Oscar Benjamin

I really don't mind. Black's 88 is also a sort of standard in Python land now. I would like to point out that a lot of this linting stuff started...

I think that there are advantages to automatic simplification in many cases but the problem with the design of both SymPy and SymEngine is the fact that there isn't a...

> Transition between inert and semi-canonicalized expressions could be done with the `refine` function. I'm not sure what you mean by this. Fundamentally what happens if I use symengine's Python...

> In any case there needs to be a system that can detect which lowering would be best to use. See SymPy's polys module and functions like `construct_domain`, `primitive_element`, `Domain.unify`...

I suspect that SymEngine would need different types or data structures for inert expressions. I think that things would be designed very differently in that case. It can be possible...

> @oscarbenjamin in your experience, what are the best data structures to use to represent the symbolic tree "as is"? It depends what you want to do with the tree...

Note that you don't need anything as exotic as the zeta function to run into performance issues with automatic evaluation: ```python In [3]: import symengine In [4]: x = symengine.symbols('x')...

I think that they can live together but designing it so that it is understandable for users needs careful thought. I would probably not allow the two types to be...

A decision needs to be made whether this is going to: - Use canonical representations like Poly or - Just generate large unsimplified expressions. A function like cancel should only...

Your code keeps converting backwards and forwards between Expr and Poly. When I say that the whole calculation would be done with a canonical representation I mean that you convert...