Section 2.4 : How the DATA Step Works

Two conventional ways to structure a DATA step depend if you are reading
an external file or to modify existing SAS data set(s).  The general form
for each case is described below:

* Write a new SAS data set by reading an external text file;

DATA <>;
< enter declarative statements >
INFILE <> ;
INPUT <>;
< enter executable statements >
OUTPUT < >;
RUN;

* Modify existing SAS data set(s);

DATA <>;
SET <>;   or   MERGE <> ;   or UPDATE <>;
< enter declarative statements >
< enter executable statements >
OUTPUT <>;
RUN;

When a SAS program is submitted, all statements in each block are checked
for accuracy (e.g., spelling, stray characters, presence/absence of
semi-colons, etc.). Statements in the DATA step are defined to be of two
types: declarative (i.e., they supply information to SAS and can usually
appear anywhere in the DATA step) and executable (i.e., they result in
some action during the iterations and the order of their placement in the
step matters).

Declarative statements such as ARRAY, ATTRIB, DROP, FORMAT, INFORMAT,
KEEP, LABEL, LENGTH, RETAIN, RENAME, and so forth, supply information to
SAS to define the working environment. With a few exceptions, they can be
placed anywhere in the current block of DATA step statements; however, it
usually aids readability to put them near the top (after the DATA and SET
statements) to avoid mixing them among executable statements. Also, it may
be necessary to place statements that refer to variables created in the
DATA step 'prior' to an executable statement which contains the first
appearance of the variable name.

Executable statements are processed in a sequential manner, returning to 
the first statement when all statements in the DATA step for the current 
observation have been processed. Unless a RETAIN statement is entered 
containing specific variable names, all variables input read from an 
external dataset or computed in the DATA step are assigned missing values 
at the beginning of the next iteration. The SET statement then reads data 
from the next observation in the existing dataset and based on their 
values assigns values to variables in the new dataset (see Chapter 4.1: 
The DATA and SET Statements for examples).

If you're reading an external file, an INPUT statement reads the next 
observation and assigns its values to the named variables. Once the end of 
DATA step is reached, the observation is automatically OUTPUT to the newly 
defined dataset. SAS then returns to the beginning of the loop and 
automatically resets all variables to missing values before processing the 
next observation (unless told otherwise by the RETAIN statement). The 
INPUT statement then reads the next line/observation.