JavaScriptServices/src/Microsoft.AspNetCore.SpaServices

Microsoft.AspNetCore.SpaServices

If you're building an ASP.NET Core application, and want to use Angular 2, React, Knockout, or another single-page app (SPA) framework, this NuGet package contains useful infrastructure for you.

This package enables:

• Server-side prerendering for universal (a.k.a. isomorphic) applications, where your Angular 2 / React / etc. components are first rendered on the server, and then transferred to the client where execution continues
• Webpack middleware so that, during development, any webpack-built resources will be generated on demand, without you having to run webpack manually or compile files to disk
• Hot module replacement so that, during development, your code and markup changes will be pushed to your browser and updated in the running application automatically, without even needing to reload the page
• Routing helpers for integrating server-side routing with client-side routing

Behind the scenes, it uses the Microsoft.AspNetCore.NodeServices package as a fast and robust way to invoke Node.js-hosted code from ASP.NET Core at runtime.

Requirements

• Node.js
  • To test this is installed and can be found, run node -v on a command line
  • Note: If you're deploying to an Azure web site, you don't need to do anything here - Node is already installed and available in the server environments
• .NET Core, version 1.0 RC2 or later

Installation into existing projects

• Add Microsoft.AspNetCore.SpaServices to the dependencies list in your project.json file

• Run dotnet restore (or if you use Visual Studio, just wait a moment - it will restore dependencies automatically)

• Install supporting NPM packages for the features you'll be using:

  • For server-side prerendering, install aspnet-prerendering
  • For server-side prerendering with Webpack build support, also install aspnet-webpack
  • For webpack dev middleware, install aspnet-webpack
  • For webpack dev middleware with hot module replacement, also install webpack-hot-middleware
  • For webpack dev middleware with React hot module replacement, also install aspnet-webpack-react

  For example, run npm install --save aspnet-prerendering aspnet-webpack to install aspnet-prerendering and aspnet-webpack.

Creating entirely new projects

If you're starting from scratch, you might prefer to use the aspnetcore-spa Yeoman generator to get a ready-to-go starting point using your choice of client-side framework. This includes Microsoft.AspNetCore.SpaServices along with everything configured for webpack middleware, server-side prerendering, etc.

See: Getting started with the aspnetcore-spa generator

Server-side prerendering

The SpaServices package isn't tied to any particular client-side framework, and it doesn't force you to set up your client-side application in any one particular style. So, SpaServices doesn't contain hard-coded logic for rendering Angular 2 / React / etc. components.

Instead, what SpaServices offers is ASP.NET Core APIs that know how to invoke a JavaScript function that you supply, passing through context information that you'll need for server-side prerendering, and then injects the resulting HTML string into your rendered page. In this document, you'll find examples of setting this up to render Angular 2 and React components.

1. Enable the asp-prerender-* tag helpers

Make sure you've installed the Microsoft.AspNetCore.SpaServices NuGet package and the aspnet-prerendering NPM package. Together these contain the server-side and client-side library code you'll need.

Now go to your Views/_ViewImports.cshtml file, and add the following line:

@addTagHelper "*, Microsoft.AspNetCore.SpaServices"

2. Use asp-prerender-* in a view

Choose a place in one of your MVC views where you want to prerender a SPA component. For example, open Views/Home/Index.cshtml, and add markup like the following:

<div id="my-spa" asp-prerender-module="ClientApp/boot-server"></div>

If you run your application now, and browse to whatever page renders the view you just edited, you should get an error similar to the following (assuming you're running in Development mode so you can see the error information): Error: Cannot find module 'some/directory/ClientApp/boot-server'. You've told the prerendering tag helper to execute code from a JavaScript module called boot-server, but haven't yet supplied any such module!

3. Supplying JavaScript code to perform prerendering

Create a JavaScript file at the path matching the asp-prerender-module value you specified above. In this example, that means creating a folder called ClientApp at the root of your project, and creating a file inside it called boot-server.js. Try putting the following into it:

module.exports = function(params) {
    return new Promise(function (resolve, reject) {
        var result = '<h1>Hello world!</h1>'
            + '<p>Current time in Node is: ' + new Date() + '</p>'
            + '<p>Request path is: ' + params.location.path + '</p>'
            + '<p>Absolute URL is: ' + params.absoluteUrl + '</p>';

        resolve({ html: result });
    });
};

If you try running your app now, you should see the HTML snippet generated by your JavaScript getting injected into your page.

As you can see, your JavaScript code receives context information (such as the URL being requested), and returns a Promise so that it can asynchronously supply the markup to be injected into the page. You can put whatever logic you like here, but typically you'll want to execute a component from your Angular 2 / React / etc. application.

Passing data from server-side to client-side code

If, as well as returning HTML, you also want to pass some contextual data from your server-side code to your client-side code, you can supply a globals object alongside the initial html, e.g.:

resolve({
    html: result,
    globals: {
        albumsList: someDataHere,
        userData: someMoreDataHere
    }
});

When the aspnet-prerender-* tag helper emits this result into the document, as well as injecting the html string, it will also emit code that populates window.albumsList and window.userData with JSON-serialized copies of the objects you passed.

This can be useful if, for example, you want to avoid loading the same data twice (once on the server and once on the client).

4. Enabling webpack build tooling

Of course, rather than writing your boot-server module and your entire SPA in plain ES5 JavaScript, it's quite likely that you'll want to write your client-side code in TypeScript or at least ES2015 code. To enable this, you can either:

• Set up some build tool such as Babel to transpile to ES5, and always remember to run this to generate plain ES5 .js files before you run your application
• Or, more conveniently, use webpack along with the asp-prerender-webpack-config attribute so that Microsoft.AspNetCore.SpaServices can automatically build your boot module and the SPA code that it references. Then there's no need for .js files even to be written to disk - the build process is all dynamic and in memory.

To enable webpack builds for your server-side prerendering, amend your MVC view to specify the location of your webpack configuration file using an asp-prerender-webpack-config attribute, e.g.:

<div id="my-spa" asp-prerender-module="ClientApp/boot-server"
                 asp-prerender-webpack-config="webpack.config.js"></div>

You'll also need to install the NPM module aspnet-webpack if you don't have it already, e.g.:

npm install --save aspnet-webpack

This includes webpack as well as the server-side code needed to invoke it from ASP.NET Core at runtime.

Now, assuming you have a working webpack configuration at webpack.config.js, your boot module and SPA code will dynamically be built using webpack.

Example: Configuring webpack to build TypeScript

Let's say you want to write your boot module and SPA code in TypeScript. First ensure that aspnet-webpack is installed, along with the libraries needed for TypeScript compilation:

npm install --save aspnet-webpack ts-loader typescript

Next, create a file webpack.config.js at the root of your project, containing:

module.exports = {
    resolve: { extensions: [ '', '.js', '.ts' ] },
    module: {
        loaders: [
            { test: /\.ts$/, loader: 'ts-loader' }
        ]
    }
};

This tells webpack that it should compile .ts files using TypeScript, and that when looking for modules by name (e.g., boot-server), it should also find files with .js and .ts extensions.

Now you can delete ClientApp/boot-server.js, and in its place, create ClientApp/boot-server.ts, containing the TypeScript equivalent of what you had before:

export default function (params: any): Promise<{ html: string}> {
    return new Promise((resolve, reject) => {
        const html = `
            <h1>Hello world!</h1>
            <p>Current time in Node is: ${ new Date() }</p>
            <p>Request path is: ${ params.location.path }</p>
            <p>Absolute URL is: ${ params.absoluteUrl }</p>`;

        resolve({ html });
    });
}

Finally, you can tell SpaServices to use the Webpack environment you've just set up. In your MVC view where you use aspnet-prerender-module, also specify aspnet-prerender-webpack-config:

<div id="my-spa" asp-prerender-module="ClientApp/boot-server"
                 asp-prerender-webpack-config="webpack.config.js"></div>

Now your boot-server.ts code should get executed when your ASP.NET Core page is rendered, and since it's TypeScript, it can of course reference any other TypeScript modules, which means your entire SPA can be written in TypeScript and executed on the server.

Webpack is a broad and powerful tool and can do far more than just invoke the TypeScript compiler. To learn more, see the webpack website.

5(a). Prerendering Angular 2 components

If you're building an Angular 2 application, you can run your components on the server inside your boot-server.ts file so they will be injected into the resulting web page.

First install the NPM package angular2-universal - this contains infrastructure for executing Angular 2 components inside Node.js:

npm install --save angular2-universal

Now you can use the angular2-universal APIs from your boot-server.ts TypeScript module to execute your Angular 2 component on the server. The code needed for this is fairly complex, but that's unavoidable because Angular 2 supports so many different ways of being configured, and you need to provide wiring for whatever combination of DI modules you're using.

You can find an example boot-server.ts that renders arbitrary Angular 2 components here. If you use this with your own application, you might need to edit the serverBindings array to reference any other DI services that your Angular 2 component depends on.

The easiest way to get started with Angular 2 server-side rendering on ASP.NET Core is to use the aspnetcore-spa generator, which creates a ready-made working starting point.

5(b). Prerendering React components

React components can be executed synchronously on the server quite easily, although asynchronous execution is tricker as described below.

Setting up client-side React code

Let's say you want to write a React component in ES2015 code. You might install the NPM modules react react-dom babel-loader babel-preset-react babel-preset-es2015, and then prepare Webpack to build .jsx files by creating webpack.config.js in your project root, containing:

var path = require('path');

module.exports = {
    resolve: { extensions: [ '', '.js', '.jsx' ] },
    module: {
        loaders: [
            { test: /\.jsx?$/, loader: 'babel-loader' }
        ]
    },
    entry: {
        main: ['./ClientApp/react-app.jsx'],
    },
    output: {
        path: path.join(__dirname, 'wwwroot', 'dist'),
        filename: '[name].js'
    },
};

You will also need a .babelrc file in your project root, containing:

{
    "presets": ["es2015", "react"]
}

This is enough to be able to build ES2015 .jsx files via Webpack. Now you could implement a simple React component, for example the following at ClientApp/react-app.jsx:

import * as React from 'react';

export class HelloMessage extends React.Component
{
    render() {
        return <h1>Hello {this.props.message}!</h1>;
    }
}

... and the following code to run it in a browser at ClientApp/boot-client.jsx:

import * as React from 'react';
import * as ReactDOM from 'react-dom';
import { HelloMessage } from './react-app';

ReactDOM.render(<HelloMessage message="World" />, document.getElementById('my-spa'));

At this stage, run webpack on the command line to build wwwroot/dist/main.js. Or, to avoid having to do this manually, you could use the SpaServices package to enable Webpack dev middleware.

You can now run your React code on the client by adding the following to one of your MVC views:

<div id="my-spa"></div>
<script src="/dist/main.js"></script>

Running React code on the server

Now you have React code being built using Webpack, you can enable server-side prerendering using the aspnet-prerender-* tag helpers as follows:

<div id="my-spa" asp-prerender-module="ClientApp/boot-server"
                 asp-prerender-webpack-config="webpack.config.js"></div>

... along with the following boot module at ClientApp/boot-server.jsx:

import * as React from 'react';
import { renderToString } from 'react-dom/server';
import { HelloMessage } from './react-app';

export default function (params) {
    return new Promise((resolve, reject) => {
        resolve({
            html: renderToString(<HelloMessage message="from the server" />)
        });
    });
}

Now you should find that your React app is rendered in the page even before any JavaScript is loaded in the browser (or even if JavaScript is disabled in the browser).

Realistic React apps and Redux

The above example is extremely simple - it doesn't use react-router, and it doesn't load any data asynchronously. Real applications are likely to do both of these.

For an example server-side boot module that knows how to evaluate react-router routes and render the correct React component, see this example.

Supporting asynchronous data loading involves more considerations. Unlike Angular 2 applications that run asynchronously on the server and freely overwrite server-generated markup with client-generated markup, React strictly wants to run synchronously on the server and always produce the same markup on the server as it does on the client.

To make this work, you most likely need some way to know in advance what data your React components will need to use, load it separately from those components, and have some way of transferring information about the loaded data from server to client. If you try to implement this in a generalized way, you'll end up reinventing something like the Flux/Redux pattern.

To avoid inventing your own incomplete version of Flux/Redux, you probably should just use Redux. This is at first a very unfamiliar and tricky-looking abstraction, but does solve all the problems around server-side execution of React apps. To get a working starting point for an ASP.NET Core site with React+Redux on the client (and server-side prerendering), see the aspnetcore-spa generator.

Webpack dev middleware

If you're using webpack, the webpack dev middleware feature included in Microsoft.AspNetCore.SpaServices will streamline your development process. It intercepts requests that would match files built by webpack, and dynamically builds those files on demand. They don't need to be written to disk - they are just held in memory and served directly to the browser.

Benefits:

• You don't have to run webpack manually or set up any file watchers
• The browser is always guaranteed to receive up-to-date built output
• The built artifacts are normally served instantly or at least extremely quickly, because internally, an instance of webpack stays active and has partial compilation states pre-cached in memory

It lets you work as if the browser natively understands whatever file types you are working with (e.g., TypeScript, SASS), because it's as if there's no build process to wait for.

Enabling webpack dev middleware

After installing the Microsoft.AspNetCore.SpaServices NuGet package and the aspnet-webpack NPM package, go to your Startup.cs file, and before your call to UseStaticFiles, add the following:

if (env.IsDevelopment()) {
    app.UseWebpackDevMiddleware();
}

// You call to app.UseStaticFiles(); should be here

You will also need to edit your webpack configuration at webpack.config.js. Since UseWebpackDevMiddleware needs to know which incoming requests to intercept, specify a publicPath value on your output, for example:

module.exports = {
    // ... rest of your webpack config is here ...

    output: {
        path: path.join(__dirname, 'wwwroot', 'dist'),
        publicPath: '/dist',
        filename: '[name].js'
    },
};

Now, assuming you're running in development mode, any requests for files under /dist will be intercepted and served using Webpack dev middleware.

This is for development time only, not for production use (hence the env.IsDevelopment() check in the code above). While you could technically remove that check and serve your content in production through the webpack middleware, it's hard to think of a good reason for doing so. For best performance, it makes sense to prebuild your client-side resources so they can be served directly from disk with no build middleware. If you use the aspnetcore-spa generator, you'll get a site that produces optimised static builds for production, while also supporting webpack dev middleware at development time.

Webpack Hot Module Replacement

For an even more streamlined development experience, you can enhance webpack dev middleware by enabling Hot Module Replacement (HMR) support. This watches for any changes you make to source files on disk (e.g., .ts/.html/.sass/etc. files), and automatically rebuilds them and pushes the result into your browser window, without even needing to reload the page.

This is not the same as a simple live-reload mechanism. It does not reload the page; it replaces code or markup directly in place. This is better, because it does not interfere with any state your SPA might have in memory, or any debugging session you have in progress.

Typically, when you change a source file, the effects appear in your local browser window in under 2 seconds, even when your overall application is large. This is superbly productive, especially in multi-monitor setups. If you cause a build error (e.g., a syntax error), details of the error will appear in your browser window. When you fix it, your application will reappear, without having lost its in-memory state.

Enabling Hot Module Replacement

First ensure you already have a working Webpack dev middleware setup. Then, install the webpack-hot-middleware NPM module:

npm install --save webpack-hot-middleware

At the top of your Startup.cs file, add the following namespace reference:

using Microsoft.AspNetCore.SpaServices.Webpack;

Now amend your call to UseWebpackDevMiddleware as follows:

app.UseWebpackDevMiddleware(new WebpackDevMiddlewareOptions {
    HotModuleReplacement = true
});

Also, to work around a temporary issue in SpaServices, you must ensure that your Webpack config includes a plugins array, even if it's empty. For example, in webpack.config.js:

module.exports = {
    // ... rest of your webpack config is here ...

    plugins: [
        // Put webpack plugins here if needed, or leave it as an empty array if not
    ]
};

Now when you load your application in a browser, you should see a message like the following in your browser console:

[HMR] connected

If you edit any of your source files that get built by webpack, the result will automatically be pushed into the browser. As for what the browser does with these updates - that's a matter of how you configure it - see below.

Enabling hot replacement for React components

Webpack has built-in support for updating React components in place. To enable this, amend your UseWebpackDevMiddleware call further as follows:

app.UseWebpackDevMiddleware(new WebpackDevMiddlewareOptions {
    HotModuleReplacement = true,
    ReactHotModuleReplacement = true
});

Also, install the NPM module aspnet-webpack-react, e.g.:

npm install --save aspnet-webpack-react

Now if you edit any React component (e.g., in .jsx or .tsx files), the updated component will be injected into the running application, and will even preserve its in-memory state.

Note: In you webpack config, be sure that your React components are loaded using babel-loader (and not just directly using babel or ts-loader), because babel-loader is where the HMR instrumentation is injected. For an example of HMR for React components built with TypeScript, see the aspnetcore-spa generator.

Enabling hot replacement for other module types

Webpack has built-in HMR support for various types of module, such as styles and React components as described above. But to support HMR for other code modules, you need to add a small block of code that calls module.hot.accept to receive the updated module and update the running application.

This is documented in detail on the Webpack site. Or to get a working HMR-enabled ASP.NET Core site with Angular 2, React, React+Redux, or Knockout, you can use the aspnetcore-spa generator.

Routing helper: MapSpaFallbackRoute

In most single-page applications, you'll want client-side routing as well as your server-side routing. Most of the time, the two routing systems work independently without interfering. However, there is one case where things get challenging: identifying 404s.

If a request arrives for /some/page, and it doesn't match any server-side route, it's likely that you want to return HTML that starts up your client-side application, which probably understands the route /some/page. But if a request arrives for /images/user-512.png, and it doesn't match any server-side route or static file, it's not likely that your client-side application would handle it - you probably want to return a 404.

To help distinguish between these cases, the Microsoft.AspNetCore.SpaServices NuGet package includes a routing helper, MapSpaFallbackRoute. For example, in your Startup.cs file's Configure method, you might add:

    app.UseStaticFiles();

    app.UseMvc(routes =>
    {
        routes.MapRoute(
            name: "default",
            template: "{controller=Home}/{action=Index}/{id?}");

        routes.MapSpaFallbackRoute(
            name: "spa-fallback",
            defaults: new { controller = "Home", action = "Index" });
    });

Since UseStaticFiles goes first, any requests that actually match physical files under wwwroot will be handled by serving that static file.

Since the default server-side MVC route goes next, any requests that match existing controller/action pairs will be handled by invoking that action.

Then, since MapSpaFallbackRoute is last, any other requests that don't appear to be for static files will be served by invoking the Index action on HomeController. This action's view should serve your client-side application code, allowing the client-side routing system to handle whatever URL has been requested.

Any requests that do appear to be for static files (i.e., those that end with filename extensions), will not be handled by MapSpaFallbackRoute, and so will end up as 404s.

This is not a perfect solution to the problem of identifying 404s, because for example MapSpaFallbackRoute will not match requests for /users/albert.einstein, because it appears to contain a filename extension (.einstein). If you need your SPA to handle routes like that, then don't use MapSpaFallbackRoute - just use a regular MVC catch-all route. But then beware that requests for unknown static files will result in your client-side app being rendered.