IndexingConnector

Initializes the connector with the current context.

Initialization is the ideal time to start any threads to do extra behind-the-scenes work. The context allows access to other useful objects that can be used at any time such as the IndexingService.

If an unrecoverable fatal error occurs during initialization, throw a StartupException to cancel the startup process. If a recoverable error occurs during initialization, most Exception instances other than StartupException cause a retry of initialization after a short delay.

Implements a connector-specific traversal strategy.

Use the CloudSearch.Indexing.Datasources.Items.Push method to push a list of PushItem instances that are to be indexed. Often, many of the sent items are unchanged since the last traversal and are thus being re-pushed unnecessarily. However, the cost of pushing unchanged items is low and doing so heals any synchronization errors or cache inconsistencies that might exist in the system.

For flat or full traversal strategies, push PushItem instances for all items to be indexed. For graph traversal strategies, push only the start node of each graph.

Alternatively, a Connector implementation can issue an CloudSearch.Indexing.Datasources.Items.Index for each item in the repository instead of using CloudSearch.Indexing.Datasources.Items.Push and CloudSearch.Indexing.Datasources.Items.Poll. Such implementations might be suitable for smaller repositories requiring only a periodic full sync.

This method may take a while and implementations may want to call sleep(long) occasionally to reduce load.

If fatal errors occur, throw an IOException or RuntimeException. In the case of an error, the ExceptionHandler defined in IndexingConnectorContext determines if and when to retry.