Implementing ETS about firefly

[bcardarella]

@josuesantos1 sure!

[bcardarella]

I'll let @bitwalker weigh in on that

[bcardarella]

@hailelagi we'd welcome all contributions! @KronicDeth would be the best best to weigh in on this ticked I think

[hailelagi]

I see, thanks for the heads up! For now, I'd be fine with trying the library approach, and eventually when the runtime api is stable, doing a re-write, I don't mind honestly and looking forward to it :)

[josuesantos1]

can i make this issue?

[josuesantos1]

@ josuesantos1 com certeza!

I can call rets? (Rust-ets)

Can I change one thing or the other or does it need to be 100% equal to ets?

[hailelagi]

hi there! 👋🏽

I've been looking at lumen for a while, really great project! I wonder is the work on this issue still desired? If so I would love to take a stab at it, I've been actively learning rust and I'm fairly confident in my elixir - it might take me a few weeks but I'm pretty sure I can crack it. Thanks!

cc: @bitwalker , @bcardarella

[bitwalker]

@hailelagi There are a few things we need to get sorted internally with the runtime before we'll be ready to implement ETS, but I'm working hard to get that stuff done as soon as possible. I'll ping you here as soon as we're ready to begin implementation, and I'd be happy to let you work on that :).

The main outstanding things I'm working out are how runtime functions need to integrate with the garbage collector, and additionally how runtime functions should be written to support yielding during expensive operations. Both of those are tricky in Rust, so I'm researching an approach via async Rust that ideally will provide a uniform solution to both of those questions. The trouble with going too far with implementation until those are pinned down, is that they will have a very significant impact on the APIs and implementation of their functionality, so it's not really worth going too deep on it until then.

That said, if you wanted to take a crack at a standalone implementation of ETS in Rust, and then work on integrating that into the runtime once we're ready, that might be a viable approach for development - but I suspect that there will be so many runtime-specific idiosyncracies that you'd largely end up reimplementing most of it anyway.

[bitwalker]

See #8 as well