We at nata.house have forked this lib in order to add some functionality,docs and fixes to keep using it smoothly for our projects, while the original repo is paused. Read more here.
Pre-renders a web app into static HTML. Uses Headless Chrome to crawl all available links starting from the root. Heavily inspired by prep and react-snapshot, but written from scratch. Uses best practices to get the best loading performance.
- Enables SEO (Google, DuckDuckGo...) and SMO (Twitter, Facebook...) for SPAs.
- Works out-of-the-box with create-react-app - no code-changes required.
- Uses a real browser behind the scenes, so there are no issues with unsupported HTML5 features, like WebGL or Blobs.
- Does a lot of load performance optimization. Here are details, if you are curious.
- Does not depend on React. The name is inspired by
react-snapshotbut works with any technology (e.g., Vue). - npm package does not have a compilation step, so you can fork it, change what you need, and install it with a GitHub URL.
Zero configuration is the main feature. You do not need to worry about how it works or how to configure it. But if you are curious, here are details.
Install:
yarn add --dev react-snapChange package.json:
"scripts": {
"postbuild": "react-snap"
}Change src/index.js (for React 16+):
import { hydrate, render } from "react-dom";
const rootElement = document.getElementById("root");
if (rootElement.hasChildNodes()) {
hydrate(<App />, rootElement);
} else {
render(<App />, rootElement);
}That's it!
To do hydration in Preact you need to use this trick:
const rootElement = document.getElementById("root");
if (rootElement.hasChildNodes()) {
preact.render(<App />, rootElement, rootElement.firstElementChild);
} else {
preact.render(<App />, rootElement);
}Install:
yarn add --dev react-snapChange package.json:
"scripts": {
"postbuild": "react-snap"
},
"reactSnap": {
"source": "dist",
"minifyHtml": {
"collapseWhitespace": false,
"removeComments": false
}
}Or use preserveWhitespace: false in vue-loader.
source - output folder of webpack or any other bundler of your choice
Read more about minifyHtml caveats in #142.
Example: Switch from prerender-spa-plugin to react-snap
Only works with routing strategies using the HTML5 history API. No hash(bang) URLs.
Vue uses the data-server-rendered attribute on the root element to mark SSR generated markup. When this attribute is present, the VDOM rehydrates instead of rendering everything from scratch, which can result in a flash.
This is a small hack to fix rehydration problem:
window.snapSaveState = () => {
document.querySelector("#app").setAttribute("data-server-rendered", "true");
};window.snapSaveState is a callback to save the state of the application at the end of rendering. It can be used for Redux or async components. In this example, it is repurposed to alter the DOM, this is why I call it a "hack." Maybe in future versions of react-snap, I will come up with better abstractions or automate this process.
Make sure to use replace: false for root components
- Emotion website load performance optimization
- Load performance optimization
- recipes
- stereobooster/an-almost-static-stack
If you need to pass some options for react-snap, you can do this in your package.json like this:
"reactSnap": {
"inlineCss": true
}Not all options are documented yet, but you can check defaultOptions in index.js.
Experimental feature - requires improvements.
react-snap can inline critical CSS with the help of minimalcss and full CSS will be loaded in a non-blocking manner with the help of loadCss.
Use inlineCss: true to enable this feature.
TODO: as soon as this feature is stable, it should be enabled by default.
You can define a list of blobs that tells react-snap which routes it should ignore when crawling.
This is specially useful when you have parts of your app that don't need to be pre-rendered. It can be very helpful if you have pages filled with client-side logic, that causes react-snap to maybe behave on unexpected ways. Just exclude these complex rotues!
react-snap will look for all the available routes it can find, but maybe your application doesn't explicitly list these routes and the crawler has been unable to find and pre-render them. In this case, you can specify routes that will be crawled no matter what.
react-snap will also crawl all of the subsequent routes found from the ones listed on include.
Important: Make sure that your exclude and include configurations do not overlap, as this is not make sense and will not be allowed to execute!
Sometimes you may want react-snap to output formats different than /route/index.html in order to use it more smoothly with some web servers or CDNs. Currently there is one built-in alternative for output formats, which changes /route/index.html to just /route.html.
Another example would be: /path/to/route/index.html becomes /path/to/route.html.
This might be another useful config specially when applying react-snap to older or bigger unstable projects, that might have several failing https requests. By setting this to true you may be able to bypass some errors and successfully pre-render parts of your app.
Also known as code splitting, dynamic import (TC39 proposal), "chunks" (which are loaded on demand), "layers", "rollups", or "fragments". See: Guide To JavaScript Async Components
An async component (in React) is a technique (typically implemented as a higher-order component) for loading components on demand with the dynamic import operator. There are a lot of solutions in this field. Here are some examples:
It is not a problem to render async components with react-snap, the tricky part happens when a prerendered React application boots and async components are not loaded yet, so React draws the "loading" state of a component, and later when the component is loaded, React draws the actual component. As a result, the user sees a flash:
100% /----| |----
/ | |
/ | |
/ | |
/ |____|
visual progress /
/
0% -------------/
Usually a code splitting library provides an API to handle it during SSR, but as long as "real" SSR is not used in react-snap - the issue surfaces, and there is no simple way to fix it.
- Use react-prerendered-component. This library holds onto the prerendered HTML until the dynamically imported code is ready.
import loadable from "@loadable/component";
import { PrerenderedComponent } from "react-prerendered-component";
const prerenderedLoadable = dynamicImport => {
const LoadableComponent = loadable(dynamicImport);
return React.memo(props => (
// you can use the `.preload()` method from react-loadable or react-imported-component`
<PrerenderedComponent live={LoadableComponent.load()}>
<LoadableComponent {...props} />
</PrerenderedComponent>
));
};
const MyComponent = prerenderedLoadable(() => import("./MyComponent"));MyComponent will use prerendered HTML to prevent the page content from flashing (it will find the required piece of HTML using an id attribute generated by PrerenderedComponent and inject it using dangerouslySetInnerHTML).
- The same approach will work with
React.lazy, butReact.lazydoesn't provide a prefetch method (loadorpreload), so you need to implement it yourself (this can be a fragile solution).
const prefetchMap = new WeakMap();
const prefetchLazy = LazyComponent => {
if (!prefetchMap.has(LazyComponent)) {
prefetchMap.set(LazyComponent, LazyComponent._ctor());
}
return prefetchMap.get(LazyComponent);
};
const prerenderedLazy = dynamicImport => {
const LazyComponent = React.lazy(dynamicImport);
return React.memo(props => (
<PrerenderedComponent live={prefetchLazy(LazyComponent)}>
<LazyComponent {...props} />
</PrerenderedComponent>
));
};
const MyComponent = prerenderedLazy(() => import("./MyComponent"));- use
loadable-components2.2.3 (current is >5). The old version ofloadable-componentscan solve this issue for a "snapshot" setup:
import { loadComponents, getState } from "loadable-components";
window.snapSaveState = () => getState();
loadComponents()
.then(() => hydrate(AppWithRouter, rootElement))
.catch(() => render(AppWithRouter, rootElement));If you don't use babel plugin, don't forget to provide modules:
const NotFoundPage = loadable(() => import("src/pages/NotFoundPage"), {
modules: ["NotFoundPage"]
});
loadable-componentswere deprecated in favour of@loadable/component, but@loadable/componentdroppedgetState. So if you want to useloadable-componentsyou can use old version (2.2.3latest version at the moment of writing) or you can wait untilReactwill implement proper handling of this case with asynchronous rendering andReact.lazy.
See: Redux Server Rendering Section
// Grab the state from a global variable injected into the server-generated HTML
const preloadedState = window.__PRELOADED_STATE__;
// Allow the passed state to be garbage-collected
delete window.__PRELOADED_STATE__;
// Create Redux store with initial state
const store = createStore(counterApp, preloadedState || initialState);
// Tell react-snap how to save Redux state
window.snapSaveState = () => ({
__PRELOADED_STATE__: store.getState()
});Caution: as of now, only basic "JSON" data types are supported: e.g. Date, Set, Map, and NaN won't be handled correctly (#54).
You can block all third-party requests with the following config:
"skipThirdPartyRequests": truereact-snap can capture all AJAX requests. It will store json requests in the domain in window.snapStore[<path>], where <path> is the path of the request.
Use "cacheAjaxRequests": true to enable this feature.
This feature can conflict with the browser cache. See #197 for details. You may want to disable cache in this case: "puppeteer": { "cache": false }.
By default, create-react-app uses index.html as a fallback:
navigateFallback: publicUrl + '/index.html',You need to change this to an un-prerendered version of index.html - 200.html, otherwise you will see index.html flash on other pages (if you have any). See Configure sw-precache without ejecting for more information.
Puppeteer (Headless Chrome) may fail due to sandboxing issues. To get around this, you may use:
"puppeteerArgs": ["--no-sandbox", "--disable-setuid-sandbox"]Read more about puppeteer troubleshooting.
"inlineCss": true sometimes causes problems in containers.
To run react-snap inside docker with Alpine, you might want to use a custom Chromium executable. See #93 and #132.
heroku buildpacks:add https://github.com/jontewks/puppeteer-heroku-buildpack.git
heroku buildpacks:add heroku/nodejs
heroku buildpacks:add https://github.com/heroku/heroku-buildpack-static.git
See this PR. At the moment of writing, Heroku doesn't support HTTP/2.
Semantic UI is defined over class substrings that contain spaces (e.g., "three column"). Sorting the class names, therefore, breaks the styling. To get around this, use the following configuration:
"minifyHtml": { "sortClassName": false }From version 1.17.0, sortClassName is false by default.
Once JS on the client is loaded, components initialized and your JSS styles are regenerated, it's a good time to remove server-side generated style tag in order to avoid side-effects
This basically means that JSS doesn't support rehydration. See #99 for a possible solutions.
See #135.
You can use navigator.userAgent == "ReactSnap" to do some checks in the app code while snapping—for example, if you use an absolute path for your API AJAX request. While crawling, however, you should request a specific host.
Example code:
const BASE_URL =
process.env.NODE_ENV == "production" && navigator.userAgent != "ReactSnap"
? "/"
: "http://xxx.yy/rest-api";See alternatives.
 |
 |
 |
|---|
Please provide a reproducible demo of a bug and steps to reproduce it. Thanks!
Tweet it, like it, share it, star it. Thank you.
You can also contribute to minimalcss, which is a big part of react-snap. Also, give it some stars.