@jankomarohnic @nateberkopec Pubgrub was most of the gain in the resolver, yes.
I haven’t kept up with Bundler dev in the time since, so probably totally out of date, but at the time, I believe Gel was also independently faster at download, install, and require… plus no need for ‘exec’
@sj26 Unfortunately not… I got as far as quotes and have then been procrastinating.
My designer recommended https://t.co/eib02aCU2s, who are not your brand of local, but might show the shape of company to look for? 🤷🏻♂️
@nateberkopec@tenlinesofcode@JacobDaddario Genuinely curious about the sorts of invocations you're seeing for relation.present? that don't lead straight into .each
Previously fun-policed in https://t.co/Gc8dZMMRIe — I haven't changed my mind yet, but you see more apps than I do, so maybe I'm missing some common pattern?
@jhawthorn@tenderlove There’s also separately the small matter of “at 100+ max conns, you should almost certainly be using pgbouncer”… but that’s PG scaling wisdom rather than Rails. (And in a far future, with fine-grained/per-statement connection pooling inside AR, maybe not required?)
@jhawthorn@tenderlove I want to make the AR pool unlimited by default, to save the “max = however many threads I intend to spawn” setting (now that we do scale down an idle pool). That leaves the DB-side limit, but I guess at least you’d be getting outright “server says no” errors then 🤷🏻♂️
Since this comes up often: IMO the rule of thumb for whether or not you should yank a gem release is: don't.
Two exceptions:
- The gem was pushed by an attacker who took over your account.
- Legal reasons (e.g. copyright, etc).
1/4
@ryanbigg I swear every time we fix something like this it quietly uncovers latent bugs for some people, and allows many more to write a nicer thing in future… but only hearing from all the people who insist the now-deprecated behaviour was clearly more correct is just so exhausting 😕