eth/filters: retrieve logs in async

[jsvisa]

Try to implement #15063 and discussed with @s1na,
In the first phase, we want to make the filter.Logs() function returns a channel instead of a large array

[s1na]

some early comments

[jsvisa]

@s1na please take another look, thanks

[holiman]

I haven't reviewed this in any depth, so maybe I'm missing something... but it looks to me like this spins up two routines that may be stuck forever trying to deliver, in case the other side is no longer reading, for whatever reason.
Wouldn't it make sense it use the ctx to check if it's time to abort these routines?

[jsvisa]

Good point. But seems we have handled this issue inside the indexedLogs https://github.com/ethereum/go-ethereum/blob/101ac683e447a2c23fc0141d33bbb3f2e0e32607/eth/filters/filter.go#L266

I wonder if I've lost something

[s1na]

If we want to protect against potential receiver failure, we'd have to wrap every channel send with a select, as in:

select {
  case logchan <- log:
  case ctx.Done():
    return ctx.Error()
}

I guess I'm just wondering if this has an impact on performance or that should be negligible

[s1na]

This PR is looking very nice overall! I'm gonna do some testing.

[s1na]

LGTM!

[jsvisa]

@s1na @holiman please take another look, thanks

[holiman]

One remaining nit, otherwise lgtm

[s1na]

I saw the timeout testcase I added fail once so I re-adjusted the timeout. It might turn out to be a flaky one. In that case I'll remove it in a future PR.

[s1na]

Needs rebase

[jsvisa]

I'll rebase it right now

[jsvisa]

Rebased, and please take another look @s1na @holiman :)

[holiman]

LGTM. We might be putting out a release pretty shortly, so I'd prefer to wait with merging until afterwards