Om Next Quick Start | Hacker News Reader

Om Next Quick Start

github.com

71 pointsds_22 comments

Lots of new stuff in Om Next. Unlike Relay and Falcor we have recovered HTTP caching https://github.com/omcljs/om/wiki/Remote-Synchronization-Tut.... And due to the redesign Om Next now has a really fantastically simple automated testing story https://github.com/omcljs/om/wiki/Applying-Property-Based-Te...

Happy to take any questions.

Could you clarify a bit the state of JS on ClojureScript land?

I just spend a weekend watching videos about Om Next and Devcards. I got Om Next inside Devcards working, but the journey from there to getting my JS based React components there was less than smooth.

I believe that things like GraphQL, Relay and Redux could provide what I have been missing on my web development story, namely code working with the data regardless of client/server separation. Om Next looks to give these things with a better language.

I'm now starting to build things with Om Next, since playing with Cljs on Devcards was really fun. I just had to rage quit a couple times trying to use my existing stuff.

I saw support for directly importing JS modules mentioned, but I could not find docs on how it could be used. Other option was to modify build configuration to get my JS files inserted into package space or converting them (I assume) to Clojure packages with CLJSJS. Finally I just mangled several build configs together which somehow worked.

I realize trying to keep one feet on JS land and other on Cljs side might not be high in your priorities, but my use case happens to require this. I keep trying, though, so thank you very much for what you've done.

Using the most popular JavaScript libraries is absolutely not a problem. However integrating random React components in a ClojureScript build is a work in progress (you could also just pre-build your React bits first and avoid this issue entirely). Maria Geller and others in the community have pushed the CommonJS integration forward an incredible distance, we're at the point now where the devil is in the details. I suspect in 6 months or so using a random React component in the ClojureScript build process will not be so challenging.

Ok, thank you.

I have to see how to go about making JS with Webpack -> Cljs a bit more automated.

Importing amd.js/umd/es6 modules would be preferred, but you could also include your .js files using the preamble option.

I think I used foreign-libs at the end. I might have had preamble there at one point, but that iteration did not work. Using Figwheel + Devcards build in Leininged seems to swallow compile information, or at least I couldn't see any error messages unless I broke the build very badly...

I have what could be a newbie/clueless question: with Om and Clojurescript, is there a way to automatically add dependencies in parts of a project, like 'npm i -S package_name> does for package.json rather than having to add those manually in a file together with the version number?

I also have a deeper question: you were saying Falcor's reliance on strings for the query types isn't ideal and Om Next would do that differently. How? Will it be a type of binary mapping for greater efficiency?

Also, a suggestion for the cljs quick start: make it a step-by-step which works straight away for newbies (and move the deliberate trip-up with a missing inclusion to the end, perhaps in an appendix for 'troubleshooting').

Kudos for all the great work on Om, cljs and the great presentations that I usually look for on Youtube. Your demo of an old Apple with BASIC reminded me of how I started in the 80's too: Oric Atmos 48K and BASIC with immediate feedback (and line numbers, RUN and GOTO!). This made me understand why I like REPLs and the immediate feedback environments so much and I even found an Oric Atmos emulator for Mac OS X that I installed and re-coded my first program in - a white circle on a black background (and thought the web/ui thing was badly engineered until the React guys came along) :P

Fun years ahead.

I use Reagent (another Clojurescript React library), and it seems this might widen the gap between the two in terms of ease-of-getting-started. Reagent is dead simple to get going with, and requires you learn basically nothing to build even moderately complicated things. Om is... somewhat notorious for having more complex abstractions built around deep thought about maintaining e.g. a single point of mutable data. This new stuff is... not something I think I could get right the first try.

That all being said, David Nolan is clearly a gamer -- Om will be interesting to watch even for people who aren't using it.

Reagent is very nice and by itself very easy to get started with.

That said, I think Om Next is more comparable to Re-frame (https://github.com/Day8/re-frame) as far as being more opinionated about handling data flow.

In that comparison, I don't think it involves any more complex abstractions or deep thought. In both cases, there is some of that, but both cases are also shining examples of explanatory documentation, which to me were edifying to read because they turn those abstractions into concrete implementations, with examples, and explain the deep thought in a clear manner.

If you use re-frame with reagent (and you should) then the ideas aren't that difficult to grasp. Om.next is more oppinionated regarding where you're allowed to run queries (subscriptions in re-frame) and transactions (handlers in re-frame), and forces you to build a single large query from the root component (wheras re-frame allows you to run several queries from every component).

Link for the curious: https://holmsand.github.io/reagent/

What I always worry about is not the ease-of-getting-started, but more the ease-of-scaling-modification-and-extending :)

The jury remains out on that one too. Om is a bit marmite. :)

I was hoping that in om next boilerplate will be cleaned up, but seems like it's not really different from original om. No reify is nice though.

Works great with React Devtools [0]!

[0] https://chrome.google.com/webstore/detail/react-developer-to...

As the author of Coils, another Clojure framework (coils.cc) I am really impressed to see Om Next, and I think that Om Next will become the de facto framework for building interactive websites

What about performance? How many data queries can it handle per component, realistically? How many connections can it handle? I.e., where will this solution break down?

Are the transactions (issued from within the browser) performed optimistically? If so, will this not give "flicker" in the browser in case a transaction needs to be rolled back? Is it also possible to easily perform a transaction from the server (i.e., not optimistically)?

Don't understand the question. Om.Next allows you to have a single root component, that component is responsible for building a query for every sub-component. So you really only have a single query, then render based on the results of that query. Whenever data that is touched by the query changes, affected components are re-rendered. I would also assume that the time it takes to render the view is much greater than the time it takes to run a query, although that depends on your app/query.

Ok, perhaps I should rephrase the question. Let's assume there are X items in your component (say a simple list box), and there are Y users, each editing one or more items in that same component. At what values of X and Y will Om break down?

Why I ask this is that before I get hooked on Om, I want to be sure that I'm not in a dead-end alley :) I want my application to be able to scale as necessary.

Hard to answer that with concrete numbers but Om Next handles optimistic and non-optimistic updates quite elegantly. Roughly speaking: each query expression is parsed twice, once locally and once in remote mode, and as long as your backend can handle Om Next query expressions, it Just Works:

Reading: https://github.com/swannodette/om-next-demo/blob/master/todo...

Mutating: https://github.com/swannodette/om-next-demo/blob/master/todo...

Depends entirely on your queries/mutations and app requirements. Om does produce regular React components, and due to immutable by default collections and data, Om can even beat React when it comes to rendering performance. At the very least, you should expect React-like performance. Then again, I've written a couple of frontend applications, and React/Reagent/Om has never been a problem for me from a performance standpoint. If it was, ClojureScript makes it easy to write code that uses regular js-objects and arrays, so it's always possible to optimize with "low-level" code when necessary. Like I said though, it has never been necessary.

> Are the transactions (issued from within the browser) performed optimistically? If so, will this not give "flicker" in the browser in case a transaction needs to be rolled back? Is it also possible to easily perform a transaction from the server (i.e., not optimistically)?

I was at the Lisp NYC talk where David talked about Om Next. He specifically spent some time talking about this and demonstrated how to perform optimistic updates that do not flicker.

I can't really say anything else (the talk was great, but way above my head at this point), but it seems to be able to do that.

Reply on news.ycombinator.com