I have embedded a web browser control in my c++ application. I want javascript running in the web browser control to be able to call a c++ function/method.
I have found mentions of three ways to do this:
- Implement an ActiveX component that acts as a middle man. (Implementation details here: http://blogs.msdn.com/b/nicd/archive/2007/04/18/calling-into-your-bho-from-a-client-script.aspx)
- Use window.external. (Also discussed in the link above, but no implementation provided)
- Add a custom object to the window object
I want to go with the third option, but I haven't found any working examples on how to do that. Can someone please show me how to do it, or link to a working example on the net somewhere.
The closest to an example that I have found is the first reply by Igor Tandetnik in a thread in the webbrowser_ctl news group. But I'm afraid I need more help than that.
I'm embedding an IWebBrowser2 control and I am not using MFC, ATL or WTL.
EDIT:
Going by the pseudo-code given by Igor in the thread I linked earlier, and code found in the codeproject article "Creating JavaScript arrays and other objects from C++" I've produced some code.
void WebForm::AddCustomObject(IDispatch *custObj, std::string name)
{
IHTMLDocument2 *doc = GetDoc();
IHTMLWindow2 *win = NULL;
doc->get_parentWindow(&win);
if (win == NULL) {
return;
}
IDispatchEx *winEx;
win->QueryInterface(&winEx);
if (winEx == NULL) {
return;
}
int lenW = MultiByteToWideChar(CP_ACP, MB_PRECOMPOSED, name.c_str(), -1, NULL, 0);
BSTR objName = SysAllocStringLen(0, lenW);
MultiByteToWideChar(CP_ACP, MB_PRECOMPOSED, name.c_str(), -1, objName, lenW);
DISPID dispid;
HRESULT hr = winEx->GetDispID(objName, fdexNameEnsure, &dispid);
SysFreeString(objName);
if (FAILED(hr)) {
return;
}
DISPID namedArgs[] = {DISPID_PROPERTYPUT};
DISPPARAMS params;
params.rgvarg = new VARIANT[1];
params.rgvarg[0].pdispVal = custObj;
params.rgvarg[0].vt = VT_DISPATCH;
params.rgdispidNamedArgs = namedArgs;
params.cArgs = 1;
params.cNamedArgs = 1;
hr = winEx->InvokeEx(dispid, LOCALE_USER_DEFAULT, DISPATCH_PROPERTYPUT, ¶ms, NULL, NULL, NULL);
if (FAILED(hr)) {
return;
}
}
The code above runs all the way through, so everything looks alright that far.
I call AddCustomObject when I receive the DISPID_NAVIGATECOMPLETE2 DWebBrowserEvents2 event passing this as *custObj
:
class JSObject : public IDispatch {
private:
long ref;
public:
// IUnknown
virtual HRESULT STDMETHODCALLTYPE QueryInterface(REFIID riid, void **ppv);
virtual ULONG STDMETHODCALLTYPE AddRef();
virtual ULONG STDMETHODCALLTYPE Release();
// IDispatch
virtual HRESULT STDMETHODCALLTYPE GetTypeInfoCount(UINT *pctinfo);
virtual HRESULT STDMETHODCALLTYPE GetTypeInfo(UINT iTInfo, LCID lcid,
ITypeInfo **ppTInfo);
virtual HRESULT STDMETHODCALLTYPE GetIDsOfNames(REFIID riid,
LPOLESTR *rgszNames, UINT cNames, LCID lcid, DISPID *rgDispId);
virtual HRESULT STDMETHODCALLTYPE Invoke(DISPID dispIdMember, REFIID riid,
LCID lcid, WORD wFlags, DISPPARAMS *Params, VARIANT *pVarResult,
EXCEPINFO *pExcepInfo, UINT *puArgErr);
};
Noteworthy implementations might be
HRESULT STDMETHODCALLTYPE JSObject::QueryInterface(REFIID riid, void **ppv)
{
*ppv = NULL;
if (riid == IID_IUnknown || riid == IID_IDispatch) {
*ppv = static_cast<IDispatch*>(this);
}
if (*ppv != NULL) {
AddRef();
return S_OK;
}
return E_NOINTERFACE;
}
and
HRESULT STDMETHODCALLTYPE JSObject::Invoke(DISPID dispIdMember, REFIID riid,
LCID lcid, WORD wFlags, DISPPARAMS *Params, VARIANT *pVarResult,
EXCEPINFO *pExcepInfo, UINT *puArgErr)
{
MessageBox(NULL, "Invoke", "JSObject", MB_OK);
return DISP_E_MEMBERNOTFOUND;
}
Unfortunately I never get the "Invoke" message box when I try to use the "JSObject" object from the javascript code.
JSObject.randomFunctionName(); // This should give me the c++ "Invoke" message
// box, but it doesn't
EDIT 2:
I implemented GetIDsOfNames
like so:
HRESULT STDMETHODCALLTYPE JSObject::GetIDsOfNames(REFIID riid,
LPOLESTR *rgszNames, UINT cNames, LCID lcid, DISPID *rgDispId)
{
HRESULT hr = S_OK;
for (UINT i = 0; i < cNames; i++) {
std::map<std::wstring, DISPID>::iterator iter = idMap.find(rgszNames[i]);
if (iter != idMap.end()) {
rgDispId[i] = iter->second;
} else {
rgDispId[i] = DISPID_UNKNOWN;
hr = DISP_E_UNKNOWNNAME;
}
}
return hr;
}
and this is my constructor
JSObject::JSObject() : ref(0)
{
idMap.insert(std::make_pair(L"execute", DISPID_USER_EXECUTE));
idMap.insert(std::make_pair(L"writefile", DISPID_USER_WRITEFILE));
idMap.insert(std::make_pair(L"readfile", DISPID_USER_READFILE));
}
with the DISPID_USER_* constants defined as private class members
class JSObject : public IDispatch {
private:
static const DISPID DISPID_USER_EXECUTE = DISPID_VALUE + 1;
static const DISPID DISPID_USER_WRITEFILE = DISPID_VALUE + 2;
static const DISPID DISPID_USER_READFILE = DISPID_VALUE + 3;
// ...
};
EDIT 3, 4 and 5:
Moved to a separate question
EDIT 6:
Made a separate question out of the "returning a string" edits. That way I can accept Georg's reply as that answers the original question.
EDIT 7:
I have gotten a few requests for a fully working, self contained, example implementation. Here it is: https://github.com/Tobbe/CppIEEmbed. Please fork and improve if you can :)