I’ll be speaking at DConf this year. I haven’t yet decided how to structure the talk yet, but I’d like to take a page out of Dan Sak’s excellent CppCon 2016 talk that covered technical and psychological aspects of language adoption. In his case he was addressing a C++ audience on the difficulty of convincing C programmers to adopt C++ (an admirable goal!). In mine, it’s how do I out-C++ C++?
The genesis for my talk was my decision a few years ago to write tests for a legacy C codebase. I wanted to use D, but even as a fan of the language I chose to write the tests in C++ instead. And if even I wouldn’t have picked D for that task, who would?
I still think it was the right decision to make. The first step in”importing” the legacy project into a C++ test looks like this:
extern "C" {
#include "legacy.h"
}
There is no second step. From here one #includes a test library such as catch2 and starts writing tests for the C functions in the API.
As far as I know, every language under the sun can interface with C. There’s usually (always?) some way to do FFI, and examples invariably show how to call a function that takes 2 integers, maybe even a const char* parameter. Unfortunately, that’s not what real codebases need to do. They need to call a function that takes a pointer to a structure that’s defined in a different header, which has 3 pointers to structures that are themselves defined in different headers, which…
You get the idea. Writing the definitions by hand error-prone, boring, and time-consuming. Even if correct, nobody seems to talk about the elephant in the C API room: macros. Not once have I encountered a non-trivial C API that doesn’t require the C preprocessor to use properly. C++ users are at an advantage here: just use the macros. Everyone else has to get by with ad-hoc solutions. And if the function-like macro implementations change, the code will break.
I think that the only way to make it so that C++ isn’t the obvious and/or only choice is that a challenger allows one to write `#include “legacy.h”` and have it work just as easily as it does natively. That’s why I started the dpp project, and I’m looking forward to talking about the technical challenges I’ve encountered on my journey so far.
See you at DConf!
Átila on Code 