A very convenient tecnique in c++ programming is the one known with many names: "d-pointer" (found in Qt/KDE contexts), shadow pointer, "pimpl", opaque pointer. Basically the idea is to hide all the private details of a class in a forward declared private pointer, which will be the only real private member of the class. Since this member will never change, this will guarantee binary compatibility among different versions of a library.

But there are other advantages in using d-pointers: one is compilation speedup during development (usually if you change a private member of a class, the header changes, and you have to recomplile all units using that header, while with d-pointer you change only the .cpp file), and another is code cleanup: you will have very compact header files, describing your class public interface without private stuff pollution. Also, sometimes you may want to add your d-pointer definition in a separate header file, ending up with three well defined files:

  • myclass.h: class declaration, only public stuff, without private stuff
  • myclass_p.hpp: private class declaration: only private stuff, no implementations
  • myclass.cpp: only implementations

The classical approach is to create a plain raw pointer to a forward declared class, initialize it in the constructor, and delete it on the destructor. A nice addition is to have the private class be a nested type, so that you can avoid polluting your IDE class list.

When programming in C++ it can often happen to be using C-style API.
These usually come in the form:

int some_api_call(char *inputParameter, char **outputParameter);

where the return value is never a real output value, but instead an exit code, and usually 0 means success.

To handle such API in a sequence of operations, one is then usually blinded to do something like this:

 int result = first_c_api_call();
 if(result != 0) {
 cerr << "Error executing first_c_api_call: " << result << endl;
 return;
 }

result = second_c_api_call();
if(result != 0) {
cerr << "Error executing second_c_api_call: " << result << endl;
return;
}

result = third_c_api_call();
.....

and so on, which is kinda boring when you have to call lots of API functions in one method.

I have been trying to write some kind of wrapper that can help making this a bit easier.
In a real life example, I’ve been trying to use gphoto2 api in a c++11 application.
Using c++11 lambdas and RAII this is what I’ve been able to do:

 void GPhotoCamera::connect() {
      CameraAbilities abilities;
      GPPortInfo portInfo;
      CameraAbilitiesList *abilities_list = nullptr;
      GPPortInfoList *portInfoList = nullptr;
      CameraText camera_summary;
      CameraText camera_about;
      int model, port;
      gp_api{ {
        sequence_run( [&]{ return gp_abilities_list_new (&abilities_list); } ),
        sequence_run( [&]{ return gp_abilities_list_load(abilities_list, d->context); } ),
        sequence_run( [&]{ model = gp_abilities_list_lookup_model(abilities_list, d->model.toLocal8Bit()); return model; } ),
        sequence_run( [&]{ return gp_abilities_list_get_abilities(abilities_list, model, &abilities); } ),
        sequence_run( [&]{ return gp_camera_set_abilities(d->camera, abilities); } ),
        sequence_run( [&]{ return gp_port_info_list_new(&portInfoList); } ),
        sequence_run( [&]{ return gp_port_info_list_load(portInfoList); } ),
        sequence_run( [&]{ return gp_port_info_list_count(portInfoList); } ),
        sequence_run( [&]{ port = gp_port_info_list_lookup_path(portInfoList, d->port.c_str()); return port; } ),
        sequence_run( [&]{ return gp_port_info_list_get_info(portInfoList, port, &portInfo); return port; } ),
        sequence_run( [&]{ return gp_camera_set_port_info(d->camera, portInfo); } ),
        sequence_run( [&]{ return gp_camera_get_summary(d->camera, &camera_summary, d->context); } ),
        sequence_run( [&]{ return gp_camera_get_about(d->camera, &camera_about, d->context); } ),
      }, make_shared<QMutexLocker>(&d->mutex)}
      .on_error([=](int errorCode, const std::string &label) {
        qDebug() << "on " << label << ": " << gphoto_error(errorCode);
        emit error(this, gphoto_error(errorCode));
      }).run_last([&]{
        d->summary = QString(camera_summary.text);
        d->about = QString(camera_about.text);
        emit connected();    
      });  
      // TODO d->reloadSettings();
      gp_port_info_list_free(portInfoList);
      gp_abilities_list_free(abilities_list);
}

I can then declare some variables in the first part of the method, and inside the “gp_api” block i can execute a sequence of operation, each one returning an int value. This value is automatically checked for an error, and if it it’s a success exit code, the next sequence block is executed.
run_last is finally executed if all steps are completed successfully. An optional mutex locker (QMutexLocker) is passed to the gp_api block as the last constructor argument, to automatically lock the c api for multithreading.

How have I accomplished this?

This is the main class so far:

#include <functional>
#include <list>
#include <mutex>

typedef std::shared_ptr<std::unique_lock<std::mutex>> default_lock;
template<typename T, T defaultValue, typename check_operator = std::equal_to<T>, typename RAII_Object = default_lock>
class sequence {
public:
  typedef std::function<T()> run_function;
  typedef std::function<void(const T &, const std::string &)> on_error_f;
  struct run {
    run_function f;
    std::string label;
    T check;
    run(run_function f, const std::string &label = {}, T check = defaultValue) : f(f), label(label), check(check) {}
  };
  sequence(const std::list<run> &runs, const RAII_Object &raii_object = {}) : runs(runs), _check_operator(check_operator{}), raii_object(raii_object) {}
  ~sequence() {
    for(auto r: runs) {
      T result = r.f();
      if(! _check_operator(result, r.check)) {
    _run_on_error(result, r.label);
    return;
      }
    };
    _run_last();
  }
  sequence &on_error(on_error_f run_on_error) { _run_on_error = run_on_error; return *this; }
  sequence &run_last(std::function<void()> run_last) { _run_last = run_last; return *this; }
  sequence &add(run r) { runs.push_back(r); }
private:
  std::list<run> runs;
  on_error_f _run_on_error = [](const T&, const std::string&) {};
  check_operator _check_operator;
  std::function<void()> _run_last = []{};
  RAII_Object raii_object;
};
#define sequence_run(...) { __VA_ARGS__ , #__VA_ARGS__}

The sequence class accepts a list of runs as construction parameters. These are stored as a class field, and sequentially executed at class destruction.
Sequence is a template class: you can define the return value type, the success value, a comparison operator to check each function result code against the success value, and finally a generic RAII_Object, which can be as previously told a mutex locker, or some other kind of resource to unlock after API executions.

The define directive at the end of the code is used to automatically create a run object which already contains a description of the code being executed (stringified).
You get this description in the on_error callback.

Near my gphoto class I also added a typedef to conveniently call the proper sequence template class with correct template parameters:

typedef sequence<int, GP_OK, std::greater_equal<int>, std::shared_ptr<QMutexLocker>> gp_api;

Which means that gp_api accepts code blocks returning int values, that the “ok” value is GP_OK (0), and that the returned value must be equal or greater than GP_OK to be considered a success run.
It also accepts a QMutexLocker shared pointer for thread locking.
As you can see in my first example I didn’t assign the gp_api object to any variable; this means that it is immediatly created, executed and destructed, for synchronous run.

So this is a simplified usage example:

gp_api{ {
  sequence_run([&]{ return first_c_api_call(); }),
  sequence_run([&]{ return second_c_api_call(); }),
}, std::make_shared<QMutexLocker>(&&;mutex)}
  .on_error([=](int errorCode, const std::string &label) {
      std::cerr << "Error at code block " << label << ": " << errorCode << std::endl;
    })
  .run_last([&]{
    // run when everything runned smoothly
  });