tw1ddle / geometrize-lib Goto Github PK

Hi
First of all, thanks for great open source library!

Are there any performance tests or any other tests that you'd recommend running before proposing any silly perf optimizations?

Shape cloning seems to be good candidate for low hanging fruit optimizations:

• currently it always relies on heap allocations and since it happens many times per step it adds up quickly;
• it uses shared_ptr though shared ownership does not seem to be really required from the first look, perhaps price of atomic counters ticking may be avoided?

Since set of supported shapes is finite and static it may be worth trying replacing inheritance, heap allocs and shared ownership with smth stack-based and std::variant<shape1...,shapeN>-powered

If ideas like above weren't dismissed before and if PRs are accepted I can give it a go if there's agreed way to measure baseline

Lib hangs with geometrizing with rectangles.
All other shapes including rotated rectangles work normally.