@rauh can you please expand your idea?

@tawus so you basically had edn values over websocket before your optimization, and edn values over websocket after that. sorry that I don't get it, but why it started to work faster?

is it that datomic tree datastructure was serialized instead of json-like values and that made it faster, right?

Encoding the path as a single int with a level shift, is this a well known technique? https://github.com/tonsky/datascript/blob/master/src/datascript/btset.cljc#L20

Can anyone point me to any reading?

Presumably it's a performance optimization, what specifically does it help with?

I probably learned that from Clojure’s hash map implementation

I see, I found some explanation here: http://blog.higher-order.net/2009/09/08/understanding-clojures-persistenthashmap-deftwice.html I'll see if I can work out what the benefit is

@leov I wanna try out my idea some more, then I'll put some code up at some point. Maybe a gist or possibly even a full library. Right now it seems to work rather nicely and I'm happy

So I only had 2 queries in my app and otherwise mostly just did raw index access. Once I have an entry point my react components just walk along the graph using entities. I just removed the query engine from datascript and it shaved off 22kb (gzipped) and ~130k minified. Now datascript is really lightweight and still provides tons of functionality

Hi, I’m experimenting with datascript and reactive streams. It’s fun but could you point me to some nice prior art? I know posh and I’m dissecting it’s sources but I would like go for streams instead of atoms and I’m wondering if someone took this aproach? Could be datascript or datomic.