ctools

A first working solution is implemented in https://cta-gitlab.irap.omp.eu/dtiziani/ctools/tree/2714-joint-binned

I use the value of the parameter outcube for the filename of the output observation definition XML file if stack=no.
Maybe this is misleading and the parameter name should be changed? Or should we introduce a new parameter for that?

I integrated your code and it’s now merged into the devel branch.

I changed the outcube parameter to outobs.

I also did some minor adaptions in the code. For example the presence of a cache for sky directions, which was added by Josh to speed-up computations, posed a problem in the modified code. The cache is now only used for stacking, while on-the-fly computation is used for multiple counts cubes. Also, each counts cube may have a different centre, which can be enabled by the usepnt=yes parameter. Typically you want to do this when you make a joint binned analysis. I modified the code so that this also works. Finally, I used slot 0 of the arrays to store the stacked counts cube, avoiding having separate data members for that case.

I also changed a bit the construction of the filename for multiple counts cubes. An identifier is only unique for a given instrument, but you could imagine having a H.E.S.S. and a CTA observation in the same observation definition file with the same identifier, hence I added the instrument name to the filename. I also converted all blanks into underscores, since it is allowed to have blanks in an identifier, but they may not work well on all filesystems.

I also added a cubes() method that returns the number of counts cubes that can be accessed through the cube() method. There is just a single cube() method now with a default argument of zero, hence it can be used like before, but also with an index.

You may check if everything works satisfactory on your side, so that we can close the issue at the end of the sprint.