Hans Klunder

@robertsLando My bad, I should have explained it better! The original aedes-persistence was synchronous. One would get an instance as soon as one called memory() https://github.com/moscajs/aedes/blob/main/aedes.js#L63 This works great for...

well actually, the broker needs to be passed to setup and only aedes knows about the broker (currently) If we want to do the explicit setup that leaves us 2...

Automated Benchmark testing suggests more than 10℅ perf decrease. I will try to figure out what is causing this. Kind regards, Hans

It turns out the difference was actually positive instead of negative, tested on Codespaces using 4 cores. ---- # Overall Benchmark Results +x% is better, -x% is worse, current threshold...

A fix for the benchmarking is in https://github.com/moscajs/aedes/pull/1046 Kind regards, Hans

@robertsLando - The objectPool has been removed leading to no material performance difference HEAD is still approx 18% faster than MAIN - timer.refresh() has been applied (I didn't know it...

@robertsLando I also added a test to test/wills.js to: - show that the new clearwills mechanism works with a lager number of wills - increase test coverage

@robertsLando also updated handlers/subscribe.js: as `completeSubscribe ()` was no longer called with an 'err' parameter, the `if (err)` would never be called decreasing coverage. The only file without 100% coverage...

FYI: with all the changes the github review comments are a bit hard to track. I think everything is in, if I missed anything please let me know. I will...

@robertsLando @mcollina Benchmark had been done, see https://github.com/moscajs/aedes/pull/1045#discussion_r2334328190 Can you please review again ? Kind regards, Hans