The article is available at : http://msdn.microsoft.com/msdnmag/issues/04/06/DataPoints/
Salient points of the article:
- A DataReader is a stream of data that is returned from a database query. When the query is executed, the first row is returned to the DataReader via the stream. The stream then remains connected to the database, poised to retrieve the next record. The DataReader reads one row at a time from the database and can only move forward, one record at a time. As the DataReader reads the rows from the database, the values of the columns in each row can be read and evaluated, but they cannot be edited.
- The DataReader can only get its data from a data source through a managed provider. The DataSet can also get its data from a data source via a managed provider, but the data source can also be loaded manually, even from an XML file on a hard drive.
- The DataReader supports access to multiple resultsets, one at a time, in the order they are retrieved. (Methods to checkout are NextResult() and Read() )
- When the SqlDataAdapter's Fill method is executed it opens its associated SqlConnection object (if not already open) and issues its associated SqlCommand object against the SqlConnection. Behind the scenes, a SqlDataReader is created implicitly and the rowset is retrieved one row at a time in succession and sent to the DataSet. Once all of the data is in the DataSet, the implicit SqlDataReader is destroyed and the SqlConnection is closed.
- The DataSet's disconnected nature allows it to be transformed into XML and sent over the wire via HTTP if appropriate. This makes it ideal as the return vehicle from business-tier objects and Web services. A DataReader cannot be serialized and thus cannot be passed between physical-tier boundaries where only string (XML) data can go.
- There are other times when a DataReader can be the right choice, such as when populating a list or retrieving 10,000 records for a business rule. When a huge amount of data must be retrieved to a business process, even on a middle tier, it can take a while to load a DataSet, pass the data to it on the business tier from the database, and then store it in memory. The footprint could be quite large and with numerous instances of it running (such as in a Web application where hundreds of users may be connected), scalability would become a problem. If this data is intended to be retrieved and then traversed for business rule processing, the DataReader could speed up the process as it retrieves one row at a time and does not require the memory resources that the DataSet requires.