13:53 <ChrisBr> headius: enebo: happy new year, long time no see! How are things going? Is there any update regarding the new hash implementation?

14:28 <headius> ChrisBr: I have played around with some experiments, some drastic (move your impl completely out to a separate class) but have not managed to find a path I really like

14:28 <headius> I think a key concurrency issue is switching from linear to addressed

14:29 <headius> the experiment I did split those into separate classes altogether and then RubyHash would switch from one to the other...but it adds a couple levels of abstraction

14:29 <headius> Happy new year to you as well!

14:30 <headius> it's tough to clean up concurrency edges because there's so many state changes

14:31 <headius> I'm wondering if we should only use linear for hashes constructed from literal hash syntax, since in many (most?) cases they don't change size after that

14:31 <headius> well also for keyword args since they also tend to get packed and unpacked without being mutated

14:39 <ChrisBr> hm, so having two different hash implementations?

14:39 <ChrisBr> completely seperated

14:39 <ChrisBr> not intermixed like it is currently?

14:40 <headius> That was my experiment

14:41 <headius> it worked reasonably well but I was still promoting it to open addressing when it got larger

14:42 <headius> may be some useful research out there on making open addressing hashes concurrency-safe, like https://dl.acm.org/citation.cfm?id=2851196

14:51 <ChrisBr> hm, ok! So anything I could do / help to get it done? Do you still plan / want to get it in?

14:51 <ChrisBr> did you already read the linked paper?

14:53 <headius> I have not read the paper, no...and I definitely still want us to get it in!

14:54 <headius> I'm not sure about direction...hopefully enebo will be around in a bit and we can chat about that

14:54 <headius> I was never able to reproduce the concurrency issues he had but it sounded fairly consistent

15:00 <ChrisBr> ok

15:00 <ChrisBr> I will be around the next 1-2 hours and maybe tonight as well, just ping me

15:06 <headius> ok sure

21:50 <headius> ChrisBr: I guess enebo is still out sick today