To make this current approach work you'd need to pass StringBuilder
instances rather than string
. That's because the data is flowing from caller to callee. The strings are out parameters. And that means that the caller has to allocate the buffers for each string, and know how large the buffers need to be.
It's much easier to use BSTR
here. This allows you to allocate the strings in the native code, and have them deallocated in the managed code. That's because BSTR
is allocated on the shared COM heap, and the p/invoke marshaller understands them. Making this slight change means that the caller does not need to know how large the strings are up front.
The code would look like this:
SAMPLEDLL_API BOOL InitExecution(BSTR* names, int count)
{
for (int i = 0 ; i < count; i++)
names[i] = SysAllocStr(...);
return TRUE;
}
And on the C# side you write it like this:
[DllImport(@"mydll.dll", CallingConvention = CallingConvention.Cdecl)]
public static extern bool InitExecution(
[Out] IntPtr[] names,
int count
);
And then you've got a bit of work to marshal from BSTR
to C# string
.
IntPtr[] namePtrs = new IntPtr[count];
InitExecution(namePtrs, namePtrs.Length);
string[] names = new string[namePtrs.Length];
for (int i = 0; i < namePtrs.Length; i++)
{
names[i] = Marshal.PtrToStringBSTR(namePtrs[i]);
Marshal.FreeBSTR(namePtrs[i]);
}