4.2 Modifying Variable Attributes with LENGTH, LABEL, FORMAT & RENAME=

4.2 Modifying Variable Attributes with LENGTH, LABEL, FORMAT & RENAME=

Quick Answer: The A00-231 content guide lists "Modify variable attributes using options and statements in the DATA step" as a main objective and names three tools explicitly: the RENAME= data set option, the LABEL and FORMAT statements, and the LENGTH statement. All of them are compile-time constructs — where you put them in the DATA step decides whether they work at all.

Every SAS variable carries six attributes: name, type (numeric or character), length in bytes, format, informat, and label. Type is immutable once compiled. The other five can be set or changed, and the exam repeatedly tests which statement changes which attribute and where it must be placed.

---

1. LENGTH: Reserving Bytes Before the First Reference

LENGTH variable-1 <$> length-1 variable-2 <$> length-2 ... ;

The $ marks a character variable; without it SAS creates a numeric.

data work.orders_typed;
   length Status $ 13 Region $ 2 OrderID 8;
   set work.orders_raw;
   if Amount < 50 then Status = 'Low';
   else if Amount < 200 then Status = 'Medium';
   else Status = 'High Priority';
run;

The rule that matters: LENGTH executes during compilation, and compilation reads the step top to bottom. A variable's length is fixed by whichever construct mentions it first. Move that LENGTH statement below the IF block and Status becomes 3 bytes — the width of 'Low' — silently truncating every longer value.

Character versus Numeric Lengths

data work.precision_trap;
   length SmallCode 3;         /* 3-byte numeric */
   SmallCode = 8388609;        /* beyond 3-byte integer precision */
   FullCode  = 8388609;        /* default 8 bytes - exact         */
run;

Exam Trap: shortening a numeric variable saves disk but is safe only for small integers such as flags and short codes. A 3-byte numeric stores integers exactly only up to 8,192 and a 4-byte numeric only to 2,097,152. Never shorten a numeric that holds an identifier, a currency amount, or a SAS datetime value.

---

2. LABEL: Descriptive Column Headings

LABEL variable-1 = 'descriptive text' variable-2 = 'descriptive text';

A label is descriptive text of up to 256 characters that reporting procedures can print instead of the variable name.

/* Permanent: stored in the descriptor portion of WORK.SALES_LABELED */
data work.sales_labeled;
   set work.sales;
   label GrossAmt = 'Gross Sales Amount (USD)'
         RegCode  = 'Sales Region Code';
run;

/* Temporary: applies to this PROC PRINT only */
proc print data=work.sales label;
   label GrossAmt = 'Gross Sales Amount (USD)';
run;

Exam Trap: a label in a PROC PRINT step does nothing unless the LABEL (or SPLIT=) option appears on the PROC PRINT statement. PROC MEANS, PROC FREQ, and PROC REPORT display labels by default, so the trap is specific to PROC PRINT.

---

3. FORMAT and INFORMAT: Display and Reading Instructions

A format controls how a stored value is displayed; an informat controls how raw text is read in. Neither changes the stored value.

data work.formatted;
   set work.transactions;
   format TxnDate  date9.
          Amount   dollar12.2
          Rate     percent8.1;
run;

/* Removing a format: name the variable with no format after it */
data work.unformatted;
   set work.formatted;
   format TxnDate;          /* TxnDate now prints as the raw day count */
run;

That last idiom is worth memorizing. format Variable; with nothing after the variable name clears the format. It is the correct answer whenever a question asks how to see raw stored values, because PROC PRINT has no UNFORMATTED option.

---

4. ATTRIB: All Attributes in One Statement

data work.attrib_demo;
   attrib Amount  length=8  format=dollar12.2 label='Gross Amount'
          Status  length=$13                  label='Order Status'
          TxnDate length=8  format=date9.     informat=mmddyy10.
                                              label='Transaction Date';
   set work.orders_raw;
run;

ATTRIB obeys the same compile-time placement rule as LENGTH: it must precede the variable's first reference to control its length.

---

5. RENAME Statement versus RENAME= Data Set Option

This distinction is a reliable exam item because the two constructs look similar and behave differently.

/* Input option: subsequent logic MUST use the new name */
data work.renamed_in;
   set sashelp.class(rename=(Age=YearsOld));
   if YearsOld > 13;              /* correct - Age no longer exists */
run;

/* RENAME statement: logic uses the OLD name */
data work.renamed_out;
   set sashelp.class;
   if Age > 13;                   /* correct - rename happens on output */
   rename Age = YearsOld;
run;

Referring to Age after an input-side RENAME= produces ERROR: Variable Age is not on file; referring to YearsOld before a RENAME statement produces an uninitialized-variable note and a column of missing values.

---

6. Attribute Precedence Summary