Opening a base table cursor

A base table cursor is a means to read and write table data without using the SQL query processor. Through this object, one can read and write data in an ISAM-like fashion. All you need to know is the table name and optionally an index name. To open a base table cursor and read data from it, we must do quite a bit of work. We start by getting an IOpenRowset pointer from our session object through QueryInterface.

IOpenRowset
This interface has only one method: OpenRowset. Set the second parameter to the requested table name (set the DBID structure with eKind = DBKIND_NAME and pwszName pointing to a string containing the UNICODE table name) and the third to the optional index name (set to NULL if you don't want to use any index order). Set the fourth parameter to the IID_IRowset constant so that upon success you get an IRowset pointer in the last parameter. The remaining two parameters are used to specify the requested cursor properties. After successfully calling this method, you get an IRowset interface pointer.

IRowset
The IRowset interface is essentially used to navigate through the data rows (RestartPosition and GetNextRows) and to read data from them (GetData). This is done independently of the data structure of the table (or query result). To map between the data row structure and the consumer's memory we need another interface.

IAccessor
To retrieve data from the row, one or more accessors must be created through the IAccessor interface (retrieved via QueryInterface on the IRowset). Each accessor maps a set of columns in the data row to memory addresses in the consumer memory through CreateAccessor. Each column mapping is represented by an entry in the DBBINDING array that the cosumer passes as the third parameter. This array may either be filled with application-provided bindings (if you just want to bind a subset of the columns), or with data from the provider itself. To get this type of information, you typically request an IColumnsInfo interface pointer and call the GetColumnInfo method to retrieve all the column information for the table. The process of mapping between the two is not linear and requires some thought and attention to the provider's peculiarities. But we will see all of this on the next post.