Use sparse columns

Applies to: SQL Server 2016 (13.x) and later Azure SQL Database Azure SQL Managed Instance

Sparse columns are ordinary columns that have an optimized storage for null values. Sparse columns reduce the space requirements for null values at the cost of more overhead to retrieve non-NULL values. Consider using sparse columns when the space saved is at least 20 percent to 40 percent. Sparse columns and column sets are defined by using the CREATE TABLE or ALTER TABLE statements.

Sparse columns can be used with column sets and filtered indexes:

  • Column sets

    INSERT, UPDATE, and DELETE statements can reference the sparse columns by name. However, you can also view and work with all the sparse columns of a table that are combined into a single XML column. This column is called a column set. For more information about column sets, see Use Column Sets.

  • Filtered indexes

    Because sparse columns have many null-valued rows, they are especially appropriate for filtered indexes. A filtered index on a sparse column can index only the rows that have populated values. This creates a smaller and more efficient index. For more information, see Create Filtered Indexes.

Sparse columns and filtered indexes enable applications, such as Windows SharePoint Services, to efficiently store and access a large number of user-defined properties by using SQL Server.

Properties of Sparse Columns

Sparse columns have the following characteristics:

  • The SQL Server Database Engine uses the SPARSE keyword in a column definition to optimize the storage of values in that column. Therefore, when the column value is NULL for any row in the table, the values require no storage.

  • Catalog views for a table that has sparse columns are the same as for a typical table. The sys.columns catalog view contains a row for each column in the table and includes a column set if one is defined.

  • Sparse columns are a property of the storage layer, rather than the logical table. Therefore a SELECT ... INTO statement does not copy over the sparse column property into a new table.

  • The COLUMNS_UPDATED function returns a varbinary value to indicate all the columns that were updated during a DML action. The bits that are returned by the COLUMNS_UPDATED function are as follows:

    • When a sparse column is explicitly updated, the corresponding bit for that sparse column is set to 1, and the bit for the column set is set to 1.

    • When a column set is explicitly updated, the bit for the column set is set to 1, and the bits for all the sparse columns in that table are set to 1.

    • For insert operations, all bits are set to 1.

    For more information about columns sets, see Use Column Sets.

The following data types cannot be specified as SPARSE:

geography
geometry
image
ntext

text
timestamp
user-defined data types

Estimated space savings by data type

Sparse columns require more storage space for non-NULL values than the space required for identical data that is not marked SPARSE. The following tables show the space usage for each data type. The NULL Percentage column indicates what percent of the data must be NULL for a net space savings of 40 percent.

Fixed-Length Data Types

Data type Nonsparse bytes Sparse bytes NULL percentage
bit 0.125 5 98%
tinyint 1 5 86%
smallint 2 6 76%
int 4 8 64%
bigint 8 12 52%
real 4 8 64%
float 8 12 52%
smallmoney 4 8 64%
money 8 12 52%
smalldatetime 4 8 64%
datetime 8 12 52%
uniqueidentifier 16 20 43%
date 3 7 69%

Precision-Dependent-Length Data Types

Data type Nonsparse bytes Sparse bytes NULL percentage
datetime2(0) 6 10 57%
datetime2(7) 8 12 52%
time(0) 3 7 69%
time(7) 5 9 60%
datetimetoffset(0) 8 12 52%
datetimetoffset (7) 10 14 49%
decimal/numeric(1,s) 5 9 60%
decimal/numeric(38,s) 17 21 42%
vardecimal(p,s) Use the decimal type as a conservative estimate.

Data-Dependent-Length Data Types

Data type Nonsparse bytes Sparse bytes NULL percentage
sql_variant Varies with the underlying data type
varchar or char 2* 4* 60%
nvarchar or nchar 2* 4*+ 60%
varbinary or binary 2* 4* 60%
xml 2* 4* 60%
hierarchyid 2* 4* 60%

*The length is equal to the average of the data that is contained in the type, plus 2 or 4 bytes.

In-Memory overhead required for updates to sparse columns

When designing tables with sparse columns, keep in mind that an additional 2 bytes of overhead are required for each non-null sparse column in the table when a row is being updated. As a result of this additional memory requirement, updates can fail unexpectedly with error 576 when the total row size, including this memory overhead, exceeds 8019, and no columns can be pushed off the row.

Consider the example of a table that has 600 sparse columns of type bigint. If there are 571 non-null columns, then the total size on disk is 571 * 12 = 6852 bytes. After including additional row overhead and the sparse column header, this increases to around 6895 bytes. The page still has around 1124 bytes available on disk. This can give the impression that additional columns can be updated successfully. However, during the update, there is additional overhead in memory which is 2*(number of non-null sparse columns). In this example, including the additional overhead - 2 * 571 = 1142 bytes - increases the row size on disk to around 8037 bytes. This size exceeds the maximum allowed size of 8019 bytes. Since all the columns are fixed-length data types, they cannot be pushed off the row. As a result, the update fails with the 576 error.

Restrictions for using sparse columns

Sparse columns can be of any SQL Server data type and behave like any other column with the following restrictions:

  • A sparse column must be nullable and cannot have the ROWGUIDCOL or IDENTITY properties. A sparse column cannot be of the following data types: text, ntext, image, timestamp, user-defined data type, geometry, or geography; or have the FILESTREAM attribute.

  • A sparse column cannot have a default value.

  • A sparse column cannot be bound to a rule.

  • Although a computed column can contain a sparse column, a computed column cannot be marked as SPARSE.

  • A data mask can be defined on a sparse column, but not on a sparse column that is part of a column set.

  • A sparse column cannot be part of a clustered index or a unique primary key index. However, both persisted and non-persisted computed columns that are defined on sparse columns can be part of a clustered key.

  • A sparse column cannot be used as a partition key of a clustered index or heap. However, a sparse column can be used as the partition key of a nonclustered index.

  • A sparse column cannot be part of a user-defined table type, which are used in table variables and table-valued parameters.

  • Sparse columns are incompatible with data compression. Therefore sparse columns cannot be added to compressed tables, nor can any tables containing sparse columns be compressed.

  • Changing a column from sparse to non-sparse, or non-sparse to sparse, requires changing the storage format of the column. The SQL Server Database Engine uses the following procedure to accomplish this change:

    1. Adds a new column to the table in the new storage size and format.

    2. For each row in the table, updates and copies the value stored in the old column to the new column.

    3. Removes the old column from the table schema.

    4. Rebuilds the table (if there is no clustered index) or rebuilds the clustered index to reclaim space used by the old column.

    Note

    Step 2 can fail when the size of the data in the row exceeds the maximum allowable row size. This size includes the size of the data stored in the old column and the updated data stored in the new column. This limit is 8060 bytes for tables that do not contain any sparse columns or 8018 bytes for tables that contain sparse columns. This error can occur even if all eligible columns have been pushed off-row.

  • When you change a non-sparse column to a sparse column, the sparse column will consume more space for non-null values. When a row is close to the maximum row size limit, the operation can fail.

SQL Server technologies that support sparse columns

This section describes how sparse columns are supported in the following SQL Server technologies:

  • Transactional replication

    Transactional replication supports sparse columns, but it does not support column sets, which can be used with sparse columns. For more information about column sets, see Use Column Sets.

    The replication of the SPARSE attribute is determined by a schema option that is specified by using sp_addarticle or by using the Article Properties dialog box in SQL Server Management Studio. Earlier versions of SQL Server do not support sparse columns. If you must replicate data to an earlier version, specify that the SPARSE attribute should not be replicated.

    For tables that are published, you cannot add any new sparse columns to a table or change the sparse property of an existing column. If such an operation is required, drop and re-create the publication.

  • Merge replication

    Merge replication does not support sparse columns or column sets.

  • Change tracking

    Change tracking supports sparse columns and column sets. When a column set is updated in a table, change tracking treats this as an update to the whole row. No detailed change tracking is provided to obtain the exact set of sparse columns that are updated through the column set update operation. If the sparse columns are updated explicitly through a DML statement, change tracking on them will work ordinarily and can identify the exact set of changed columns.

  • Change data capture

    Change data capture supports sparse columns, but it does not support column sets.

  • The sparse property of a column is not preserved when the table is copied.

Examples

In this example, a document table contains a common set that has the columns DocID and Title. The Production group wants a ProductionSpecification and ProductionLocation column for all production documents. The Marketing group wants a MarketingSurveyGroup column for marketing documents. The code in this example creates a table that uses sparse columns, inserts two rows into the table, and then selects data from the table.

Note

This table has only five columns to make it easier to display and read. Declaring the sparse columns to be nullable is optional if the ANSI_NULL_DFLT_ON option is set. When SET ANSI_DEFAULTS is ON, SET ANSI_NULL_DFLT_ON is enabled. ANSI_DEFAULTS is ON by default for most connection providers. For more information, see SET ANSI_DEFAULTS.

USE AdventureWorks2022;  
GO  
  
CREATE TABLE DocumentStore  
    (DocID int PRIMARY KEY,  
     Title varchar(200) NOT NULL,  
     ProductionSpecification varchar(20) SPARSE NULL,  
     ProductionLocation smallint SPARSE NULL,  
     MarketingSurveyGroup varchar(20) SPARSE NULL ) ;  
GO  
  
INSERT DocumentStore(DocID, Title, ProductionSpecification, ProductionLocation)  
VALUES (1, 'Tire Spec 1', 'AXZZ217', 27);  
GO  
  
INSERT DocumentStore(DocID, Title, MarketingSurveyGroup)  
VALUES (2, 'Survey 2142', 'Men 25 - 35');  
GO  

To select all the columns from the table returns an ordinary result set.

SELECT * FROM DocumentStore ;  

Here is the result set.

DocID Title ProductionSpecification ProductionLocation MarketingSurveyGroup

1 Tire Spec 1 AXZZ217 27 NULL

2 Survey 2142 NULL NULL Men 25-35

Because the Production department is not interested in the marketing data, they want to use a column list that returns only columns of interest, as shown in the following query.

SELECT DocID, Title, ProductionSpecification, ProductionLocation   
FROM DocumentStore   
WHERE ProductionSpecification IS NOT NULL ;  

Here is the result set.

DocID Title ProductionSpecification ProductionLocation

1 Tire Spec 1 AXZZ217 27

See also