Programmatically scrolling WebBrowser control from Visual C/C++

Introduction

For some time I struggled to do these simple tasks in Visual C++:

1. How to get browser window to scroll programmatically?

   It does not accept Windows scroll messages. Calling the MFC GetScrollInfo APIs does not do anything. But there is a way.

2. To scroll we use IHTMLElement2 API get_scrollTop/put_scrollTop... but how do I obtain IHTMLElement2?

The way is a bit obscure...

Now, the code

//
// All this code does is what
// "m_browser.Document.Body.ScrollTop = 100;"
// does in VB. Gotta love COM in C++.
//

// let's say m_browser is the WebBrowser's member variable.

HRESULT hr;

// get the document dispatch from browser
IDispatch *pDisp = m_browser.GetDocument();
ASSERT( pDisp ); //if NULL, we failed

// get document interface
IHTMLDocument2 *pDocument = NULL;
hr = pDisp->QueryInterface( IID_IHTMLDocument2, (void**)&pDocument );
ASSERT( SUCCEEDED( hr ) );
ASSERT( pDocument );

//
// this is the trick!
// take the body element from document...
//
IHTMLElement *pBody = NULL;
hr = pDocument->get_body( &pBody );
ASSERT( SUCCEEDED( hr ) );
ASSERT( pBody );

// from body we can get element2 interface,
// which allows us to do scrolling
IHTMLElement2 *pElement = NULL;
hr = pBody->QueryInterface(IID_IHTMLElement2,(void**)&pElement);
ASSERT(SUCCEEDED(hr));
ASSERT( pElement );

// now we are ready to scroll
// scroll down to 100th pixel from top
pElement->put_scrollTop( 100 );

// try to get the whole page size - but the returned number
// is not allways correct. especially with pages that use dynamic html
// tricks...
long scroll_height;
pElement->get_scrollHeight( &s );

// we can use this workaround!
long real_scroll_height;
pElement->put_scrollTop( 20000000 ); // ask to scroll really far down...
pElement->get_scrollTop( &real_scroll_height );
real_scroll_height += window_height; // will return the scroll height
// for the first visible pixel, to get whole html page size must
// add the window's height... (to obtain window_height is
// left as an exercise for the reader)


// print to debug output
TRACE( "real scroll height: %ld, get_scrollHeight: %ld\n",
                 real_scroll_height, scroll_height );

And there we have it.