Edgio
Edgio

Routing with EdgeJS

The @layer0/core package provides a JavaScript API for controlling routing and caching from your code base rather than a CDN web portal. Using this EdgeJS approach allows this vital routing logic to be properly tested, reviewed, and version controlled, just like the rest of your application code.

Using the Router, you can:

  • Proxy requests to upstream sites
  • Send redirects from the network edge
  • Render responses on the server using Next.js, Nuxt.js, Angular, or any other framework that supports server side rendering.
  • Alter request and response headers
  • Send synthetic responses
  • Configure multiple destinations for split testing

Configuration

You define routes for Edgio using the routes.js file.

Before continuing, if you have not already initialized your project with Edgio, do so using the instructions in Web CDN.

The routes.js file should export an instance of @layer0/core/router/Router:

JavaScript./routes.js
1const { Router } = require('@layer0/core/router')
2
3module.exports = new Router()

Declare Routes

Declare routes using the method corresponding to the HTTP method you want to match.

JavaScript./routes.js
1const { Router } = require('@layer0/core/router')
2
3module.exports = new Router().get('/some-path', ({ cache, proxy }) => {
4 // handle the request here
5})

All HTTP methods are available:

  • get
  • put
  • post
  • patch
  • delete
  • head

To match all methods, use match:

JavaScript./routes.js
1const { Router } = require('@layer0/core/router')
2
3module.exports = new Router().match('/some-path', ({ cache, proxy }) => {
4 // handle the request here
5})

Route Execution

When Edgio receives a request, it executes each route that matches the request in the order in which they are declared until one sends a response. The following methods return a response:

Multiple routes can therefore be executed for a given request. A common pattern is to add caching with one route and render the response with a later one using middleware. In the following example we cache then render a response with Next.js:

JavaScript
1const { Router } = require('@layer0/core/router')
2const { nextRoutes } = require('@layer0/next')
3
4// In this example a request to /products/1 will be cached by the first route, then served by the `nextRoutes` middleware
5new Router()
6 .get('/products/:id', ({ cache }) => {
7 cache({
8 edge: { maxAgeSeconds: 60 * 60, staleWhileRevalidateSeconds: 60 * 60 },
9 })
10 })
11 .use(nextRoutes)

Alter Requests and Responses

Edgio offers APIs to manipulate request and response headers and cookies. The APIs are:

OperationRequestUpstream ResponseResponse sent to Browser
Set headersetRequestHeadersetUpstreamResponseHeadersetResponseHeader
Add cookie*addUpstreamResponseCookieaddResponseCookie
Update headerupdateRequestHeaderupdateUpstreamResponseHeaderupdateResponseHeader
Update cookie*updateUpstreamResponseCookieupdateResponseCookie
Remove headerremoveRequestHeaderremoveUpstreamResponseHeaderremoveResponseHeader
Remove cookie*removeUpstreamResponseCookieremoveResponseCookie

* Adding, updating, or removing a request cookie can be achieved with updateRequestHeader applied to cookie header.

You can find detailed descriptions of these APIs in the @layer0/core documentation.

Embedded Values

You can inject values from the request or response into headers or cookies as template literals using the ${value} format. For example: setResponseHeader('original-request-path', '${path}') would add an original-request-path response header whose value is the request path.

ValueEmbedded valueDescription
HTTP method${method}The value of the HTTP method used for the request (e.g. GET)
URL${url}The complete URL path including any query strings (e.g. /search?query=docs). Protocol, hostname, and port are not included.
Path${path}The URL path excluding any query strings (e.g. /search)
Query string${query:<name>}The value of the <name> query string or empty if not available.
Request header${req:<name>}The value of the <name> request header or empty if not available.
Request cookie${req:cookie:<name>}The value of the <name> cookie in cookie request header or empty if not available.
Request named parameter${req:param:<name>}The value of the <name> param defined in the route or empty if not available.
Response header${res:<name>}The value of the <name> response header or empty if not available.

Route Pattern Syntax

The syntax for route paths is provided by path-to-regexp, which is the same library used by Express.

Named Parameters

Named parameters are defined by prefixing a colon to the parameter name (:foo).

JavaScript
1new Router().get('/:foo/:bar', res => {
2 /* ... */
3})

Please note: Parameter names must use “word characters” ([A-Za-z0-9_]).

Custom Matching Parameters

Parameters can have a custom regexp, which overrides the default match ([^/]+). For example, you can match digits or names in a path:

JavaScript
1new Router().get('/icon-:foo(\\d+).png', res => {
2 /* ... */
3})

Tip: Backslashes need to be escaped with another backslash in JavaScript strings.

Custom Prefix and Suffix

Parameters can be wrapped in {} to create custom prefixes or suffixes for your segment:

JavaScript
1new Router().get('/:attr1?{-:attr2}?{-:attr3}?', res => {
2 /* ... */
3})

Unnamed Parameters

It is possible to write an unnamed parameter that only consists of a regexp. It works the same the named parameter, except it will be numerically indexed:

JavaScript
1new Router().get('/:foo/(.*)', res => {
2 /* ... */
3})

Modifiers

Modifiers must be placed after the parameter (e.g. /:foo?, /(test)?, /:foo(test)?, or {-:foo(test)}?).

Optional

Parameters can be suffixed with a question mark (?) to make the parameter optional.

JavaScript
1new Router().get('/:foo/:bar?', res => {
2 /* ... */
3})

Tip: The prefix is also optional, escape the prefix \/ to make it required.

Zero or More

Parameters can be suffixed with an asterisk (*) to denote zero or more parameter matches.

JavaScript
1new Router().get('/:foo*', res => {
2 /* res.params.foo will be an array */
3})

The captured parameter value will be provided as an array.

One or More

Parameters can be suffixed with a plus sign (+) to denote one or more parameter matches.

JavaScript
1new Router().get('/:foo+', res => {
2 /* res.params.foo will be an array */
3})

The captured parameter value will be provided as an array.

Matching Method, Query Parameters, Cookies, and Headers

Match can either take a URL path, or an object which allows you to match based on method, query parameters, cookies, or request headers:

JavaScript
1router.match(
2 {
3 path: '/some-path', // value is route-pattern syntax
4 method: /GET|POST/i, // value is a regular expression
5 cookies: { currency: /^(usd)$/i }, // keys are cookie names, values are regular expressions
6 headers: { 'x-moov-device': /^desktop$/i }, // keys are header names, values are regular expressions
7 query: { page: /^(1|2|3)$/ }, // keys are query parameter names, values are regular expressions
8 },
9 () => {},
10)

Body Matching for POST requests

You can also match HTTP POST requests based on their request body content as in the following example:

JavaScript
1router.match(
2 {
3 body: { parse: 'json', criteria: { operationName: 'GetProducts' } }, // the body content will parsed as JSON and the parsed JSON matched against the presence of the criteria properties (in this case a GraphQL operation named 'GetProducts')
4 },
5 () => {},
6)

Currently the only body content supported is JSON. Body content is parsed as JSON and is matched against the presence of the fields specified in the criteria field. The POST Body Matching Criteria section below contains examples of using the criteria field.

Body matching can be combined with other match parameters such as headers and cookies. For example,

JavaScript
1router.match(
2 {
3 // Only matches GetProducts operations to the /graphql endpoint
4 // for logged in users
5 path: '/graphql',
6 cookies: { loginStatus: /^(loggedIn)$/i }, // loggedin users
7 body: { parse: 'json', criteria: { operationName: 'GetProducts' } },
8 },
9 () => {},
10)

Caching & POST Body Matching

When body matching is combined with cache in a route, the HTTP request body will automatically be used as the cache key. For example, the code below will cache GraphQL GetProducts queries using the entire request body as the cache key:

JavaScript
1router.match(
2 {
3 body: { parse: 'json', criteria: { operationName: 'GetProducts' } },
4 },
5 ({ cache }) => {
6 cache({
7 edge: {
8 maxAgeSeconds: 60 * 60,
9 staleWhileRevalidateSeconds: 60 * 60 * 24, // this way stale items can still be prefetched
10 },
11 })
12 },
13)

You can still add additional parameters to the cache key using the normal EdgeJS key property. For example, the code below will cache GraphQL GetProducts queries separately for each user based on their userID cookie and the HTTP body of the request.

JavaScript
1router.match(
2 {
3 body: { parse: 'json', criteria: { operationName: 'GetProducts' } },
4 },
5 ({ cache }) => {
6 cache({
7 edge: {
8 maxAgeSeconds: 60 * 60,
9 staleWhileRevalidateSeconds: 60 * 60 * 24, // this way stale items can still be prefetched
10 },
11 key: new CustomCacheKey().addCookie('userID'), // Split cache by userID
12 })
13 },
14)

POST Body Matching Criteria

The criteria property can be a string or regular expression.

For example, the router below,

JavaScript
1router.match(
2 {
3 body: { parse: 'json', criteria: { foo: 'bar' } },
4 },
5 () => {},
6)

would match an HTTP POST request body containing:

JavaScript
1{
2 "foo": "bar",
3 "bar": "foo"
4}

Regular Expression Criteria

Regular expressions can also be used as criteria. For example,

JavaScript
1router.match(
2 {
3 body: { parse: 'json', criteria: { operationName: /^Get/ } },
4 },
5 () => {},
6)

would match an HTTP POST body containing:

JavaScript
1{
2 "operationName": "GetShops",
3 "query": "...",
4 "variables": {}
5}

Nested JSON Criteria

You can also use a nested object to match a field at a specific location in the JSON. For example,

JavaScript
1router.match(
2 {
3 body: {
4 parse: 'json',
5 criteria: {
6 operation: {
7 name: 'GetShops',
8 },
9 },
10 },
11 },
12 () => {},
13)

would match an HTTP POST body containing:

JavaScript
1{
2 "operation": {
3 "name": "GetShops",
4 "query": "..."
5 }
6}

GraphQL Queries

The EdgeJS router provides a graphqlOperation method for matching GraphQL.

JavaScript
1router.graphqlOperation('GetProducts', res => {
2 /* Handle the POST for the GetProducts query specifically */
3})

By default, the graphqlOperation assumes your GraphQL endpoint is at /graphql. You can alter this behavior by using the path property as shown below:

JavaScript
1router.graphqlOperation({ path: '/api/graphql', name: 'GetProducts' }, res => {
2 /* Handle the POST for the GetProducts query specifically */
3})

Note that when the graphqlOperation function is used, the HTTP request body will automatically be included in the cache key.

The graphqlOperation function is provided to simplify matching of common GraphQL scenarios. For complex GraphQL matching (such as authenticated data), you can use the generic Body Matching for POST requests feature.

See the guide on Implementing GraphQL Routing in your project.

Request Handling

The second argument to routes is a function that receives a ResponseWriter and uses it to send a response. Using ResponseWriter you can:

  • Proxy a backend configured in layer0.config.js
  • Serve a static file
  • Send a redirect
  • Send a synthetic response
  • Cache the response at edge and in the browser
  • Manipulate request and response headers

See the API Docs for Response Writer

Blocking Search Engine Crawlers

If you need to block all search engine bot traffic to specific environments (such as your default or staging environment), the easiest way is to include the x-robots-tag header with the same directives you would otherwise set in a meta tag.

To block search engine traffic for Edgio edge links and permalinks, you can use the built-in .noIndexPermalink() call on the router:

JavaScript
1router.noIndexPermalink()

This will match requests with the host header matching /layer0.link|layer0-perma.link/ and set a response header of x-robots-tag: noindex.

Additionally, you can customize this to block traffic to development or staging websites based on the host header of the request:

JavaScript
1router
2 .noIndexPermalink()
3 .get(
4 {
5 headers: {
6 // Regex to catch multiple hostnames
7 host: /dev.example.com|staging.example.com/,
8 },
9 },
10 ({ setResponseHeader }) => {
11 setResponseHeader('x-robots-tag', 'noindex')
12 },
13 )

Full Example

This example shows typical usage of @layer0/core, including serving a service worker, next.js routes (vanity and conventional routes), and falling back to a legacy backend.

JavaScript./routes.js
1const { Router } = require('@layer0/core/router')
2
3module.exports = new Router()
4 // adds `x-robots-tag: noindex` response header to Edgio
5 // edge links and permalinks to prevent bot indexing
6 .noIndexPermalink()
7 .get('/service-worker.js', ({ serviceWorker }) => {
8 // serve the service worker built by webpack
9 serviceWorker('dist/service-worker.js')
10 })
11 .get('/p/:productId', ({ cache }) => {
12 // cache products for one hour at edge and using the service worker
13 cache({
14 edge: {
15 maxAgeSeconds: 60 * 60,
16 staleWhileRevalidateSeconds: 60 * 60,
17 },
18 browser: {
19 maxAgeSeconds: 0,
20 serviceWorkerSeconds: 60 * 60,
21 },
22 })
23 proxy('origin')
24 })
25 .fallback(({ proxy }) => {
26 // serve all unmatched URLs from the origin backend configured in layer0.config.js
27 proxy('origin')
28 })

Errors Handling

You can use the router’s catch method to return specific content when the request results in an error status (For example, a status code of 537). Using catch, you can also alter the statusCode and response on the edge before issuing a response to the user.

JavaScript
1router.catch(RegExp | string | number, (routeHandler: Function))

Examples

For example, to issue a custom error page when the origin returns any 5xx status code:

JavaScriptroutes.js
1const { Router } = require('@layer0/core/router')
2
3module.exports = new Router()
4 // Example route that returns with a 5xx error status code
5 .get('/failing-route', ({ proxy }) => {
6 proxy('broken-origin')
7 })
8 // So let's assume that the route above returns 5xx, so instead of rendering
9 // the broken-origin response we can alter that by specifing .catch
10 .catch(/5[0-9][0-9]/, ({ serveStatic }) => {
11 // The file below is present at the root of the directory
12 serveStatic('customized-error-page.html', { statusCode: 502 })
13 })

The .catch method allows the edge router to render a response based on the result preceeding routes. So in the example above whenever we receive a 5xx, we respond with customized-error-page.html from the application’s root directory, and change the status code to 502.

  • Your catch callback is provided a ResponseWriter instance. You can use any ResponseWriter method except proxy inside .catch.
  • We highly recommend keeping catch routes simple. Serve responses using serveStatic instead of send to minimize the size of the edge bundle.

Environment Edge Redirects

In addition to sending redirects at the edge within the router configuration, this can also be configured at the environment level within the Edgio Developer Console.

Under <Your Environment> Configuration, click Edit to draft a new configuration. Scroll down to the Redirects section:

redirects

Click Add A Redirect to configure the path or host you wish to redirect to:

add redirect

Note: you will need to activate and redeploy your site for this change to take effect.