Clojurescript Support for Browser

[jsmassa]

Should we support ClojureScript in the browser?

[chuckberrypi]

I don’t know if my situation is at all typical, but browser support would be huge for me. I’m not a professional developer, but I know enough that I’m able to write simple tools that would make my job much easier/more pleasant. The catch is that getting new applications approved on my firm’s machines is a pain, so writing programs to run in the browser is the only realistic way for me to build and use my own tools. Having something like this project to handle persistence in, say, indexeddb would be huge for my particular, admittedly-maybe-not-so-common use case. I’ve tried interfacing with it directly and have gotten stymied (again, not a pro…)

[garbados]

Hi! Any news here?

It seems there is an outstanding PR and the branch mentioned in the Datahike in ClojureScript with IndexedDB support blog post. It seems there has been a lot going on, now and then. I eagerly await this feature's arrival.

[whilo]

Hey @garbados!

We have focused to get the JVM version running smoothly first, cover also native builds (which are harder to recover later once you have problematic dependencies) and deliver the distributed memory model with the default transaction interface of the new REST server with JSON support. @pkpkpk and I are now looking into porting the durability in the persistent sorted set to ClojureScript, which is the first step to complete Datahike ClojureScript support. Hopefully this can be done in the next few weeks and then we will have to plan for how to render the Datahike codebase async+sync.

What are you particularly interested in? Also more help would definitely speed things up.

[whilo]

If you replicate the persistent-sorted-set then both Datahike and DataScript will effectively get replication. Datahike has not recently been compiled for ClojureScript, but assuming a synchronous execution (as is done in tonsky/persistent-sorted-set#14) fixing Datahike cljs support would mostly be working through relatively simple compiler errors. This effort is much easier than porting the code base to support asynchronous IO, which is what we did for the ClojureScript port that is now stale, https://github.com/replikativ/datahike/tree/206-cljs-support. The latter has benefits, in particular it allows to lazily load index fragments from large remote DBs instead of always having to fetch fragments first, but the drawback being asynchronous execution which can be slower and less nice to handle in JS.

You have covered a nice set of deeply related technologies and I see your dat problem. There are potential fixes for that unless you cannot rely on physical clocks or don't know all replicas through vector clocks. But handling multiple writers/conflicts is not easy in general anyway as it can induce unintuitive app behaviour when things get undone asynchronously. Regarding social apps with replikativ, we had this demo Twitter-like app https://www.topiq.es/. It is not very polished, but it also used DataScript replication, in this case of the transaction log with CDVCS instead of the index.

How can I clarify further?

[garbados]

These are excellent leads, thank you for teeing me up like this. Give me a minute to read up and chew on the problem surface, then I'll probably have more questions.

[whilo]

@garbados Lmk if you like to have a call to discuss the implementation. It should not be very difficult to sync the persistent sorted set.

[garbados]

That would be great. I'm still chewing through compilation bugs on my off-hours, but it would be good to sync up soon. Uh, how do you want to schedule something? Should I drop a line somewhere for DMs?

[whilo]

Yes, that would be great. You can also ping me on the Clojurians slack if you like.