Multiple Energy Window Code

[ludovicabrusaferri]

This is a Pull Request for merging the multiple energy window code into master

[KrisThielemans]

Travis fails because of undefined virtual energy functions for CListModeDataGEHDF5. Probably easiest not to derive from CListEnergy?

I didn't check why Appveyor or codacy failed...

[ludovicabrusaferri]

Ok I have removed the dependency. I guess it will be the same for CListRecordECAT too? Let me know.

[KrisThielemans]

Possibly we cannot just delete it as LmToProjData uses record.energy().get_energyA_in_keV().

The only way I see to resolve that is by doing thing like

// before the loop, do a `dynamic_cast` to test if we have energy
auto clist_energy_ptr =  dynamic_cast<CListEnergy const *> (record);

for (...)
{
  if (clist_energy_ptr)
  {
    // do checks on energy
   }
}

a bit ugly. Maybe you see another way?

[KrisThielemans]

Another problem is a much bigger one: what do we do with SPECT where there's only 1 energy window. Maybe @danieldeidda has some great ideas...

[danieldeidda]

Another problem is a much bigger one: what do we do with SPECT where there's only 1 energy window. Maybe @danieldeidda has some great ideas...

sorry I need to read through this PR, but why 1 energy window would be a problem?

[KrisThielemans]

why 1 energy window would be a problem?

a lot of the code talks about pairs. This is unavoidable (and correct) for scatter, but will be a pain for SPECT, as there it doesn't make sense. Somehow we'd have to "hide" it I guess. Or replace a "pair" with a single index.

We could shift the single "pair" index into Bin I think. That way, it seems we can let LmToProjData be agnostic about energy windows. I haven't thought about this properly though...

[ludovicabrusaferri]

I can see indeed this could be an issue for spect. I need to think about it carefully too , I'll come back to this asap

[KrisThielemans]

Travis: seems some things are failing with the recon_test_pack: ROOT support test scripts (e.g. run_ROOT_gate.sh) but also run_scatter_tests.sh (although that only with error relative to sup-norm of first data-set = 0.435529 % at the very end)

Appveyor: a lot of warnings make it hard to see where the errors are (some are due to main STIR, so I'll fix the main culprit in ExamInfo) but I found

C:\projects\stir\src\scatter_buildblock\scatter_detection_modelling.cxx(244): error C2065: 'M_PI': undeclared identifier [C:\projects\stir\build\src\scatter_buildblock\scatter_buildblock.vcxproj]
C:\projects\stir\src\scatter_buildblock\scatter_detection_modelling.cxx(244): error C2789: 'den1': an object of const-qualified type must be initialized [C:\projects\stir\build\src\scatter_buildblock\scatter_buildblock.vcxproj]

For the first error, use _PI (M_PI isn't portable, _PI is defined in stir/common.h which is always included). The 2nd is probably just a consequence of the first.

[robbietuk]

As mentioned in #736, I cant figure out the benifit having energyA adn energyB.