INFO: Using ActiveX Data Objects (ADO) via #import in VC++
ID: Q169496
|
The information in this article applies to:
-
ActiveX Data Objects (ADO), versions 1.0, 1.5, 2.0, used with:
-
Microsoft Visual C++, 32-bit Editions, versions 4.2, 5.0, 6.0
SUMMARY
The #import directive in Visual C++ offers a powerful new mechanism for
manipulating OLE servers. When used with ActiveX Data Objects (ADO),
#import can simplify getting at your data. This article discusses what is
necessary to take advantage of #import with ADO.
MORE INFORMATION
Before You Instantiate Any Classes Created by #import
It's important to initialize OLE before creating any instances of classes
created by #import. For example, the following code is safe, as it declares a #import smart pointer, initializes OLE, and then instantiates the smart
pointer:
// Declare smart pointer of Recordset
_RecordsetPtr p;
void main( void )
{
// Initialize OLE.
::CoInitialize(NULL);
// Instantiate smart pointer.
HRESULT hr = p.CreateInstance( __uuidof( Recordset ) );
...
}
The next code sample, however, is not safe and generates an unhandled
exception. The global smart pointer p is both declared and instantiated (by
virtue of passing a specific uuid in the constructor):
// Declare & instantiate smart pointer of Recordset
_RecordsetPtr p( __uuidof( _Recordset ) );
void main( void )
{
// Initialize OLE
::CoInitialize(NULL);
...
}
Because p is a global variable, it is instantiated before CoInitialize is
ever called in main(). You can correct this with the following code snippet :
struct InitOle {
InitOle() { ::CoInitialize(NULL); }
~InitOle() { ::CoUninitialize(); }
} _init_InitOle_;
// Declare & instantiate smart pointer of Recordset
_RecordsetPtr p( __uuidof( _Recordset ) );
...
An instance of the struct InitOle is declared, and instantiated before p,
and, therefore, initializes OLE in it's constructor. Without this kind of
fail-safe, you will see the following error message:
Unhandled exception in [Program] (KERNEL32.DLL): 0xE06D7363
Microsoft C++ Exception.
Correct Implementation of #import
It is important to invoke ADO correctly in your program, or you can have
compiler errors. The following code demonstrates the correct way to use #import with Msado10.dll the MSADO15.dll:
#import <msado15.dll> \
no_namespace \
rename( "EOF", "adoEOF" )
Error Handling
With ADO, you may get an error in the HRESULT returned from an ADO method,
you may get an exception raised by #import generated classes, and for
either condition the ADO Errors Collection may be populated. In order to
get at the Errors Collection you need a valid connection object. For more information please see the following article in the Microsoft Knowledge Base:
Q169498 INFO: Extracting Error Information from ADO in VC++ with #import
ADO and Dbdaoint.h
Attempts to mix ADO (through #import) and either MFC DAO or the DAO SDK in the same implementation file, as follows:
#include <afxdao.h> // MFC DAO
// -or-
#include <dbdao.h> // DAO SDK
#import <msado15.dll> no_namespace ...
Generates the following six errors:
error C2011: 'EditModeEnum' : 'enum' type redefinition
error C2011: 'LockTypeEnum' : 'enum' type redefinition
error C2011: 'FieldAttributeEnum' : 'enum' type redefinition
error C2011: 'DataTypeEnum' : 'enum' type redefinition
error C2011: 'ParameterDirectionEnum' : 'enum' type redefinition
error C2011: 'RecordStatusEnum' : 'enum' type redefinition
While very nearly identical in content, the actual values in each
enumerated type differ between what is required by ADO and what is required
by DAO. You have several options to work around this:
- Separate ADO and DAO code into separate .cpp files. Keep the use of
#import or #include <afxdao.h/dbdao.h> in separate implementation
files as well.
- Modify the #import statement to create a namespace for anything
generated for ADO. This means you will have to reference the
namespace when referencing an ADO object as shown in the two functions
below. The first shows how to use ADO exclusively within a function.
The second shows how to mix-and-match ADO and DAO objects. This is
possible only by explicitly referencing the ADO namespace for any ADO
class or enumerated type:
#include <afxdao.h>
#import <msado15.dll> \
rename_namespace("AdoNS") rename( "EOF", "adoEOF" )
void ADOOnly( void )
{
using namespace AdoNS;
_RecordsetPtr prs;
// Generates Compile Errors:
CDaoRecordset rs;
}
void MixAdoAndDao( void )
{
AdoNS::_RecordsetPtr prs;
// Compiles just fine
CDaoRecordset drs;
}
Dissecting and using Msado105.tlh/Msado15.tli --------------------------------------------
#import generates two files, Msado105.tlh and Msado15.tli off of the typelib contained within Msado15.dll. The structure of the .tlh file can be broken out as follows:
- Forward References and Typedefs
- Smart Pointer Typedef and Declarations
- Type Library Items
Each is described in detail below.
Forward References and Typedefs
Forward References and Typedefs are created through the use of struct
__declspec(uuid("...")) on the GUID for any Dual Interface, Interface, and
CoClass defined in the typelib.
...
struct __declspec(uuid("00000274-0000-0010-8000-00aa006d2ea4"))
/* dual interface */ _Connection;
...
struct __declspec(uuid("00000275-0000-0010-8000-00aa006d2ea4"))
/* interface */ ICADOConnection;
...
struct /* coclass */ Connection;
...
Not all interfaces, such as Connection, have multiple implementations.
This depends on the typelib, but for ADO most interfaces are dual and not
implemented as interface or coclass.
Smart Pointer TypeDef Declarations
For Interfaces and Dual Interfaces, smart pointers are declared, which
greatly simplifies using the interface:
...
_COM_SMARTPTR_TYPEDEF(_Connection, __uuidof(_Connection));
...
_COM_SMARTPTR_TYPEDEF(ICADOConnection, __uuidof(ICADOConnection));
...
Note that no smart pointer was declared for the coclass Connection
interface.
Type Library Items
This includes any enumerated types defined in the typelib, as well
implementation of the smart pointers and typelib items:
enum CursorTypeEnum
{
adOpenUnspecified = -1,
adOpenForwardOnly = 0,
adOpenKeyset = 1,
adOpenDynamic = 2,
adOpenStatic = 3
};
...
struct __declspec(uuid("00000274-0000-0010-8000-00aa006d2ea4"))
_Connection : _ADO
{
//
// Property data.
//
_declspec(property(get=GetConnectionString,
put=PutConnectionString))
_bstr_t ConnectionString;
...
//
// Wrapper methods for error-handling.
//
_bstr_t GetConnectionString ( );
void PutConnectionString (
_bstr_t pbstr );
...
//
// Raw methods provided by interface.
//
virtual HRESULT __stdcall get_ConnectionString (
BSTR * pbstr ) = 0;
virtual HRESULT __stdcall put_ConnectionString (
BSTR pbstr ) = 0;
...
};
In the preceding code fragment, the Property Data section uses declspec to
declare get and put methods for ConnectionString. The Wrapper methods
section provides methods created by #import, which wrap these methods, and
raise an _com_error exception if they are not successful. The Raw
Methods section declares the actual method that is invoked by the
interface.
While you could call GetConnectionString or PutConnectionString, it is
really unnecessary. Since ConnectionString is a property you would
reference it as follows:
_ConnectionPtr p;
bstr bstrConnect;
...
bstrConnect = SysAllocString( L"DSN=AdoDemo;UID=admin;PWD=sa" );
p->ConnectionString = bstrConnect;
The actual implementation of GetConnectionString/PutConnectionString can be
found in the Msado15.tli file.
When it comes time to use the Connection object in your code, you would use
an instance of the smart pointer for the dual interface defined in
Msado15.tlh as follows:
_ConnectionPtr p;
bstrConnect
HRESULT hr = S_OK;
_ConnectionPtr pConn;
hr = pConn.CreateInstance( __uuidof( Connection ) );
if( !FAILED( hr ) )
hr = pConn->Open( L"pubs", L"sa", L"" );
Where pubs is an ODBC data source.
#import and Explicitly Calling Release()
The advantage of #import is that it takes care of AddRef, QueryInterface,
and Release for you automatically. However, if you decide to start calling
Release() explicitly, you can create problems for yourself.
Within _com_ptr_t is a member variable, m_pInterface. As #import is a very
thin wrapper, it makes no distinction with m_pInterface after the object is
actually released, versus just decrementing its reference count without
actually destroying the object. By explicitly calling Release()--without
very explicitly calling AddRef() to balance it--#import will gladly try to
release an object that doesn't exist, creating interesting side effects and
crashing behavior.
Best advice, you did not AddRef() it (or at least no need to), do not release it either.
REFERENCES
- Inside Com by Dale Rogerson ISBN 1-57231-349-8
- The OLE-COM Object Viewer (Oleview.exe) that ships with Visual C++ for examining the contents of a typelib.
- Visual C++ online documentation: search on #import
For more information, please see the following articles in the Microsoft Knowledge Base:
Q182389 FILE: Adovcbm.exe ADO 1.5 with #import and Getrows/Bookmarks
Q184968 FILE: Adovcsp.exe Demonstrates Using Stored Procedures with ADO
Q186387 SAMPLE: Ado2atl.exe Returns ADO Interfaces from COM
Q181733 FILE: Adovcbtd.exe #import Using UpdateBatch and CancelBatch
Q166112 PRB: Conflict with EOF when using #import with ADO
Q168354 INFO: Underlying OLE and OLEDB Provider Errors Exposed via ADO
Additional query words:
kbdse s_mfc kbadoall kbadovc
Keywords : kbcode kbusage kbADO kbVC
Version : WINDOWS:1.0,1.5,2.0
Platform : WINDOWS
Issue type : kbinfo
Last Reviewed: February 15, 1999