What is the place for tests?

[heavyionanalysis]

What is the place for tests?

[kreisl]

What I suggested is to put unit test together with the classes. Such as Axis.cpp and Axis.test.cpp in the same folder so it is kind of mandatory to write them at the same time and when we make changes the same pull request should also include the tests. That would be ideal. I have an additional tests folder, which might contain more complicated tests, which might need some manual intervention. For example a MC simulation, which is maybe difficult to formulate as an automatic test. It could create a root file or pdf file which needs inspection. Those tests in /tests would be a higher level and are not strictly needed to verify the mathematics of QnTools. However they might be nicer to look at for physicists. What do you think of this proposal?

[eugene274]

Well, I've never seen that way of organization, anyway it sounds reasonable for me. But I get used to have following structure:

src/
|-- targetA/
|    `-- ClassA.cpp
`-- targetB/
     `-- ClassB.cpp
tests/
|-- targetA/
|    `-- TestClassA.cpp
`-- targetB/
     `-- TestClassB.cpp

It allows keeping all testing infrastructure (e.g. googletest) inside tests/ and switching on/off all testing by adding or not-adding tests folder to CMake.

MC test is not actually a unit test. I would put it to examples/.

[kreisl]

In the implementation I have largely followed this project structure presented here:
http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2018/p1204r0.html
In particular I have implemented this approach:
https://api.csswg.org/bikeshed/?force=1&url=https://raw.githubusercontent.com/vector-of-bool/pitchfork/develop/data/spec.bs

I know some things are a bit controversial.
I quite like the split of unit tests which should test a single class / file. and larger tests which go into tests.
I could not keep all of the source files in the same directory because ROOT caused me some problems during the dictionary generation.

[kreisl]

I would argue that we keep a "MC closure test" for physicists to easily digest and appreciate the features of the framework from a physics perspective and that it would belong to the test category. I like your idea of having an examples folder, but I would see this as an additional thing.