Rhombus feels slow and perhaps a bit too dynamic

[sidkshatriya]

My first impression of Rhombus is that it is an excellently designed system. Wow -- everything is so principled and clean ! The Rhombus paper and OOPSLA presentation were both impressive -- thanks !

I'm facing three major issues:

1. DrRacket seems very slow with Rhombus. It's possible that there are just too many levels of abstraction between rhombus and chez. The whole editing experience does not feel very smooth.
2. racket-langserver works in Rhombus but again feels very slow and not very performant/reliable. If people want to use Rhombus they should not have switch to DrRacket -- this just makes the activation energy a bit too high
3. The language feels a bit too dynamic. When I program in C/C++ or Rust for example I'm really used to the ability to go to definition, look at the type definitions, search for usage and otherwise enjoy excellent LSP support in my editor. In Rhombus there are a lot of amazing meta programming features but that hurts the ability for the editor/IDE to help you out. I understand that dynamic/static is a tradeoff but in my view this could limit Rhombus usage. Even traditionally dynamic languages like Python are moving towards more typing. I personally would have loved to see Rhombus with a focus on typing (and additional metaprogramming facilities) rather than meta programming being front and centre and then some support for typing. Racket already has great metaprogramming and if Rhombus was more typed by default, it could have been a more compelling reason to use it (I love the new syntax in Rhombus but it may not be a sufficient pull to switch to it for a lot of people).

I think (1) and (2) can be dealt with sufficient engineering effort. (3) is more of a personal opinion

[distractedlambda]

To point (3): I think there's (a lot) more static information already in Rhombus, than the tools are currently able to take advantage of. For instance, "dot providers" make it possible to statically know methods that should be available on a particular value, and hypothetically this could drive some sort of method autocomplete. There's also a lot of places where definition/use-pairings are not detected as reliably as with Racket, even though it's all being resolved statically during compilation (I think this may be a side-effect of binding spaces?).

[jackfirth]

There are probably many instances of (3) that are fixable. If you have specific examples, opening issues for them would be helpful.