Can one of the admins verify this patch?

Hi Will!

We try to bring the behavior of the RooAbsCollections closer to the STL containers like std::vector. This means that operator[] should simply access an element as quickly as possible without any bounds checking for no overhead. That's why it changed from 6.24 to master.

Only the at() member should do the bounds checking. I still have some work to do here because its behavior is inconsistent with std::vector (it returns a nullptr if the bounds check fails).

So bringing the element-accessing behavior closer to Python behavior would unfortunately bring us further away from the STL behavior, and I would advise against the suggested change.

However, I would agree that on the pyROOT side, the RooAbsCollections should behave more pythonic and we should absolutely implement a pythonization to support element access from the back with negative indices!

Is that a fair compromise? Did you intend to use the negative indices in C++?

This sounds good to me, yes this is purely for the pythonization, if there's a dedicated place where that is coded up then yes supporting it there (with an exception if trying to access invalid index) would be great! Atm it's very user un-friendly that ROOT crashes if we access an index out of range in pyROOT

Okay, very good! Just FYI, the pythonizations are implemented in these "mirror classes" that I declared here:
https://github.com/root-project/root/tree/21fbd344e60a849622991a92d3230681e257c3b9/bindings/pyroot/pythonizations/python/ROOT/pythonization/_roofit

Every function that is defined in a mirror class is then bound to the actual RooFit class, with the original cppyy function name changed to have an underscore prefix.

I have already a PR open with some more pythonizations of RooAbsCollection: #8179. I can just add one more commit with the element access.

But in general, I'm not working too much on these pythonizations because we will most likely get a Google summer of code student for that project! So if you have any further nice ideas for RooFit pythonizations, feel free to write them down in this issue and I will what can be implemented with the student.

Thanks again for the idea! It has been implemented in #8179, so I will close this PR.

Please let us know in #7217 if you have further ideas for pythonizations!