Vert.x-Web

To use vert.x web, add the following dependency to the dependencies section of your build descriptor:

Vert.x-Web uses and exposes the API from Vert.x core, so it’s well worth getting familiar with the basic concepts of writing HTTP servers using Vert.x core, if you’re not already.

The Vert.x core HTTP documentation goes into a lot of detail on this.

Here’s a hello world web server written using Vert.x core. At this point there is no Vert.x-Web involved:

We create an HTTP server instance, and we set a request handler on it. The request handler will be called whenever a request arrives on the server.

When that happens we are just going to set the content type to text/plain, and write Hello World! and end the response.

We then tell the server to listen at port 8080 (default host is localhost).

You can run this, and point your browser at http://localhost:8080 to verify that it works as expected.

Here’s the 10000 foot view:

A Router is one of the core concepts of Vert.x-Web. It’s an object which maintains zero or more Routes.

A router will pause the incoming HttpServerRequest, to ensure that the request body or any protocol upgrades are not lost. Second, it will find the first matching route for that request, and passes the request to that route.

The route can have a handler associated with it, which then receives the request. You then do something with the request, and then, either end it or pass it to the next matching handler.

Here’s a simple router example:

It basically does the same thing as the Vert.x Core HTTP server hello world example from the previous section, but this time using Vert.x-Web.

We create an HTTP server as before, then we create a router. Once we’ve done that we create a simple route with no matching criteria so it will match all requests that arrive on the server.

We then specify a handler for that route. That handler will be called for all requests that arrive on the server.

The object that gets passed into the handler is a RoutingContext - this contains the standard Vert.x HttpServerRequest and HttpServerResponse but also various other useful stuff that makes working with Vert.x-Web simpler.

For every request that is routed there is a unique routing context instance, and the same instance is passed to all handlers for that request.

Once we’ve set up the handler, we set the request handler of the HTTP server to pass all incoming requests to handle.

So, that’s the basics. Now we’ll look at things in more detail:

When Vert.x-Web decides to route a request to a matching route, it calls the handler of the route passing in an instance of RoutingContext. A route can have different handlers, that you can append using handler

If you don’t end the response in your handler, you should call next so another matching route can handle the request (if any).

You don’t have to call next before the handler has finished executing. You can do this some time later, if you want:

In the above example route1 is written to the response, then 5 seconds later route2 is written to the response, then 5 seconds later route3 is written to the response and the response is ended.

Note, all this happens without any thread blocking.

Handlers are quite powerful as they allow you to build quite complex applications. For simple responses, for example, returning an asynchronous response from a vert.x API directly, the router includes a shortcut to a handler that ensures that:

However you can also use it for non JSON responses if the function supplied calls either write or end:

Sometimes, you might have to do something in a handler that might block the event loop for some time, e.g. call a legacy blocking API or do some intensive calculation.

You can’t do that in a normal handler, so we provide the ability to set blocking handlers on a route.

A blocking handler looks just like a normal handler but it’s called by Vert.x using a thread from the worker pool not using an event loop.

You set a blocking handler on a route with blockingHandler. Here’s an example:

By default, any blocking handlers executed on the same context (e.g. the same verticle instance) are ordered - this means the next one won’t be executed until the previous one has completed. If you don’t care about orderering and don’t mind your blocking handlers executing in parallel you can set the blocking handler specifying ordered as false using blockingHandler.

Note, if you need to process multipart form data from a blocking handler, you MUST use a non-blocking handler FIRST in order to call setExpectMultipart(true). Here is an example:

A route can be set-up to match the path from the request URI. In this case it will match any request which has a path that’s the same as the specified path.

In the following example the handler will be called for a request /some/path/. We also ignore trailing slashes so it will be called for paths /some/path and /some/path// too:

Often you want to route all requests that begin with a certain path. You could use a regex to do this, but a simply way is to use an asterisk * at the end of the path when declaring the route path.

In the following example the handler will be called for any request with a URI path that starts with /some/path/.

For example /some/path/foo.html and /some/path/otherdir/blah.css would both match.

With any path it can also be specified when creating the route:

It’s possible to match paths using placeholders for parameters which are then available in the context pathParam.

Here’s an example

The placeholders consist of : followed by the parameter name. Parameter names consist of any alphabetic character, numeric character or underscore. In certain cases this is a bit limited, so users may switch to an extended name rule which includes 2 extra characters - and $. The extended param rule is enabled as a system property:

In the above example, if a POST request is made to path: /catalogue/products/tools/drill123/ then the route will match and productType will receive the value tools and productID will receive the value drill123.

A parameter is not required to be a path segment. For example path parameters like the following ones are valid too:

Note: You can also capture as path param .

Regular expressions can also be used to match URI paths in routes.

Alternatively the regex can be specified when creating the route:

You can also capture path parameters when using regular expressions, here’s an example:

In the above example, if a request is made to path: /tools/drill123/ then the route will match and productType will receive the value tools and productID will receive the value drill123.

Captures are denoted in regular expressions with capture groups (i.e. surrounding the capture with round brackets)

Using int index param names might be troublesome in some cases. It’s possible to use named capture groups in the regex path.

In the example above, named capture groups are mapped to path parameters of the same name as the group.

Additionally, you can still access group parameters as you would with normal groups (i.e. params0, params1…​)

By default a route will match all HTTP methods.

If you want a route to only match for a specific HTTP method you can use method

Or you can specify this with a path when creating the route:

If you want to route for a specific HTTP method you can also use the methods such as get, post and put named after the HTTP method name. For example:

If you want to specify a route will match for more than HTTP method you can call method multiple times:

If you are creating an application that require custom HTTP verbs, for example, an WebDav server, then you can specify custom verbs like:

By default routes are matched in the order they are added to the router.

When a request arrives the router will step through each route and check if it matches, if it matches then the handler for that route will be called.

If the handler subsequently calls next the handler for the next matching route (if any) will be called. And so on.

Here’s an example to illustrate this:

In the above example the response will contain:

As the routes have been called in that order for any request that starts with /some/path.

If you want to override the default ordering for routes, you can do so using order, specifying an integer value.

Routes are assigned an order at creation time corresponding to the order in which they were added to the router, with the first route numbered 0, the second route numbered 1, and so on.

By specifying an order for the route you can override the default ordering. Order can also be negative, e.g. if you want to ensure a route is evaluated before route number 0.

Let’s change the ordering of route2 so it runs before route1:

then the response will now contain:

If two matching routes have the same value of order, then they will be called in the order they were added.

You can also specify that a route is handled last, with last

Note: Route order can be specified only before you configure an handler!

You can specify that a route will match against matching request MIME types using consumes.

In this case, the request will contain a content-type header specifying the MIME type of the request body. This will be matched against the value specified in consumes.

Basically, consumes is describing which MIME types the handler can consume.

Matching can be done on exact MIME type matches:

Multiple exact matches can also be specified:

Matching on wildcards for the sub-type is supported:

And you can also match on the top level type

If you don’t specify a / in the consumers, it will assume you meant the sub-type.

The HTTP accept header is used to signify which MIME types of the response are acceptable to the client.

An accept header can have multiple MIME types separated by ‘,’.

MIME types can also have a q value appended to them* which signifies a weighting to apply if more than one response MIME type is available matching the accept header. The q value is a number between 0 and 1.0. If omitted it defaults to 1.0.

For example, the following accept header signifies the client will accept a MIME type of only text/plain:

With the following the client will accept text/plain or text/html with no preference.

With the following the client will accept text/plain or text/html but prefers text/html as it has a higher q value (the default value is q=1.0)

If the server can provide both text/plain and text/html it should provide the text/html in this case.

By using produces you define which MIME type(s) the route produces, e.g. the following handler produces a response with MIME type application/json.

In this case the route will match with any request with an accept header that matches application/json.

Here are some examples of accept headers that will match:

You can also mark your route as producing more than one MIME type. If this is the case, then you use getAcceptableContentType to find out the actual MIME type that was accepted.

In the above example, if you sent a request with the following accept header:

Then the route would match and acceptableContentType would contain text/html as both are acceptable but that has a higher q value.

You can configure that a Route will match against the request hostname.

Request are checked against the Host header to a match and patterns allow the usage of wildcards, as for example .vertx.io or fully domain names as www.vertx.io.

You can combine all the above routing criteria in many different ways, for example:

You can disable a route with disable. A disabled route will be ignored when matching.

You can re-enable a disabled route with enable

Your application may be behind a proxy server, HAProxy for example. When working under this setup accessing the client connection details will not properly return the expected results. For example the client host ip address will be the proxy server ip address, not the client’s one.

In order to get the right connection information, a special header Forward has been standardized to include the right information. However this standard is not very old, so many proxies out there have been using other headers that usually start with the prefix: X-Forward. Vert.x web allows the usage and parsing of these headers but not by default.

The reason why these headers are disabled by default is to prevent malicious applications to forge their origin and hide where they are really coming from.

As previously mentioned forward is disabled by default, to enable you should use:

Same rule applies to explicitly disable the parsing of the headers:

To read more about the format of the header format, please consult:

Behind the scenes, what this feature does is changing the following values of your connection (either HTTP or WebSocket):

You can use the context data in the RoutingContext to maintain any data that you want to share between handlers for the lifetime of the request.

Here’s an example where one handler sets some data in the context data and a subsequent handler retrieves it:

You can use the put to put any object, and get to retrieve any object from the context data.

A request sent to path /some/path/other will match both routes.

Alternatively you can access the entire context data map with data.

While the context allows you to store data during the request-response lifecycle, there are times it is important to have runtime metadata available. For example, to build API documentation, or to hold specific configuration for a given route.

The metadata functionality works similarly to the context data. You can either access the holding Map or use the dedicated getter and setter defined both on the Router and Route interfaces:

While the routing context will allow you to access the underlying request and response objects, sometimes it will be more productive if a few shortcuts would be present to help with common tasks. A few helpers are present in the context to facilitate with this task.

Serve an "attachment", an attachment is a response that will trigger the browser to open the response on the OS application configured to handle a specific mime type. Imagine you’re generating a PDF:

Perform a redirect to a different page or host. One example is to redirect to an HTTPS variant of the application:

Send a JSON response to the client:

Simple content type check:

Verify if a request is "fresh" with respect to the cache headers and the current values of last modified/ etag.

And a few other simple self explanatory shortcuts:

Until now all routing mechanism allow you to handle your requests in a sequential way, however there might be times where you will want to go back. Since the context does not expose any information about the previous or next handler, mostly because this information is dynamic there is a way to restart the whole routing from the start of the current Router.

So from the code you can see that if a request arrives at /some/path if first add a value to the context, then moves to the next handler that re routes the request to /some/path/B which terminates the request.

You can reroute based on a new path or based on a new path and method. Note however that rerouting based on method might introduce security issues since for example a usually safe GET request can become a DELETE.

Reroute is also allowed on the failure handler, however due to the nature of re router when called the current status code and failure reason are reset. In order the rerouted handler should generate the correct status code if needed, for example:

It should be clear that reroute works on paths, so if you need to preserve and or add state across reroutes, one should use the RoutingContext object. For example you want to reroute to a new path with a extra parameter:

Reroute will re-parse the query params too. Be aware that previously query params will be discarded. The method will also silently discard and ignore any html fragment from the path. This is to keep the semantics of reroute consistent between a regular request and a re route.

If more information is required to be passed to the new request, it should use the context that is preserved all the lifetime of the HTTP transaction.

Sometimes if you have a lot of handlers it can make sense to split them up into multiple routers. This is also useful if you want to reuse a set of handlers in a different application, rooted at a different path root.

To do this you can mount a router at a mount point in another router. The router that is mounted is called a sub-router. Sub routers can mount other sub routers so you can have several levels of sub-routers if you like.

Let’s look at a simple example of a sub-router mounted with another router.

This sub-router will maintain the set of handlers that corresponds to a simple fictional REST API. We will mount that on another router. The full implementation of the REST API is not shown.

Here’s the sub-router:

If this router was used as a top level router, then GET/PUT/DELETE requests to urls like /products/product1234 would invoke the API.

However, let’s say we already have a web-site as described by another router:

We can now mount the sub router on the main router, against a mount point, in this case /productsAPI

This means the REST API is now accessible via paths like: /productsAPI/products/product1234.

There are a couple of rules that must be fulfilled before sub routers can be used:

Validation happens at the time the router is added to the http server. This means that you cannot get any validation error during the build time because of the dynamic nature of sub routers. They depend on the context to be validated.

Vert.x Web parses the Accept-Language header and provides some helper methods to identify which is the preferred locale for a client or the sorted list of preferred locales by quality.

The main method acceptableLanguages will return the ordered list of locales the user understands, if you’re only interested in the user prefered locale then the helper: preferredLanguage will return the 1st element of the list or null if no locale was provided by the user.

If no routes match for any particular request, Vert.x-Web will signal an error depending on match failure:

You can manually manage those failures using errorHandler

As well as setting handlers to handle requests you can also set handlers to handle failures in routing.

Failure handlers are used with the exact same route matching criteria that you use with normal handlers.

For example you can provide a failure handler that will only handle failures on certain paths, or for certain HTTP methods.

This allows you to set different failure handlers for different parts of your application.

Here’s an example failure handler that will only be called for failure that occur when routing to GET requests to paths that start with /somepath/:

Failure routing will occur if a handler throws an exception, or if a handler calls fail specifying an HTTP status code to deliberately signal a failure.

If an exception is caught from a handler this will result in a failure with status code 500 being signalled.

When handling the failure, the failure handler is passed the routing context which also allows the failure or failure code to be retrieved so the failure handler can use that to generate a failure response.

For the eventuality that an error occurs when running the error handler related usage of not allowed characters in status message header, then the original status message will be changed to the default message from the error code. This is a tradeoff to keep the semantics of the HTTP protocol working instead of abruptly creash and close the socket without properly completing the protocol.

The BodyHandler allows you to retrieve request bodies, limit body sizes and handle file uploads.

You should make sure a body handler is on a matching route for any requests that require this functionality.

The usage of this handler requires that it is installed as soon as possible. The handler will resume the HttpServerRequest processing, since it installs handlers to consume the HTTP request body.

Vert.x-Web has out of the box cookies support.

Vert.x-Web provides out of the box support for sessions.

Sessions last between HTTP requests for the length of a browser session and give you a place where you can add session-scope information, such as a shopping basket.

Vert.x-Web uses session cookies to identify a session. The session cookie is temporary and will be deleted by your browser when it’s closed.

We don’t put the actual data of your session in the session cookie - the cookie simply uses an identifier to look-up the actual session on the server. The identifier is a random UUID generated using a secure random, so it should be effectively unguessable.

Cookies are passed across the wire in HTTP requests and responses so it’s always wise to make sure you are using HTTPS when sessions are being used. Vert.x will warn you if you attempt to use sessions over straight HTTP.

To enable sessions in your application you must have a SessionHandler on a matching route before your application logic.

The session handler handles the creation of session cookies and the lookup of the session, so you don’t have to do that yourself.

Sessions data is saved to a session store automatically after the response headers have been sent to the client. But note that, with this mechanism, there is no guarantee the data is fully persisted before the client receives the response. There are occasions though when this guarantee is needed. In this case you can force a flush. This will disable the automatic saving process, unless the flushing operation failed. This allows to control the state before completing the response like:

Vert.x Session Handler state by default uses a cookie to store session ID. Session ID is a unique string, used to recognize individual visitor between visits. But, if client’s web browser doesn’t support cookies or visitor has disabled cookies in web browser’s settings, we can’t store session id on client’s machine. In this case, new session will be created for every request. This behavior is useless because we can’t remember information for certain visitor between two requests. We can say that, by default, sessions can’t work if browser doesn’t support cookies.

Vert.x Web supports sessions without cookies, known as "cookieless" sessions. As an alternative, Vert.x Web can embed session id inside of page URL. On this way, all page links will contain session id string. When visitor clicks on some of these links, it will read session id from page URL, so we don’t need cookies support to have functional sessions.

To enable cookieless sessions:

It is important to know that in this mode the session ID should be passed by the application to the end user, usually by rendering it on the HTML page or script. There are some important rules. The session id is identified by the following pattern on the path /optional/path/prefix/'('sessionId')'/path/suffix.

As an example, given the path: http://localhost:2677/WebSite1/(S(3abhbgwjg33aqrt3uat2kh4d))/api/ the session Id will be: 3abhbgwjg33aqrt3uat2kh4d.

Main security problem when working with sessions is a possibility that malicious user will find out others' session id. If two users share same session id, they share same session variables too and website is considering them as one visitor. This could be a security risk if session is used for any private or sensitive data, or to allow access to restricted areas of web site. When cookies are used, session id can be protected using SSL and by marking a cookie as secure. But, in case of cookieless session, session id is part of URL and is much more vulnerable.

Vert.x comes with some out-of-the-box handlers for handling both authentication and authorization. In vert.x web the meanings for the 2 words are:

While Authentication is tightened to a well known protocol, e.g.:

Authorization in vert.x is quite generic and can be used regardless of the prior. Yet it is also possible and a valid use case to use the same provider module for both cases.

Vert.x-Web comes with an out of the box handler for auditing security events. This handler is useful for logging to an external system like EDR/XDR. EDR stands for Endpoint Detection and Response and XDR stands for Extended Detection and Response. These are security products that are used to detect and respond to threats on endpoints. The handler is called SecurityAuditLoggerHandler and it used as a top level handler as:

Vert.x-Web comes with an out of the box handler for serving static web resources so you can write static web servers very easily.

To serve static resources such as .html, .css, .js or any other static resource, you use an instance of StaticHandler.

Any requests to paths handled by the static handler will result in files being served from a directory on the file system or from the classpath. The default static file directory is webroot but this can be configured.

In the following example all requests to paths starting with /static/ will get served from the directory webroot:

For example, if there was a request with path /static/css/mystyles.css the static serve will look for a file in the directory webroot/css/mystyle.css.

It will also look for a file on the classpath called webroot/css/mystyle.css. This means you can package up all your static resources into a jar file (or fatjar) and distribute them like that.

When Vert.x finds a resource on the classpath for the first time it extracts it and caches it in a temporary directory on disk so it doesn’t have to do this each time.

The handler will handle range aware requests. When a client makes a request to a static resource, the handler will notify that it can handle range aware request by stating the unit on the Accept-Ranges header. Further requests that contain the Range header with the correct unit and start and end indexes will then receive partial responses with the correct Content-Range header.

Cross Origin Resource Sharing is a safe mechanism for allowing resources to be requested from one domain and served from another.

Vert.x-Web includes a handler CorsHandler that handles the CORS protocol for you.

Here’s an example:

There are cases where your application needs to handle more than just 1 tenant. In this case a helper handler is provided that simplifies setting up the application.

In the case the tenant is identified by a HTTP header, say for example X-Tenant, then creating the handler is as simple as:

You now should register what handler should be executed for the given tenant:

This is useful for security situations:

The tenant id can be read at any moment from the context, for example to decide which resource to load, or which database to connect to:

Multi tenant is a powerful handler that will allow applications to live side by side, however it provides no sandboxing for execution. It should not be used as isolation as wrongly written applications may leak state across tenants.

Vert.x-Web includes dynamic page generation capabilities by including out of the box support for several popular template engines. You can also easily add your own.

Template engines are described by TemplateEngine. In order to render a template render is used.

The simplest way to use templates is not to call the template engine directly but to use the TemplateHandler. This handler calls the template engine for you based on the path in the HTTP request.

By default the template handler will look for templates in a directory called templates. This can be configured.

The handler will return the results of rendering with a content type of text/html by default. This can also be configured.

When you create the template handler you pass in an instance of the template engine you want. Template engines are not embedded in vertx-web, so you must configure your project to access them. Configuration is provided for each template engine.

Here’s an example with the Handlebars template engine: