Convert VARIANT to...? - c++

Note:
Attempting to invoke a method of an interface, of which the return type is _variant_t
Code:
_variant_t resultsDataString;
_bstr_t simObjectNames;
simObjectNames = SysAllocString (L"TEST example 3");
resultsDataString = pis8->GetSimObject (simObjectNames);
inline function illustrated below, contained in .tli FILE:
inline _variant_t IS8Simulation::GetSimObject ( _bstr_t Name ) {
VARIANT _result;
VariantInit(&_result);
HRESULT _hr = get_SimObject(Name, &_result);
if (FAILED(_hr)) _com_issue_errorex(_hr, this, __uuidof(this));
return _variant_t(_result, false);
}
Note:
resultsDataString is of struct tagVARIANT:
VARTYPE vt is 9 (unsigned short)
IDispatch IDispatch interface pointer
Question
How can I then convert or extract the value?
Possible using VariantChangeType?
Another way?
Edit:
Note:
Looking to transform the following, Visual Basic code to Visual C++
MFC or ATL, if need be
Ideally, pure C++
Visual basic equivalent:
Public example1, example2 As SIMUL8.S8SimObject
Dim numberOfexamples As Variant
Dim resultString As Variant
Set example1 = MySimul8.SimObject("Example 1")
Set example2 = MySimul8.SimObject("Example 2")
numberOfexamples = example1.CountContents + example2.CountContents
resultString = CStr(numberOfexamples) & "*"

It appears you are using C++ as a COM client by relying on the VC++ compiler's built-in COM support. To make coding the client "easier" you've used #import to generate C++ wrapper classes that attempt to hide all the COM details from you - or at least make the COM details simpler. So you're not using the COM SDK directly, but are using a client-side framework (I think of it like a light-weight COM-only framework akin to ATL or MFC).
Your example code, however, seems to be mixing the direct low-level COM SDK (VARIANTs, BSTR, SysAllocString) with the #import COM framework (_variant_t, _bstr_t, XXXXPtr). COM from C++ is complicated at first - so in a perfect world I would suggest getting to know the basics of COM before going too far forward.
However, if you just want something to work I would guess this is #import-style-of-COM-clients version of the VB code you provided:
_variant_t example1Var;
_variant_t example1Var;
SIMUL8::S8SimObjectQIPtr example1; // I'm guessing at this type-name from the VB code
SIMUL8::S8SimObjectQIPtr example2;
example1Var = pis8->GetSimObject(_bstr_t(L"Example 1"));
example2Var = pis8->GetSimObject(_bstr_t(L"Example 2"));
if (example1Var.vt == VT_DISPATCH && example2Var.vt == VT_DISPATCH)
{
// **UPDATE** to try to spoon feed the QI ptr...
example1 = IDispatchPtr((IDispatch*)example1Var);
example2 = IDispatchPtr((IDispatch*)example2Var);
// Does this screw-up reference counting?
int numberOfexamples = example1->CountContents + example2->CountContents;
}
UPDATE:
Documentation on #import This makes using COM from C++ much easier, but is yet one other thing to learn...

Take a look at the MSDN docs for _variant_t
There is a ChangeType method as well as Extractors.
Edit: Once you have an IDispatch pointer, you need to use QueryInterface to get a specific object type with members and methods, or use IDispatch::Invoke. You can look at the MSDN Docs for IDispatch. Either way, you need to know about the object type that is being returned by the IS8Simulation::GetSimObject call.
Edit #2: Based on your VB code update, you want to use the same kind of code for C++, i.e. use the S8SimObject type in its C++ form (look in the generated .tlh file for _COM_SMARTPTR_TYPEDEF). Then you can directly do the same thing with CountContents. Otherwise, you would need to look up the CountContents DISPID with IDispatch::GetIDsOfNames and then call invoke.

You don't have to convert the _variant_t. It contains an IDispatch interface pointer, so you can get it out like this:
if (v.vt == VT_DISPATCH)
{
IDispatchPtr pDispatch = v.pdispVal;
// Do something with pDispatch.
// e.g. assign it to a strongly-typed interface pointer.
}

There is a concept of "value" for an object (DISPID_VALUE).
Here is a codeproject.com article on resolving VARIANTs that might help.

Related

Why do my C# and C++ dlls exhibit different behavior?

I am working on a project that involves the creation of a dll that honours a certain interface in order to plug into some software in order to add functionality to it. This is done by a dll that calls my dll (I do not have the source code for the dll that does the calling). Originally I was given an interface and a C# implementation that created a COM visible dll. However after using this for a while I found I wanted to make use of some large C++ libraries and as creating wrappers would take a long time I thought about creating a C++ ATL COM dll instead. I did this and the methods of my class appear to be called correctly (I register my dll, run the program and the methods appear to be called in the correct order), however I have found some of the behavior to be different.
I am not sure how to go about explaining this as my code relates to a closed source API but perhaps if I describe an example someone might have some ideas as to where I might want look.
For instance, in the C# dll I attempt to open a file by doing this:
FMANFileControl fileControl = new FMANFileControl();
FMANFile wFile = null;
const string filePath = #"C:\Data\April 4\Data_IDA.wiff";
wFile = fileControl.GetFileObject(filePath, 1);
long numSamples = wFile.GetNumberOfSamples();
I get the correct number of samples.
In my C++ dll I have this (with some of the HRESULT checks removed in order to keep the code shorter):
std::string filePath = "C:\\Data\\April 4\\Data_IDA.wiff";
_bstr_t fileName(filePath.c_str());
IFMANFilePtr ipFMANFile;
IFMANFileControlPtr ipFMANFileControl;
hr = ipFMANFileControl.CreateInstance(__uuidof(FMANFileControl));
hr = ipFMANFile.CreateInstance(__uuidof(FMANFile));
ipFMANFile = ipFMANFileControl->GetFileObject(fileName, 1);
long numSamples = ipFMANFile->GetNumberOfSamples();
but the files does not open correctly, resulting is zero samples.
Using oleview I looked at the typelib and it says this for the function:
[id(0x00000001), helpstring("method GetWiffFileObject")]
IFMANWiffFile* GetWiffFileObject( [in] BSTR WiffFileName, [in] long sample);
The file I get information from is one that is being written to during an experiment and just before it obtains more data it calls my method and I should be able to obtain the newest file. In the C# dll this is possible, but in the C++ dll this is not. While I realize the specifics of this is hidden, I am wondering is anyone has any idea why a C++ COM dll and a C#, comvisible dll that make use of the same interface would exhibit different behavour when being called by the same dll.
I am pretty stumped at this moment so any ideas at all would be appreciated, even if they turn out to be way off base. I can share my source code if anyone thinks they might be able to help.
EDIT:
I tried the solution to answer 1, however I could not compile my code. When reading about this I found this post:
Differences between [in, out] and [out, retval] in COM IDL definitions
that seems to suggest that since the FMANFile pointer is marked [out, retval] that the method becomes:
IFMANFilePtr ExploreData::IFMANFileControl(BSTR filename, long sample);
or am I misinterpreting that article?
EDIT 2:
Got it working though I am not really sure why.
Originally I had the variables declared in the header as private member variables of the class, like this:
class ATL_NO_VTABLE CUserIDA :
public CComObjectRootEx<CComSingleThreadModel>,
public CComCoClass<CUserIDA, &CLSID_UserIDAObject>,
public IUserIDA
{
.
.
.
public:
STDMETHOD(GetSwitchCriteria)(DOUBLE* intensity, DOUBLE* minMass, DOUBLE* maxMass, VARIANT_BOOL *selectIntensity, LONG* numOfDepCycles);
.
.
.
private:
ExploreDataObjects::IFMANWiffFilePtr ipFMANWiffFile;
ExploreDataObjects::IFMANWiffFile2Ptr ipFMANWiffFile2;
};
Just to try it I moved them to the top of the class delcaration like this:
class ATL_NO_VTABLE CUserIDA :
public CComObjectRootEx<CComSingleThreadModel>,
public CComCoClass<CUserIDA, &CLSID_UserIDAObject>,
public IUserIDA
{
ExploreDataObjects::IFMANWiffFilePtr ipFMANWiffFile;
ExploreDataObjects::IFMANWiffFile2Ptr ipFMANWiffFile2;
I thought that by default these would also be private members and the same as before so I am at a loss to explain why this seemed to work. Can someone explain this?
Your C++ code is correct, except for the following line:
hr = ipFMANFile.CreateInstance(__uuidof(FMANFile));
It doesn't make any sens, because ipFMANFile is initialized once again in the next statement.
Unfortunately, this IDL declaration:
IFMANWiffFile* GetWiffFileObject([in] BSTR WiffFileName, [in] long sample);
is limited for debugging purposes since it doesn't support the native COM mechanisme for the exception reporting via HRESULT. The COM compliant declaration would be:
HRESULT GetWiffFileObject([in] BSTR WiffFileName, [in] long sample, [out, retval] IFMANWiffFile** fileInstance);
I believe that you are unable to change the library's code so I suggest you to try some external debugging tools like 'procmon.exe' and 'dbgview.exe' to inspect the application events when you run the CPP test case. Look for all failed actions.
I hope this will help you somehow

Given a COM DLL, extract all classes CLSID and corresponding interface name

My question is similar to Getting CLSID for a DLL file?, I think.
I have a directory with some DLLs, each one implementing one or more COM interfaces. I would like to get:
1) Each interface name
2) The CLSID of the class implementing the interface
For each DLL. It's important that everything can be done programatically (So I can't use some sort of COM browser and manually look up for that information).
Later I will lookup the CLSID given the interface name and call some methods using IDispatch.
One alternative seems to be scanning the registry trying to match the type, interface and class GUID and the .dll filename. But that seems to be slow and not robust.
Does someone has a clear solution to this problem?
EDIT:
With the response of Ben Voigt, I came with the following code which suit my needs:
ITypeLib *typelib;
ITypeInfo *typeinfo;
LoadTypeLibEx(_T("c:\\mydir\\mycom1"), REGKIND_NONE, &typelib);
for (UINT i = 0;i < typelib->GetTypeInfoCount();++i) {
TYPEKIND typekind;
typelib->GetTypeInfoType(i, &typekind);
if (typekind == TKIND_COCLASS) {
// class!
CComBSTR className;
TYPEATTR *typeattr;
typelib->GetTypeInfo(i, &typeinfo);
typeinfo->GetDocumentation(MEMBERID_NIL, &className, NULL, NULL, NULL);
typeinfo->GetTypeAttr(&typeattr);
GUID classGUID = typeattr->guid;
for (UINT j = 0;j < typeattr->cImplTypes;++j) {
// interface!
CComBSTR interfaceName;
HREFTYPE hreftype;
ITypeInfo *classtypeinfo;
typeinfo->GetRefTypeOfImplType(j, &hreftype);
typeinfo->GetRefTypeInfo(hreftype, &classtypeinfo);
classtypeinfo->GetDocumentation(MEMBERID_NIL, &interfaceName, NULL, NULL, NULL);
// associate interfaceName with classGUID here
}
}
}
You can't get that from the COM DLL, but you can get it from the typelib. I'm pretty sure the MIDL compiler has a switch to decompile a typelib, but parsing IDL wouldn't be as easy as using the TypeLib API.
To complicate matters, the typelib is often stored as a resource inside the DLL. So you'd extract the resource, and open it with the TypeLib API.
Start with LoadTypeLibEx which will return you an ITypeLib* interface pointer (you knew you were going to need COM in order to get information about COM libraries, right?). This will actually do the resource extraction step for you.
Then, call ITypeLib::GetTypeInfoCount to find out how many types there are. Call ITypeLib::GetTypeInfoType for each one to find the interfaces and coclasses. And call ITypeLib::GetTypeInfo followed by ITypeInfo::GetDocumentation to get the name.
It looks like you have all of this so far. Next you need the GUID of the type, which is gotten with ITypeInfo::GetTypeAttr (not ITypeLib::GetLibAttr). That gives you a TYPEATTR structure, which has a guid field.
From the same TYPEATTR structure, you'll need the cImplTypes field. That together with ITypeInfo::GetRefTypeOfImplType will let you match up each coclass to the interfaces it implements.
Note that there's not guaranteed to be a 1:1 relationship between interfaces and implementation coclasses. And the interface can be in a different library from the coclass.
Few caveats to Ben Voigt's answer: not every COM DLL has a typelib. In your case, it seems, it does; but that's not a requirement. The only rock-solid requirement is that the DLL exports the function DllGetClassObject(), where you pass a CLSID and get back an object factory.
You could load the library and call DllGetClassObject for every registered CLSID on the system (scan the registry under HKCR\CLSID for the list of those). The ones where you get a valid object back are the ones that the DLL supports. Now, in theory, it's not even a requirement that the CLSIDs the DLL supports are registered; I can envision a DLL that implements private object classes that only the intended client knows about and no one else should. But this is a very, very exotic scenario; for one thing, the regular COM mechanism of looking up the DLL path by the CLSID will break for those. The client would have to load the DLL directly.
You could also scan the registry for the CLSID's where the DLL under consideration is registered as InprocServer32; this, again, will break in case of private classes. You can do a registry search in regedit, search in data. Also, you'd have to deal with filename case issues, short vs. long names, hardlinks, environment variable substitution and so on. So I would not recommend it.
EDIT: just thought of another way. Download Regmon, run it, and call regsvr32 on the DLL. Then watch what CLSID's are touched.
You might want to look at the source code in WiX's heat utility, which does the same thing:
http://wix.sourceforge.net/manual-wix3/heat.htm

Writing a simple ActiveX control for IE that has one method

I'm learning how to write a scriptable ActiveX control. My goal is to have a tiny control that can check to see if something is installed on the system. What I've done so far is:
Create a MFC ActiveX control project in VS2008
Add some 'safe for scripting' bits that I found here.
Extend the IDL to provide my "IsInstalled" method, which for now returns TRUE unconditionally (but will later read some keys from the registry.)
Build the control and run regsvr32 on it. I verified that it does show up in HKEY_CLASSES_ROOT, and when I instantiate the object, the IE Developer Tools "Locals" pane shows that the object is of type _D[my plugin name]. Not only that, but my IsInstalled() method shows up underneath that object.
However, when I call IsInstalled(), I just can't get it to work:
JScript Debugger - Breaking on JScript runtime error -(n http://img138.imageshack.us/img138/1586/whycomwhy.png
I'm at a loss. I've also tried making IsInstalled a property instead of a method, using VARIANT_BOOL instead of boolean instead of BOOL in the IDL, you name it.
Here's the relevant excerpts of code.
The header:
afx_msg VARIANT_BOOL IsInstalled();
The implementation:
afx_msg VARIANT_BOOL
CMyAXCtrl::IsInstalled()
{
return TRUE;
}
The dispatch map:
BEGIN_DISPATCH_MAP(CMyAXCtrl, COleControl)
DISP_FUNCTION_ID(CMyAXCtrl, "IsInstalled", dispidIsInstalled, IsInstalled, VT_BOOL, VTS_NONE)
END_DISPATCH_MAP()
The dispatch part of the IDL:
[ uuid(6B662202-CF13-4144-AA33-C3FEE9C2C962),
helpstring("Dispatch interface for My Control")]
dispinterface _Daxplugin
{
properties:
methods:
[id(1)] VARIANT_BOOL IsInstalled();
};
If there's any other relevant bits of code I should provide, let me know. But I'm stumped here. Thank you in advance!
You almost certainly have the wrong prototype for a scriptable function. OLE Automation for scripting languages tends to rely on returning a HRESULT then using an out parameter for the actual return code.
So change it to
[id(1)] HRESULT IsInstalled(VARIANT_BOOL* p);
Also TRUE != VARIANT_TRUE, you must return VARIANT_TRUE which is equal to -1 instead of 1.
Hope some of that actually helps, but without the actual error I might be mistaken :)
You could mark your control as save for scripting by implementing IObjectSafety or by marking the Object as save while registering it (as supposed by the link you provided).
Have you run regsvr32 after adding the code to mark it save for scripting?
You can check the registry if your control has the safe for scripting bits set.
If the bits are set you will find the two keys {7DD95802-9882-11CF-9FA9-00AA006C42C4} (Safe for Initialization)
{7DD95801-9882-11CF-9FA9-00AA006C42C4}(Safe For Scripting) as subkeys of ImplementedCategories in your object.
I would suggest to implement IObjectSafety since it doesn't depend on your class to register itself.

How to conditionally choose the C# class I invoke via COM in my C++ DLL?

After much help from all my StackOverFlow brethren, I managed to create a C++ DLL that calls my C# classes via COM and passes data back and forth to an external application. There was much celebration in the kingdom after that code started working.
Now I have a new problem. I'm expanding the DLL so that it can call different classes (all implementing the same interface). I need to decide what class to call depending on a char array passed to the DLL when it is loaded. The methods I call are exactly the same regardless of which class I use. What is the best way to switch between classes when calling the DLL?
// Initialize COM.
HRESULT hr = CoInitialize(NULL);
// I want to do something like this....but how?
if (strcmp(modelType, "Model1") == 0) {
IUnitModelPtr pIUnit(__uuidof(ClassOne));
}
if (strcmp(modelType, "Model2") == 0) {
IUnitModelPtr pIUnit(__uuidof(ClassTwo));
}
//call method 1
//call method 2
CoUninitialize();
//exit
This is probably a fairly simple question, but I really do not know any C++. Just getting COM going was a major challenge for me.
edit: Im sure there are some super elegant ways to achieve this (reflection?) but please limit your suggestions to stuff that can be implemented easily....efficiency is not important here and maintainability is not really an issue.
Do smth like this:
GUID classId = GUID_NULL;
if( strcmp( modelType, "Model1" ) == 0 ) {
classId = __uuidof( class1 );
} else if( strcmp( modelType, "Model2" ) == 0 ) {
classId = __uuidof( class2 );
} else if(... etc, continue for all possible model types
}
IUnitModelPtr unit;
unit.CreateInstance( classId );
// interface methods can be called here
Two options.
As each different interface implementation is a coclass, and your C++ code uses the right prog-id/classid to perform the create.
Alternately the exposed coclass is a factory with a method to return the actual implementation class.
Both need logic to map your modelType to implementation class. You can do this in the C++ code or in the .NET code. Without more information about the overall context (is the string --> coclass mapping logically part of the COM component or of the caller).
(In a pure C++ COM implementation you can create your own custom COM instance factory, but that's more advanced COM and not available for .NET COM interop.)

How to use msxml with Visual Studio 2008 Express (no ATL classes) without becoming crazy?

It is not really a question because I have already found a solution. It took me a lot of time, that's why I want to explain it here.
Msxml is based on COM so it is not really easy to use in C++ even when you have helpful classes to deal with memory allocation issues. But writing a new XML parser would be much more difficult so I wanted to use msxml.
The problem:
I was able to find enough examples on the internet to use msxml with the help of CComPtr (smart pointer to avoid having to call Release() for each IXMLDOMNode manually), CComBSTR (to convert C++ strings to the COM format for strings) and CComVariant. This 3 helpful classes are ATL classes and need an #include <atlbase.h>.
Problem: Visual Studio 2008 Express (the free version) doesn't include ATL.
Solution:
Use comutil.h and comdef.h, which include some simple helper classes:
_bstr_t replaces more or less CComBSTR
_variant_t replaces more or less CComVariant
_com_ptr_t replaces indirectly CComPtr through the use of _COM_SMARTPTR_TYPEDEF
Small example:
#include <msxml.h>
#include <comdef.h>
#include <comutil.h>
// Define some smart pointers for MSXML
_COM_SMARTPTR_TYPEDEF(IXMLDOMDocument, __uuidof(IXMLDOMDocument)); // IXMLDOMDocumentPtr
_COM_SMARTPTR_TYPEDEF(IXMLDOMElement, __uuidof(IXMLDOMElement)); // IXMLDOMElementPtr
_COM_SMARTPTR_TYPEDEF(IXMLDOMNodeList, __uuidof(IXMLDOMNodeList)); // IXMLDOMNodeListPtr
_COM_SMARTPTR_TYPEDEF(IXMLDOMNamedNodeMap, __uuidof(IXMLDOMNamedNodeMap)); // IXMLDOMNamedNodeMapPtr
_COM_SMARTPTR_TYPEDEF(IXMLDOMNode, __uuidof(IXMLDOMNode)); // IXMLDOMNodePtr
void test_msxml()
{
// This program will use COM
CoInitializeEx(NULL, COINIT_MULTITHREADED);
{
// Create parser
IXMLDOMDocumentPtr pXMLDoc;
HRESULT hr = CoCreateInstance(__uuidof (DOMDocument), NULL, CLSCTX_INPROC_SERVER, IID_IXMLDOMDocument, (void**)&pXMLDoc);
pXMLDoc->put_validateOnParse(VARIANT_FALSE);
pXMLDoc->put_resolveExternals(VARIANT_FALSE);
pXMLDoc->put_preserveWhiteSpace(VARIANT_FALSE);
// Open file
VARIANT_BOOL bLoadOk;
std::wstring sfilename = L"testfile.xml";
hr = pXMLDoc->load(_variant_t(sfilename.c_str()), &bLoadOk);
// Search for node <testtag>
IXMLDOMNodePtr pNode;
hr = pXMLDoc->selectSingleNode(_bstr_t(L"testtag"), &pNode);
// Read something
_bstr_t bstrText;
hr = pNode->get_text(bstrText.GetAddress());
std::string sSomething = bstrText;
}
// I'm finished with COM
// (Don't call before all IXMLDOMNodePtr are out of scope)
CoUninitialize();
}
Maybe try using the #import statement.
I've used it in a VS6 project I have hanging around, you do something like this (for illustrative purposes only; this worked for me but I don't claim to be error proof):
#import "msxml6.dll"
...
MSXML2::IXMLDOMDocument2Ptr pdoc;
HRESULT hr = pdoc.CreateInstance(__uuidof(MSXML2::DOMDocument60));
if (!SUCCEEDED(hr)) return hr;
MSXML2::IXMLDOMDocument2Ptr pschema;
HRESULT hr = pschema.CreateInstance(__uuidof(MSXML2::DOMDocument60));
if (!SUCCEEDED(hr)) return hr;
pschema->async=VARIANT_FALSE;
VARIANT_BOOL b;
b = pschema->loadXML(_bstr_t( /* your schema XML here */ ));
MSXML2::IXMLDOMSchemaCollection2Ptr pSchemaCache;
hr = pSchemaCache.CreateInstance(__uuidof(MSXML2::XMLSchemaCache60));
if (!SUCCEEDED(hr)) return hr;
_variant_t vp=pschema.GetInterfacePtr();
pSchemaCache->add(_bstr_t( /* your namespace here */ ),vp);
pdoc->async=VARIANT_FALSE;
pdoc->schemas = pSchemaCache.GetInterfacePtr();
pdoc->validateOnParse=VARIANT_TRUE;
if (how == e_filename)
b = pdoc->load(v);
else
b = pdoc->loadXML(bxmldoc);
pXMLError = pdoc->parseError;
if (pXMLError->errorCode != 0)
return E_FAIL; // an unhelpful return code, sigh....
Another option would to use another XML parser that is already done, such as eXpat. It avoids having to use ATL and the complexities of COM, and is way easier than implementing your own. I suggest this only becasue you've stated that the reason you're looking at msxml is because you don't want to implement your own parser.
You can use TinyXML. It is open source and more over platform independent.
I'm happy I posted my question although I already had a solution because I got several alternative solutions. Thanks for all your answers.
Using another parser such as eXpat or the maybe smaller (not so powerfull but enough for my needs) TinyXML could actually be a good idea (and make it easier to port the program to another operating system).
Using an #import directive, apparently a Microsoft specific extension to simplify the use of COM, is also interesting and brought me to the following web page MSXML in C++ but as elegant as in C#, which explain how to simplify the use of msxml as much as possible.
Why don't you use some MSXML wrapper that would shield you form COM, such as Arabica?