DougLeonard

> > Please make sure that if the `__new__` return is untyped (returns `Any`) then it uses the class type implicitly. Not sure if that is already covered, but I...

It's in my linked reply above, but not a problem. From PEP484: > Any function without annotations should be treated as having the most general type possible, or ignored, by...

> If the return type is left as `Any`, then created objects that have an un-annotated `__new__` would have type `Any`, right? So nothing would be rejected. You would just...

Thanks for the attention to it. Maybe I was too enthusiastic in the context of displayed objects and interactive root sessions. I'm not an expert, but still learning. Th RDirectory...

Another minor related point: ``` - TH1F *test = new TH1F("test","This is test histogram",100,-4,4); + auto test = RDirectory::Heap().Create("test", "test","This is test histogram",100,-4,4); ``` I guess this gives us what,...

TCanvas is not fully compatible with unique_ptr reset in root 6 though. If you reset it to a TCanvas of the same name, garbage collection causes it to be double...

> Thanks for your comments! Thanks for the discussion. I think many students learn C++ from root examples so it matters. Thanks for the link to the docs. I got...

As a bonus I think my code is more potentially thread safe. It's not something I'm real expert on, but it looks pretty thread safe. It's dealing with whatever it's...

> The intent is the opposite. The RDirectory's internal representation and what it returns is a kind of shared_ptr. The user register the same name twice then if they hold...

For what it's worth, for fun I wrote optimized an assembly version of the inner loop using ~maximally unrolled~ maximally gratuitous SSE, the core of the code (that might even...