Performance optimizations in Reagent – part 2

Have I mentioned that it’s a nightmare to install React Devtools in Electron? Turns out, it’s worth it. But it’s still a pain. If you want to try, you’ll need to first install the extension in a chrome-based browser (like Brave, the one I use) and then install it by code. At Chlorine, what I did was to copy the whole extension folder to a devtools directory and then changed the script to load electron to the script below (see the BrowserWindow.addDevToolsExtension):

const path = require('path')
const {app, BrowserWindow} = require('electron')

app.on('ready', () => {
  let browser = new BrowserWindow({
    width: 900, height: 600,
    webPreferences: {nodeIntegration: true}})
  browser.loadURL(path.join('file://', __dirname, '/index.html'))

  BrowserWindow.addDevToolsExtension('./devtools')
})

Originally, the React Devtools was installed at ~/.config/BraveSoftware/Brave-Browser/Default/Extensions/fmkadmapgofadopljbjfkapdkoienihi/4.21.0_0 (tilde is my home folder). If you’re trying to install on Atom editor, you’ll have to open developer tools and issue the following script:

r = require('remote')
r.BrowserWindow.addDevToolsExtension('./devtools') // or the right directory

This command will fail, but somehow when you reload Atom, the React Devtools will be there. There’s only one feature that I use – highlight which elements were updated. And that’s where our performance tuning starts again.
(more…)

The magical wonders of programming with pure data

I’ll start with a disclaimer – if you’re developing in ClojureScript because of performance, you’re doing the wrong thing. Although a Clojure code can be really close to Java performance, ClojureScript simply can’t match JS performance – specially because in Javascript, the fundamental data structures like arrays, sets, and objects are implemented in a faster language, and as of today, no “user space” data structure can get even close the “native” ones. And, let’s not forget that JS is slower than Java.

So, if you want to beat JS performance in ClojureScript, you have to make use of what ClojureScript offers you – the REPL, the fast experimentation, and the wonderful experience of programming with data. So, this is a tale of a huge performance improvement that I’m currently working in Chlorine.

Rendering EDN

In Chlorine and Clover, when you evaluate something, the result will be rendered in a tree-like way at the console. This works quite well for Clojure, because UNREPL makes structures that are too big small, and you can query for more data. So far, so good.

In ClojureScript, Clojerl, Babashka, etc, things are not so good. Big data structures will be rendered fully on the view. They can lock or crash your editor, they can occupy all your memory, etc. The reason for that is that tree structures are hard – when you’re rendering the “parent”, you don’t know what the children will be. Currently, in Chlorine, rendering (range 80000) locks my editor for 4 seconds until it calculates everything, layouts all data, etc… and I wanted to change this.

Reagent vs Helix vs React

I was always intrigued on how much performance hit I get when I use Reagent’s Hiccup data structure instead of React objects – after all, there must be some performance problems, right? After all, that’s the promise of Helix – to not pay this performance hit because you’re closer to what React wants.

So, first things first: profiling Javascript inside Electron IS A NIGHTMARE. I decided to install React Devtools extension on Electron (inside the Atom editor) but that got me into so much trouble, false-positives, wrong profiles that I ended up deciding against it and simply used the “performance” tab to do my profiling. There, I could see batching.cljs hogging my performance, so the next move was easy – move away from Reagent and use Helix.

Except… that Helix uses some macros and I ended up doing the code in React. My testcase was simple: render a vector of 80000 elements and see how it would perform, obviously without all the bells and whistles that Chlorine offers today (otherwise this experiment would be waaay longer). And that’s where things get surprising: with Reagent, I was hitting 1800ms of scripting, and about 120ms of rendering. With React… 1650ms of scripting, and about 200ms of rendering. I decided to do more benchmarks and probably because of OS caching, warm-up, or whatever, the results got even closer, with Reagent sometimes performing better than React – but still too slow.
(more…)

The power of Pathom

Most people don’t know the power of Pathom. Most people that I know of think that Pathom is about graphs.

They are wrong.

I mean, yes, in the end you’ll have a graph, with dependencies, but that’s not the point. The point of Pathom is the ability to transform your code into a soup or attributes. It’s also probably the best usage of qualified keywords for me, if not the only one that justifies the downsides (I’ll not enter in details here – just know that having to convert from qualified keywords to unqualified multiple times is not fun at all).

The name “soup of attributes” may not be a beautiful one, but believe me – it’s incredible. The idea is quite simple – instead of trying to handle all possible conversions from multiple sources to multiple destinations, you just define which attributes can be computed in terms of others, and Pathom does the rest. As always, things are better with examples, so let’s go.

I had to work on a system that somehow had strange rules – it needed to generate a bunch of text files for different companies. Each company expected a different file name and different fields. To generate the file, we had to accumulate data that came from a payload, from an external system that we called via REST API, and also from some data we had on our database. To make things worse, some companies would expect some of the data that was returned from REST on the filename, and there were also some state changes – like, if a file was already processed, the company would send us a return file, and we had to read some content of this file, move this file to another directory, renaming the file in the process.
(more…)

Static Typing – the dangers of incomplete info

Ok, so I’m going to use this post – Making Illegal States Unrepresentable – and I’ll add my experience to it. For people that don’t know F# (or that don’t want to check all the post to see what’s the point), the idea is that he’s trying to construct a type that will only be valid if a user does at least an e-mail address or a postal contact. Then, he ends with the following type (I’m “inventing” a way to represent this type that’s close to Scala, but easier to read for people that don’t know Scala or Haskell or F#):

type Contact {
  Name: String
  AND Contact: ContactInfo
}

type ContactInfo {
  EmailOnly: EmailInfo
  OR PostOnly: PostalInfo
  OR EmailAndPost: EmailInfo with PostalInfo
}

// Types EmailInfo and PostalInfo have to be defined also

Then, he uses 13 lines to construct a ContactInfo, and another 12 to update a contact info. He ends up concluding that these complicated types are necessary because the logic is complicated. And that’s where we start to disagree.
(more…)

Implementing shadow.remote API

Since version 0.8.0 of Chlorine, there’s a new way to evaluate ClojureScript code: that’s the Shadow-CLJS Remote API. It is basically a new REPL (not nREPL, no Socket REPL) over WebSockets to try to solve problems when translating other REPLs to ClojureScript. So, to understand why these problems exist, I’ll first introduce the difference between ClojureScript and Clojure.

On Clojure, you’re always inside a JVM. This means that compilation happens on the same JVM that your REPL, and your code is running. If you practice REPL-Driven Development, even your tests are running on the same JVM. In practical terms, it means that when you fire up your REPL, you already have everything ready to run code, compile code, and evaluate forms.

On ClojureScript, the compiler is written in Clojure – that means it’s running on the JVM. So, to produce Javascript code you don’t need a Javascript environment – and that’s when things become confusing, because when exactly will you run the REPL? Let’s try from another angle: if you start the REPL on compilation time, you can’t evaluate code (because there’s no Javascript generated, nor any Javascript engine running). If you start the REPL when you run the compiled code, this REPL can become unusable if you stop the Javascript environment, and also you have to coordinate lots of state and translations between formats.
(more…)

Reagent Mastermind

One of these days, a friend of mine posted about his experience writing the “Mastermind” game in React (in portuguese only). The game is quite simple:

  1. You have six possible colors
  2. A combination of 4 colors is chosen randomly (they can be repeated – for example, blue,blue,blue,blue is a valid combination) – you have to guess that number
  3. You have up to 10 guesses of 4 colors. For each color on the right position, you win a “black” point. For each color in the wrong position, you win a “white” point
  4. If you can’t guess on the 10th try, you loose.

So, first, we’ll create a shadow-cljs app – create a package.json file, fill it with {"name": "Mastermind"}, then run npm install shadow-cljs. Traditional stuff.

Then, we’ll create the shadow-cljs.edn file. It’ll only contain a single target (:browser), opening up a dev-http server so we can serve our code, and we’ll add reagent library dependency. I also added the material-ui dependency, but you don’t really need it for the code. Now, running npx shadow-cljs watch browser will start a webserver at port 3000, and we can start to develop things.
(more…)

ClojureScript vs clojure.core.async

I’m going to make a somewhat bold statement: core.async does not work with ClojureScript. And, in this post, I’m going to show some examples why this is true, at least for the current versions of core.async.

So let’s start by understanding a little bit about the runtime: Javascript is a single-threaded runtime that implicitly runs an event-loop. So, for example, when you ask to read a file, you can do it synchronously or asynchronously. If you decide to run in that asynchronously, it means that as soon as you issue the fs.readFile command, you need to register a callback and the control is returned to the “main thread”. It’ll keep running until it runs out of commands to execute, then the runtime will wait the result from the callback; when it returns, the function that you registered will be called with the file contents. When the function ends, the JS runtime will await to see if there’s any other pending call, and it’ll exit if there’s nothing else to do.

The same thing happens in browser environment, but in this case the callbacks are events from the DOM: like clicking on buttons or listening for changes in some elements. The same rules apply here: the runtime is single threaded and when something happens it will first execute everything that needs to be executed, then it will be called back with the event that happened.

So maybe we can change these callbacks with core.async channels right? But the answer is no, because core.asyncs go blocks will not run in different threads (because, again, the runtime is single-threaded). Instead, it creates a state machine and it’ll control of when each of these go blocks will be called, at what time, eventually replacing the event-loop that Javascript environment already have.
(more…)