How to Easily Create an Empty SAS Dataset

Two Primary Methods for Empty Dataset Creation

There are two common and highly effective methods for generating an empty dataset structure within a SAS library. The choice between them depends entirely on whether you are designing a structure from scratch or mirroring an existing schema.

• Method 1: Creating from Scratch (Defining the Schema): This technique involves using the DATA and ATTRIB statements to manually define every variable attribute, which is ideal for building entirely new data structures where no template exists.
• Method 2: Creating from an Existing Dataset (Copying Metadata): This approach uses the SET statement combined with the STOP statement to clone the structure (but not the data) of an already existing dataset, providing a rapid way to replicate complex schemas.

Method 1: Create Empty Dataset from Scratch

The first approach allows for granular control over every aspect of the dataset’s design. This technique utilizes the ATTRIB statement to specify critical properties for each variable, such as LENGTH, FORMAT, and LABEL. Crucially, the STOP statement is employed to prevent the DATA step from processing any observations, ensuring the resulting dataset remains empty.

data empty_data;
attrib 
    var1 length=8 format=best12. label="var1"
    var2 length=$30 format=$30. label="var2"
    var3 length=8 format=best12. label="var3"
stop;
run;

Method 2: Create Empty Dataset from Existing Dataset

When the desired structure already exists in another dataset, the fastest way to create an empty shell is using the SET statement. This method copies all variable definitions, including attributes like formats and labels, from the source dataset into the new one, bypassing the need for extensive ATTRIB statements.

data empty_data;
    set existing_data;
    stop;
run;

Understanding the Critical Role of the STOP Statement

In both creation methods, the inclusion of the STOP statement is paramount to achieving a truly empty dataset. When SAS executes a DATA step, it typically loops through the input data until the end-of-file is reached, creating an observation during each iteration. The STOP statement immediately halts the execution of the DATA step before the first observation is written to the output dataset.

If the STOP statement were omitted in Method 1 (Creating from Scratch), the DATA step would execute once, creating a single observation containing missing values for all defined variables, thus resulting in a dataset that is technically not empty. In Method 2, omitting STOP would cause the entire source dataset to be copied, defeating the purpose of creating an empty schema.

The result of utilizing the STOP statement is a dataset object containing defined variable names and metadata, but with a record count of zero observations. The following practical examples demonstrate the application and verification of both methods.

Example 1: Defining a New Dataset Schema from Scratch

This example demonstrates Method 1 by defining a new dataset named empty_data. We specify four variables relevant to employee sales tracking, ensuring we establish the correct data type (numeric or character) and presentation (format) for each column using the ATTRIB statement.

Notice the detailed assignment of attributes:

• employee_ID and sales are defined as LENGTH=8 (standard numeric).
• employee_Name is defined as LENGTH=$30 (a character variable of length 30).
• A specific date format (date9.) is applied to sales_date.

/*create empty dataset*/
data empty_data;
    attrib 
    employee_ID length=8 format=best12. label="Employee ID"
    employee_Name length=$30 format=$30. label="Employee Name"
    sales length=8 format=best12. label="Sales"
    sales_date length=8 format=date9. label="Sales Date";
    stop;
run;

Verifying the Empty Dataset Structure

To confirm that the dataset was created successfully with the required metadata but without any records, we use the PROC CONTENTS procedure. This procedure reads the descriptor portion of the dataset, providing key information about its structure and attributes.

/*view contents of dataset*/
proc contents data=empty_data;

The resulting output clearly shows that the dataset possesses the necessary structural elements (four variables defined), but the crucial metric, Observations, is listed as zero. This confirms the dataset is truly empty of data rows.

Furthermore, reviewing the variable listing confirms the successful application of the ATTRIB statement, detailing the variable name, type, length, and assigned label for all columns:

Example 2: Leveraging Existing Structure for Metadata Cloning

Method 2 is often preferred for efficiency, especially when dealing with complex data structures or when ensuring compatibility with an existing system. Here, we create an empty dataset named empty_dat by referencing the structure of sashelp.Comet, which is a predefined dataset built into the SAS system library.

The SET sashelp.Comet; line instructs the DATA step to inherit all structural properties of the source dataset. The subsequent STOP; statement prevents the reading of any actual data records, ensuring only the structure is copied.

/*create empty dataset from existing dataset*/
data empty_dat;
    set sashelp.Comet;
    stop;
run;

Validating the Copied Structure

Just as in the previous example, we use PROC CONTENTS to inspect the newly created dataset empty_dat and verify that the structural copy was successful without including any observations.

/*view contents of dataset*/
proc contents data=empty_data;

The output confirms that empty_dat successfully adopted the schema from sashelp.Comet, listing all its variables. Crucially, the observation count remains at zero, confirming the successful use of the SET and STOP combination for metadata cloning.

A closer look at the variables section shows the inherited variable names and their predefined attributes, ready for data insertion:

Conclusion and Next Steps

Mastering the creation of empty datasets is a crucial skill for any SAS programmer, enabling robust data management and streamlined ETL processes. Whether you manually define the structure using ATTRIB or leverage existing definitions using SET, the key is the efficient use of the STOP statement to preserve the schema while ensuring zero observations.

Once the empty structure is established, the next logical step involves techniques for population, appending data, or merging multiple sources into this new template. Understanding how to manage the dataset metadata is foundational for more advanced data manipulation tasks.

The following tutorials explain how to perform other common tasks in SAS: