refactor: Add macro to simplify algorithm binding

[benjaminhuth]

While working with the python bindings, I thought its quite a lot of repetive boiler plate code to add an Algorithm to the python bindings. With the help of Boost.Preprocessor this can be simplified quite a lot:

ACTS_PYTHON_DECLARE_ALGORITHM(
    ActsExamples::FooAlgorithm,
    "FooAlgorithm", 
    inputA, inputB, output)

expands to

{
  using Alg = ActsExamples::FooAlgorithm;
  using Config = Alg::Config;
  auto alg = py::class_<Alg, ActsExamples::BareAlgorithm, std::shared_ptr<Alg>>(
                 mex, "FooAlgorithm")
                 .def(py::init<const Config &, Acts::Logging::Level>(),
                      py::arg("config"), py::arg("level"))
                 .def_property_readonly("config", &Alg::config);
  auto c = py::class_<Config>(alg, "Config").def(py::init<>());
  ACTS_PYTHON_STRUCT_BEGIN(c, Config);
  ACTS_PYTHON_MEMBER(inputA);
  ACTS_PYTHON_MEMBER(inputB);
  ACTS_PYTHON_MEMBER(output);
  ACTS_PYTHON_STRUCT_END();
}

This supports any number of arguments due to the preprocessor loops of Boost. Since we use macros anyways here to simplify declaration, I think this could be an acceptable use case of another macro.

I have this already appied for the Exa.TrkX algorithm, but wanted to wait for feedback (@paulgessinger) before doing the rest of the algorithms.

[codecov]

Codecov Report

Merging #1510 (4de4b56) into main (eef2627) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main    #1510   +/-   ##
=======================================
  Coverage   48.54%   48.54%           
=======================================
  Files         381      381           
  Lines       20789    20789           
  Branches     9539     9539           
=======================================
  Hits        10092    10092           
  Misses       4112     4112           
  Partials     6585     6585           

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[paulgessinger]

Existing python tests seem to fail with this change. Can you check if you can reproduce locally, @benjaminhuth?

[benjaminhuth]

Yeah I already know what the problem is. Most writers/readers do not provide a config() accessor in C++, so I left the corresponding macro without defining that in python.

However, it seems like that the config property it is used in one place at least.

I see three options there

• remove that one use case
• Add config accessors to all readers and writers and defining that in the macro
• Don't use the macro in the one special case (not a good solution I think)

I would go with the second option, because I think it is useful to have this accessor... What do you think @paulgessinger ?

[paulgessinger]

Yes I agree, adding the config accessor everywhere seems like the right thing to do.

[benjaminhuth]

@paulgessinger This should be ready to go in now, could you have a look and approve?

[paulgessinger]

Looks good I think. If the output doesn't change, I'm happy.