问题
I am new to C# but have worked extensively with C++. I have a C++ function that needs to be called from C#. After reading some answers from SO and some googling, I conclude that I need to make a pure C interface to the function. I have done this, but am still confused about how to call it from C#.
The function in C++ looks like this:
int processImages(
std::string& inputFilePath, // An input file
const std::vector<std::string>& inputListOfDirs, // Input list of dirs
std::vector<InternalStruct>& vecInternalStruct, // Input/Output struct
std::vector<std::vector< int > >& OutputIntsForEachFile,
std::vector< std::vector<SmallStruct> >& vecVecSmallStruct, // Output
int verboseLevel
);
The same function, converted in C, looks like this:
int processImagesC(
char* p_inputFilePath, // An input file
char** p_inputListOfDirs, // Input list of dirs
size_t* p_numInputDirs, // Indicating number of elements
InternalStruct* p_vecInternalStruct, // Input/Output struct
size_t* p_numInternalStructs,
int** p_OutputIntsForEachFile, // a 2d array each row ending with -1
size_t* p_numOutputIntsForEachFile //one number indicating its number of rows
SmallStruct** p_vecVecSmallStruct, // Output
size_t* p_numInVecSmallStruct,
int verboseLevel
);
This is based on this advice.
Now I need to call this from C#, which is where the confusion is. I have tried my best to convert the structures.
The C# code looks like this:
[DllImport(
@"C:\path\to\cppdll.dll", CallingConvention=CallingConvention.Cdecl,
EntryPoint="processImagesC", SetLastError=true)]
[return: MarshalAs(UnmanagedType.I4)]
unsafe public static extern int processImagesC(
String inputFilePath,
String[] inputListOfDirs,
ref uint numInputListOfDirs,
// Should I use ref InternalStruct * vecInternalStruct?
ref InternalStruct[] vecInternalStruct,
ref uint numInternalStruct,
// Or ref int[] or ref int[][] or int[][]?
ref int[][] OutputIntsForEachFile,
ref uint numOutputIntsForEachFile,
// again, ref ..[], [][], or ref [][]?
ref SmallStruct[][] vecVecSmallStruct,
int verboseLevel
);
There are memory allocations for all the output variables (pointers) done within the C/C++ code. This likely means we need to declare the code as unsafe, correct?
How do we handle memory deallocation? Should I write another API (function) that does the deallocation of objects/arrays allocated by C/C++?
The C++ code needs to be standard compliant and platform independent, so I cannot insert any windows-specific things in it.
I hope someone could make sense of this and provide an answer or at least point me in the right direction.
回答1:
Since there seems to be some interest in using It Just Works (IJW) with C++/CLI, I'll post some info about that, further google searches and research will need to be done to figure it all. C++/CLI can be enabled with a single compiler flag (/CLI, enabled through Property Page->General->Common Language Runtime Support). C++/cli is NOT c++, but rather just another managed language. C++/CLI classes can be compiled into dll's and called directly from other .NET projects (C#, VB.NET, ect). However, unlike the other .NET languages it can directly interact with C++ code.
This is an ok start to learning C++/CLI. The big thing to learn is the decorations that tell you the class is managed (.NET class) and not Vanila C++. The "ref" keyword decalres the definition as a .NET definition:
public ref class Foo{ public: void bar();};//Managed class, visible to C#
public ref struct Foo{};//Managed struct, visible to C#
All reference classes are referred to with Handles rather than pointers or references. A handle is denoted by the ^ operator. To make a new handle, you use gcnew, and to access functions/members of the handle, use the -> operator.
//in main
Foo^ a = gcnew Foo();
a->bar();
You often have to move structures common from C# to native types and back again. (such as managed Array^ or String^ to void* or std::string). This process is called Marshaling. This handy table is pretty useful for figuring that out.
A common task is to create a wrapper for a native class, done like this:
//Foo.h
#include <string>
namespace nativeFoo
{
class Foo
{
private:
std::string fooHeader;
public:
Foo() {fooHeader = "asdf";}
std::string Bar(std::string& b) {return fooHeader+b;}
};
}
//ManagedFoo.h
#include "foo.h"
namespace managedFoo
{
public ref class Foo
{
private:
nativeFoo::Foo* _ptr;
public:
Foo(){_ptr = new nativeFoo::Foo();}
~Foo(){!Foo();}
!Foo(){if (_ptr){delete ptr;ptr = NULL;}}
String^ bar(String^ b)
{
return marshal_as<String^>(_ptr->bar(marshal_as<std::string>(b)));
}
};
}
Warning: I am totally missing a bunch of #include and #using statements, this is just to give a general gist of how to use this.
回答2:
Begin from this:
- How to pass structure as pointer in C dll from C#
And something about marshalling in this:
- Deep copying an array c# without serialization
Note that Marshal.Copy also overloads for arrays use. With marshalling you can get rid of ref excepting that you do want to. Just write C/C++ in their way.
And following is a little bit complicated:
- Physical disk size not correct (IoCtlDiskGetDriveGeometry)
回答3:
The 2 ways I've often seen this handled is to either have a 'FreeResource' style API, or specify in the function the size of output buffers.
Method 1
C++
void GetName(char ** _str)
{
if (!_str)
return; // error
*_str = new char[20];
strcpy(*str, "my name");
}
void FreeString(char * _str)
{
delete str;
}
Client (any language)
char * name;
GetName(&name);
...
FreeString(name);
Method 2
C++
void GetName(char * _str, size_t _len)
{
if (_len < 20)
return; // error
strcpy(str, "my name");
}
Client (any language)
char * name = new char[20];
GetName(name, 20);
...
回答4:
If you are willing to use third party tool, there is a tool named C#/.NET PInvoke Interop SDK may be helpful to you. But you can do it yourself as well. For simple classes with a few methods, you can write your own code in managed C# code.
The basic idea of instantiating a C++ object from .NET world is to allocate exact size of the C++ object from .NET, then call the constructor which is exported from the C++ DLL to initialize the object, then you will be able to call any of the functions to access that C++ object, if any of the method involves other C++ classes, you will need to wrap them in a C# class as well, for methods with primitive types, you can simply P/Invoke them. If you have only a few methods to call, it would be simple, manual coding won't take long. When you are done with the C++ object, you call the destructor method of the C++ object, which is a export function as well. if it does not have one, then you just need to free your memory from .NET.
Here is an example.
public class SampleClass : IDisposable
{
[DllImport("YourDll.dll", EntryPoint="ConstructorOfYourClass", CharSet=CharSet.Ansi, CallingConvention=CallingConvention.ThisCall)]
public extern static void SampleClassConstructor(IntPtr thisObject);
[DllImport("YourDll.dll", EntryPoint="DoSomething", CharSet=CharSet.Ansi, CallingConvention=CallingConvention.ThisCall)]
public extern static void DoSomething(IntPtr thisObject);
[DllImport("YourDll.dll", EntryPoint="DoSomethingElse", CharSet=CharSet.Ansi, CallingConvention=CallingConvention.ThisCall)]
public extern static void DoSomething(IntPtr thisObject, int x);
IntPtr ptr;
public SampleClass(int sizeOfYourCppClass)
{
this.ptr = Marshal.AllocHGlobal(sizeOfYourCppClass);
SampleClassConstructor(this.ptr);
}
public void DoSomething()
{
DoSomething(this.ptr);
}
public void DoSomethingElse(int x)
{
DoSomethingElse(this.ptr, x);
}
public void Dispose()
{
Marshal.FreeHGlobal(this.ptr);
}
}
For the detail, please see the below link,
C#/.NET PInvoke Interop SDK
The tool, xInterop NGen++ 2.0 has been released. Please check it out if you are interested in creating C# wrapper for native C++ DLL.
(I am the author of the SDK tool)
来源：https://stackoverflow.com/questions/15672351/calling-c-function-from-c-with-lots-of-complicated-input-and-output-paramete