SAS Companion for UNIX Environments

Accessing Sequential-Format Data Libraries

The sequential engines enable you to access data libraries in sequential format on tape or disk. The sequential engines do not support indexing and compression of observations.

Note: Before using sequential engines, read the information about sequential data libraries in SAS Language Reference: Dictionary. [cautionend]

Reading and Writing SAS Files on Tape

You can write SAS files directly to tape using the TAPE engine; however, it is more efficient to use a staging directory so that the files can be processed directly from disk. You can use the UNIX tar command to move SAS data sets between the staging directory and tape. (Do not use the UNIX cp command.)

A SAS library on tape can contain one or more SAS data sets; however, only one SAS data set from a given library on tape can be accessed at any given point in a SAS job.

To access Version 8 SAS files on tape, you can specify the V8TAPE or TAPE engine in the LIBNAME statement:

LIBNAME libref V8TAPE 'tape-device-pathname';
The tape-device-pathname must be a pathname for a tape device; it should be the name of the special file associated with the tape device. (Check with your system administrator for details.) The name must be enclosed in quotes. You cannot specify remote tape devices in the LIBNAME statement.

Multi-volume tape libraries are supported if you specify the TAPECLOSE=LEAVE system option when you start your SAS session.

For example, the following LIBNAME statement assigns the libref SEQ2 to the /dev/tape2 tape device. Because the tape device is specified, the engine does not have to be specified.

libname seq2 '/dev/tape2';

Reading and Writing Transport Formats on Tape

Transport formats on tape are handled in a manner similar to external files. Read Processing Files on TAPE before continuing with this topic.

For example, the following SAS statements issue the UNIX mt command to rewind the tape and create a transport file using the xport engine and PROC CPORT:

x 'mt -t /dev/rmt/0mn rewind'; 
libname tranfile xport '/dev/rmt/0mn'; 
proc cport library=sasuser file=tranfile; 
run;

The following statements import the transport file into the WORK data library:

x 'mt -t /dev/rmt/0mn rewind'; 
libname tranfile xport '/dev/rmt/0mn'; 
proc cimport infile=tranfile library=work; 
run;

Writing Sequential Data Sets to Named Pipes

You can send output to and read input from the operating system by using named pipes. For example, you may want to compress a data set or send it to a tape management system without creating intermediate files.

You can read from and write to named pipes from within your SAS session by specifying the pipe name in the LIBNAME statement:

LIBNAME libref <TAPE> 'pipename';

Since you cannot position a pipe file, SAS uses the TAPE engine to ensure sequential access. You do not have to specify the engine name; TAPE is assumed.

For example, suppose you want to create a SAS data set and compress the data set without creating an intermediate, uncompressed data set. Create a named pipe (such as mypipe) and enter the compress command:

mknod mypipe p compress <mypipe >sasds.Z

In your SAS session, assign a libref to the pipe and begin writing to the data set:

libname x 'mypipe'; 
data x.a;
  ...more SAS statements...
output; 
run;

The data is sent to mypipe, compressed, and written to the data set. When SAS closes the data set, the compress finishes, and you have a compressed, sequential data set in sasds.Z.

If you begin writing to a named pipe before the task on the other end (in this case, the compress command) begins reading, your SAS session will be suspended until the task begins to read.

Chapter Contents
Previous
Next
Top of Page