SAS Companion for the CMS Environment

Using CMS Pipelines

You can use a standard CMS pipeline specification in place of the external file specification in the FILE, FILENAME, and INFILE statements, and in the FILENAME function. This enables you to receive input from any CP or CMS command or pipeline device driver, or to route output to any pipeline device driver.

For the FILENAME statement and FILENAME function, specify PIPE as the device type. For the FILE and INFILE statements, specify PIPE as an option. If you specify PIPE as an option in a FILE or INFILE statement, you cannot specify any other options in that statement.

SAS supports inline comments as an extension of the standard CMS pipeline interface, as shown:

filename adata pipe
   '(name adata)    /* name pipeline */
   < archive data a /* read in file  */
   | unpack         /* unpack        */
   ';

Data transfer between SAS and the CMS pipeline takes place in the pipeline's FITTING stage. By default, SAS inserts a FITTING stage at the beginning of pipeline specifications that are used for output and at the end of pipeline specifications that are used for input. For example, when this ADATA fileref is referenced in an INFILE statement, the pipeline specification is internally coded as

'(name adata) < archive data a | unpack | fitting sasio'

You can override the default insertion of the FITTING stage by explicitly inserting the stage into your pipeline specification. This is allowed as long as the FITTING stage is always either the first or last stage of a stream. For example, the FANIN stage in the middle of this pipeline delivers data to the FITTING stage:

filename twostrm pipe
   '(endchar %) < jan data a
   | a: fanin
   |    fitting sasio
   %
    < feb data a
   | a:
   ';

You may prefer to write complex pipelines as REXX pipeline stages, then to specify the REXX stage name in your SAS program. For the unpack example above, you could write a file called UNPKDATA REXX:

/* UNPKDATA REXX */
parse arg fileid
'callpipe <' fileid,
    '| unpack',
    '| *:'
exit

Then you could write your FILENAME statement as

filename mydata pipe
   '(name mydata)    /* name pipeline */
   rexx unpkdata archive data a 
     /* read and unpack a file        */
   ';

Macro variables are resolved within a pipeline specification that is enclosed in double quotes.

This example illustrates the use of pipelines with INFILE, FILENAME, and FILE statements, and with macro variables.

/* Data step to get spoolid. Only accept files        */
/* sent by a specified userid                         */
data _NULL_;
   infile                    /* define the pipeline   */
      'cp query rdr          /* query all rdr files   */
         |strfind /USERABC / /* from this userid only */
         |take 1             /* only process one file */
         |specs word 2 1'    /* keep only the spoolid */
      PIPE;
   input spid $;     /* read spoolid as char variable */
                    /* save spoolid as macro variable */
   call symput('SPOOLID',spid);
run;

/* define fileref for reading netdata rdr file        */
filename nd pipe
   "reader file &spoolid  /* read the file            */
       |drop 1            /* skip TAG                 */
       |specs 2-* 1       /* skip CC byte             */
       |deblock netdata   /* interpret netdata format */
       |strfind xC0       /* keep only data records   */
       |specs 2-* 1       /* skip control byte        */
       ";

/* data step to read input from rdr file              */
/* write data to USERABC DATA in packed format        */
data abc;
   file                   /* pipe for PUT processing  */
      '  strip            /* strip blanks             */
         | pack V 80      /* write packed format      */
         | > userabc data a fixed' pipe; /* this file */
   infile nd;             /* read from ND fileref     */
   input x;               /* read var x from each rec */
   put x;   /* write to USERABC DATA in packed format */
run;

Chapter Contents
Previous
Next
Top of Page