satori VS yoga

View on GitHub

I've used satori [0] on the backend with TypeScript/Deno to render JSX as an SVG (which is then rendered to a PNG).

Satori is meant for rendering Open Graph images (e.g. the little images that come up when you post a link on Twitter/Slack/Facebook), but I found that it works well for rendering arbitrary images. It supports a subset of modern CSS, including flexbox.

My use case is posting match reports for League of Legends into a Discord text channel, e.g. person X just played a match, here are their stats.

It's quite nice because there are almost zero server-side native dependencies (the one exception is the library to convert svg -> png requires some native libraries).

Here's what a match report looks like: [1]

Here's an example of what the JSX looks like: [2]

[0]: https://github.com/vercel/satori

[1]: https://github.com/shepherdjerred/glitter/blob/main/assets/p...

[2]: https://github.com/shepherdjerred/glitter/blob/main/packages...

Another way is to write HTML/CSS and use satori [0] to convert that to SVG. It's meant for Open Graph images (the images that show up when you link a site in Discord, Slack, Twitter, etc.), but it works quite well for anything.

This is obviously not as flexible as true SVG, but it is familiar to author for anyone who's written a React application. I've used it on the backend to generate match reports for League of Legends [1]

[0]: https://github.com/vercel/satori

[1]: https://github.com/shepherdjerred/glitter-boys

Install the @vercel/og package. This library is designed to convert React code into PNG images. It is built on Satori, a library that converts HTML and CSS into SVGs.

Examples are available in the Vercel OG Playground.

In this version, we no longer use Puppeteer to capture HTML and return images. Instead, we utilize the @vercel/og library, which employs Satori as its core engine. Satori is a library that converts HTML and CSS into SVG.

This is made possible thanks to the Dynamic Open Graph Image Generation feature introduced with Next.js version 13.3, and the new Metadata API. In summary, it involves generating images using code (in our case, TSX, HTML, and CSS) with the help of the libraries @vercel/og (already integrated in the App router) and Satori. Satori converts HTML and CSS to SVG, and then resvg-js converts the SVG to a PNG image. All of this in just a few milliseconds!

Here you are returning an ImageResponse instead of the Response, alternatively you can also extend the request and response web api using 'NextRequest' and 'NextRespone', to do that you can import them using import { NextResponse, NextRequest } from 'next/server';, though for this example it is not required. Now if you refresh your browser you will get an image generated by your 'route.js' at request time. Well we are almost done. You can render whatever dynamic data in your image you want and customize your image using og playground, you can even generate 'SVG' on request as the og image. For this example we will fetch a random number from random.org api, then we will use that number as an id and fetch an image from Lorem Picsum, with the same image url we will fetch the description for the image from the Alt Image Generator and generate an image on request with the image that we fetched and the description we have fetched and use it in a design to create the og image. Kind of like that.

Use Vue files to generate SVG images by Satori. The image can be generated by running ESM script or CLI.

I'm not familiar with it but looks like it's made with nodejs, node uses the same js engine used by chrome, node renders the template and converts it to plain html/css and then they use this library to convert it to png but in the library github page it says that they don't support everything so it's kinda similar to xhtml2pdf or weasyprint

We still wanted to leverage a layout engine that could be performant and easy-to-use. After doing some performance testing with native UIKit, Autolayout, and a few other third-party options, we ended up bringing FlexLayout into the mix, which is a Swift implementation of Facebook’s Yoga layout engine. All RPL components utilize FlexLayout in order to lay out content fast and efficiently. While we’ve enjoyed using it, we’ve found a few touch points to be mindful of. There are some rough edges we’ve found, such as utilizing stack views with subviews that use FlexLayout, that often come at odds with both UIKit and FlexLayout’s layout engines.

We have our own test suite (orginally derived from the test suite of Meta's Yoga layout library [0]) which consists of text fixtures that are small HTML snippets [1] and a test harness [2] that turns those into runnable tests, utilising headless chrome both to parse the HTML and to generate the assertions based on the layout that Chrome renders (so we are effectively comparing our implementation against Chrome). We currently have 686 generated tests (covering both Flexbox and CSS Grid).

We would like to utilise the Web Platform Test suite [3], however these are not in a standard format and many of the tests require JavaScript so we are not currently able to do that.

[0]: https://github.com/facebook/yoga

[1]: https://github.com/DioxusLabs/taffy/tree/main/test_fixtures

[2]: https://github.com/DioxusLabs/taffy/tree/main/scripts/gentes...

[3]: https://github.com/web-platform-tests/wpt/tree/master/css/cs...

The "engine" is built on top of amazing https://www.lwjgl.org/ and https://github.com/bkaradzic/bgfx/, and UI system is baked by https://github.com/memononen/nanovg and https://github.com/facebook/yoga

React Native uses the Yoga engine under the hood, which allows you to use CSS properties to layout your React Native UI in a way that translates really well. Layout in Yoga is limited to Flexbox and absolute/relative positioning, however; there is no CSS grid and no display attribute. This keeps things simpler and more performant, but if developers are accustomed to using other layout techniques on the web, they’ll need to adjust to this new limitation.

PR #246 is super interesting to check out: by fixing the caching strategy, we were able to eliminate an exponential time (with respect to tree depth) performance penalty, and get comparable speeds for flat and deeply nested layouts (something I'd never expected to be possible). Preliminary benchmarks shows us significantly faster than yoga, Meta's C++ library for the same thing, especially on deep trees. Not too shabby for a tiny team of volunteers!

You're going to build the CLI using Ink, a React component-based library for building interactive CLIs. It uses Yoga to build Flexbox layouts in the terminal, so most CSS-like props are available in Ink as well. Ink is simply a React renderer for the terminal, so all the React features are supported. No need to learn a new syntax specific to Ink.

Interesting.

I was thinking that this was going to be a crazy amount of layout engine work, but now I look a little closer it appears the layout work is farmed out to yoga [0] (not trying to take away anything from the effort here). So this project is almost a wrapper around running yoga as a renderer and using SVG as a form of backend target?

I say "appears" because the yoga landing page doesn't do a great job of explaining what it does.

[0] https://github.com/facebook/yoga