Working with Query Notifications
Applies To: SQL Server 2016 Preview
Query notifications were introduced in SQL Server 2005 and SQL Server Native Client. Built upon the Service Broker infrastructure introduced in SQL Server 2005, query notifications allow applications to be notified when data has changed. This feature is particularly useful for applications that provide a cache of information from a database, such as a Web application, and need to be notified when the source data is changed.
Query notifications allow you to request notification within a specified time-out period when the underlying data of a query changes. The request for notification specifies the notification options, which include the service name, message text, and time-out value to the server. Notifications are delivered through a Service Broker queue that applications may poll for available notifications.
The syntax of the query notifications options string is:
service=<service-name>[;(local database=<database> | broker instance=<broker instance>)]
Notification subscriptions outlive the process that initiates them, as an application may create a notification subscription and then terminate. The subscription remains valid, and the notification will occur if the data changes within the time-out period specified when the subscription was created. A notification is identified by the query executed, the notification options, and the message text, and may be cancelled by setting its time-out value to zero.
Notifications are sent only once. For continuous notification of data change, a new subscription must be created by re-executing the query after each notification is processed.
SQL Server Native Client applications typically receive notifications by using the Transact-SQL RECEIVE command to read notifications from the queue associated with the service specified in the notification options.
Table names must be qualified in queries for which notification is required, for example, dbo.myTable. Table names must be qualified with two part names. Subscription is invalid if three- or four-part names are used.
The notification infrastructure is built on top of a queuing feature introduced in SQL Server 2005. In general, notifications generated at the server are sent through these queues to be processed later.
To use query notifications a queue and a service must exist on the server. These can be created using Transact-SQL similar to the following:
CREATE QUEUE myQueue CREATE SERVICE myService ON QUEUE myQueue ([http://schemas.microsoft.com/SQL/Notifications/PostQueryNotification])
The service must use the predefined contract http://schemas.microsoft.com/SQL/Notifications/PostQueryNotification as shown above.
The SQL Server Native Client OLE DB provider supports consumer notification on rowset modification. The consumer receives notification at every phase of rowset modification and on any attempted change.
Passing a notifications query to the server with ICommand::Execute is the only valid way to subscribe to query notifications with the SQL Server Native Client OLE DB provider.
In order to support query notifications through OLE DB, SQL Server Native Client adds the following new properties to the DBPROPSET_SQLSERVERROWSET property set.
The number of seconds that the query notification is to remain active.
The default is 432000 seconds (5 days). The minimum value is 1 second, and the maximum value is 2^31-1 seconds.
The message text of the notification. This is user defined, and has no predefined format.
By default, the string is empty. You can specify a message using 1-2000 characters.
The query notification options. These are specified in a string with name=value syntax. The user is responsible for creating the service and reading notifications off of the queue.
The default is an empty string.
The notification subscription is always committed, regardless of whether the statement ran in a user transaction or in auto commit or whether the transaction in which the statement ran committed or rolled back. The server notification fires upon any of the following invalid notification conditions: change of underlying data or schema, or when the timeout period is reached; whichever is first. Notification registrations are deleted as soon as they are fired. Hence upon receiving notifications, the application must subscribe again in case they want to get further updates.
Another connection or thread can check the destination queue for notifications. For example:
WAITFOR (RECEIVE * FROM MyQueue); // Where MyQueue is the queue name.
Note that SELECT * does not delete the entry from the Queue, however RECEIVE * FROM does. This stalls a server thread if the queue is empty. If there are queue entries at the time of the call, they are returned immediately; otherwise the call waits until a queue entry is made.
RECEIVE * FROM MyQueue
This statement immediately returns an empty result set if the queue is empty; otherwise it returns all queue notifications.
If SSPROP_QP_NOTIFICATION_MSGTEXT and SSPROP_QP_NOTIFICATION_OPTIONS are non-NULL and non-empty, the query notifications TDS header containing the three properties defined above are sent to the server with each execution of the command. If either of them is null (or empty), the header is not sent and DB_E_ERRORSOCCURRED is raised, (or DB_S_ERRORSOCCURRED if the properties are both marked as optional), and the status value is set to DBPROPSTATUS_BADVALUE. The validation occurs on Execute/Prepare. Similarly, DB_S_ERRORSOCCURED is raised when the query notification properties are set for connections to SQL Server versions before SQL Server 2005. The status value in this case is DBPROPSTATUS_NOTSUPPORTED.
Initiating a subscription does not guarantee that subsequent messages will be successfully delivered. In addition, no check is made as to the validity of the service name specified.
Preparing statements will never cause the subscription to be initiated; only statement execution will achieve this and query notifications are not impacted by the use of OLE DB core services.
For more information about the DBPROPSET_SQLSERVERROWSET property set, see Rowset Properties and Behaviors.
If SQL_SOPT_SS_QUERYNOTIFICATION_MSGTEXT and SQL_SOPT_SS_QUERYNOTIFICATION_OPTIONS are not NULL, the query notifications TDS header containing the three attributes defined above will be sent to the server each time the command is executed. If either of them is null, the header is not sent, and SQL_SUCCESS_WITH_INFO is returned. The validation occurs on SQLPrepare Function, SqlExecDirect, and SqlExecute, all of which fail if the attributes are not valid. Similarly, when these query notification attributes are set for SQL Server versions before SQL Server 2005, the execution fails with SQL_SUCCESS_WITH_INFO.
Prepare statements will never cause the subscription to be initiated; subscription can be initiated by statement execution.
The following data types are not supported for notifications:
If a notification request is made for a query which returns any of these types, the notification fires immediately, specifying that notification subscription was not possible.
If a subscription request is made for a batch or stored procedure, a separate subscription request is made for each statement executed within the batch or stored procedure. EXECUTE statements will not register a notification, but will send the notification request to the executed command. If it is a batch, the context will be applied to the executed statements and the same rules described above apply.
Submission of a query for notification that was submitted by the same user under the same database context and has the same template, same parameter values, same notification ID, and same delivery location of an existing active subscription, will renew the existing subscription, resetting the new specified time-out. This means that if notification is requested for identical queries, only one notification will be sent. This would apply to a query duplicated in a batch, or a query in a stored procedure that was called multiple times.