read and write datasets. EXECIO is the TSO/E command that moves lines between a dataset and a REXX stem variable, and it is the standard way REXX programs do file I/O. This page covers allocating datasets, reading with DISKR, writing with DISKW, and finishes with a complete file-processing exec.

What EXECIO does

EXECIO is a TSO/E host command, so inside REXX you run it as a quoted string: "EXECIO * DISKR INDD (STEM emp. FINIS)".
DISKR reads lines from a dataset; DISKW writes lines to a dataset.
The ddname (INDD above) must be allocated - with the ALLOCATE command in TSO, or with a JCL DD in batch.
Data lands in a stem: stem.1, stem.2, ... and stem.0 holds the number of lines.
After every EXECIO, the special variable RC holds the return code: 0 means success, anything else means a problem. Always check it.

Allocating a dataset to a ddname

In a TSO session, allocate before the EXECIO and free afterwards:

/* REXX */ "ALLOC F(INDD) DA('MYID.EMP.DATA') SHR REUSE" IF RC \= 0 THEN DO SAY 'ALLOC failed, RC =' RC EXIT 8 END /* ... EXECIO work happens here ... */ "FREE F(INDD)"

DA('MYID.EMP.DATA') - quote the dataset name or TSO prefixes it with your userid.
SHR REUSE - share the dataset, and reuse the ddname if it is already allocated.
FREE releases the dataset so it can be edited or deleted afterwards.
In batch JCL you skip ALLOCATE entirely - the DD statements allocate the ddnames (see the JCL below).

Reading a dataset: DISKR

"EXECIO * DISKR INDD (STEM emp. FINIS)" reads the whole dataset: * means all lines.
STEM emp. collects the lines; emp.0 tells you how many were read.
FINIS closes the dataset when the read completes.
Replace * with a number to read a fixed count, and use (OPEN) / (FINIS) to process a huge dataset in chunks instead of loading it all at once.

/* REXX - read the whole file at once */ "EXECIO * DISKR INDD (STEM emp. FINIS)" IF RC \= 0 THEN DO SAY 'Read failed, RC =' RC EXIT 8 END SAY 'Records read:' emp.0 DO i = 1 TO emp.0 SAY i || ':' emp.i END

/* REXX - chunked read for very large files */ "EXECIO 0 DISKR INDD (OPEN)" DO FOREVER "EXECIO 1000 DISKR INDD (STEM chunk.)" IF chunk.0 = 0 THEN LEAVE DO i = 1 TO chunk.0 /* process chunk.i here */ END END "EXECIO 0 DISKR INDD (FINIS)"

Writing a dataset: DISKW

Build the output lines in a stem, set stem.0 to the line count, then write them:

/* REXX */ out.1 = 'First report line' out.2 = 'Second report line' out.0 = 2 "ALLOC F(OUTDD) DA('MYID.EMP.REPORT') SHR REUSE" "EXECIO" out.0 "DISKW OUTDD (STEM out. FINIS)" IF RC \= 0 THEN SAY 'Write failed, RC =' RC "FREE F(OUTDD)"

"EXECIO" out.0 "DISKW ..." - the line count comes from the REXX variable out.0.
FINIS on a DISKW flushes the last lines and closes the dataset - without it, trailing records can be lost.
To create a brand-new dataset from REXX, allocate it with space parameters, or simpler, define the output DD in JCL with DISP=(NEW,CATLG).

Complete example: employee salary report

Input dataset MYID.EMP.DATA holds fixed-format lines - id in columns 1-5, name in 6-25, salary in 26-33. The exec reads every record, totals the salaries, and writes a report. Batch JCL that runs it follows.

/* REXX - PAYRPT: read employees, write a totals report */ "ALLOC F(INDD) DA('MYID.EMP.DATA') SHR REUSE" "EXECIO * DISKR INDD (STEM emp. FINIS)" IF RC \= 0 THEN DO SAY 'Read failed, RC =' RC "FREE F(INDD)" EXIT 8 END total = 0 rep.1 = 'EMPLOYEE SALARY REPORT' rep.2 = '======================' n = 2 DO i = 1 TO emp.0 PARSE VAR emp.i 1 id 6 name 26 salary total = total + salary n = n + 1 rep.n = id STRIP(name) salary END n = n + 1 rep.n = '----------------------' n = n + 1 rep.n = 'Employees:' emp.0 ' Total salary:' total rep.0 = n "ALLOC F(OUTDD) DA('MYID.EMP.REPORT') SHR REUSE" "EXECIO" rep.0 "DISKW OUTDD (STEM rep. FINIS)" IF RC \= 0 THEN SAY 'Write failed, RC =' RC "FREE F(INDD) F(OUTDD)" SAY 'Report written:' rep.0 'lines.'

//PAYRPT JOB (ACCT),'REXX PAYRPT',CLASS=A,MSGCLASS=X //STEP1 EXEC PGM=IRXJCL,PARM='PAYRPT' //SYSEXEC DD DSN=MYID.REXX.EXEC,DISP=SHR //INDD DD DSN=MYID.EMP.DATA,DISP=SHR //OUTDD DD DSN=MYID.EMP.REPORT,DISP=OLD //SYSTSPRT DD SYSOUT=* //SYSTSIN DD DUMMY

In batch, INDD and OUTDD come from JCL DDs, so the exec needs no ALLOCATE - the same exec runs in TSO with ALLOCATE or in batch with DDs.
PARSE VAR emp.i 1 id 6 name 26 salary splits each fixed-format record by column position.
EXIT 8 on read failure sets a non-zero step return code so the job visibly fails.

Backward

Forward