The http package is a simple package used for performing HTTP requests and parsing the responses. In addition, it has built-in support for keep-alive requests, and HTTP server-sent events. It also contains server request routing, session, and session continuation code similar to Seaside.
The http package requiresthe following packages:
Other packages that are very useful in addition (but not required) are:
For the sake of keeping this README trimmed down (there's a lot to cover), it will be assumed that you are familiar with the url package. It it also assumed that you know the basic anatomies of HTTP requests and responses.
Let's start at the lowest level, walk through a simple example, and then jump to the helper APIs...
First, let's make a simple HTTP request:
CL-USER > (http-make-request "google.com")
#<REQUEST GET http://google.com/>
Next, let's perform the request, which should get us a response.
CL-USER > (http-perform *)
#<RESPONSE 301 "Moved Permanently">
Hmm... looks like the resource isn't where we thought it would be. Let's follow the response to where it says the resource will be.
CL-USER > (http-follow *)
#<RESPONSE 200 "OK">
That looks more like it. Now that we've gotten an OK response, let's see what the content of the response is.
CL-USER > (resp-body *)
"<!doctype html><html>...</html>"
The above may seem like a lot of work for a simple request. However, there are a lot of helper functions that wrap most of this up. It's important to understand at each step of the way what's being done so that you can best use the library for your needs.
The http-perform function takes an optional parameter that is a socket stream. If you don't provide one, then one is created for you (this is the most common use-case). Sometimes, though, you'll want to create your own socket, and keep it alive to make several requests without having to tear it down and open another one up. To do this, you'll need to use the http-open-stream function.
(http-open-stream request &optional timeout) ;=> stream
Additionally, when you create the request, you'll need to pass :keep-alive t, so that the connection isn't closed after the response is read.
(http-make-request url :keep-alive t)
NOTE: even if you want to keep the connection alive, the server may respond with a "Connection: close" header, in which case the output side of the socket will be shutdown.
Be sure and close the sockets when you are done!
Both request and response objects derive from the same base class: headers. The http-headers accessor method can be used to both extract and set the headers of a request or response.
CL-USER > (http-headers (http-get "www.google.com"))
(("Location" "http://www.google.com/")
("Content-Type" "text/html; charset=UTF-8")
("Date" "Thu, 10 Apr 2014 18:08:21 GMT")
("Expires" "Sat, 10 May 2014 18:08:21 GMT")
("Cache-Control" "public, max-age=2592000")
("Server" "gws")
("Content-Length" "219")
("X-XSS-Protection" "1; mode=block")
("X-Frame-Options" "SAMEORIGIN")
("Alternate-Protocol" "80:quic")
("Connection" "close")
...)
The http-header method can be used to extract a single header from a request or response.
CL-USER > (http-header (http-get "google.com") "Content-Type")
"text/html; charset=ISO-8859-1"
The http-header method is also setf-able, and can be used to change or add headers to a request or response.
There are many situations in which a header may actually exist many times in a response or request (e.g. "Set-Cookie" headers). For those situations, http-header also takes an optional all keyword parameter that defaults to nil. When non-nil, it will return a list of all the headers it finds.
Since request is also a subclass of headers, you can optionally set headers whenever creating the your request.
CL-USER > (http-make-request "apple.com" :headers '(("x-my-header" 10)))
#<REQUEST GET http://apple.com>
Although the response is automatically read for you, the body of the response may not be. If you supply :read-body nil to the request (the default is T), then the body will not be read.
By default, the http package will send an "Accept-Encoding" header with your requests. However, the only accepted encoding is the "identity" encoding. With only this encoding, a server shouldn't send an zipped or encoded responses.
You can, however, register other encodings with the http package, and those encodings will automatically be included in the "Accept-Encoding" header with all requests.
(http-register-content-encoding name &key encoder decoder)
For example, if you had the functions gzip and gunzip, you could register the "gzip" encoding with those functions, like so...
(http-register-content-encoding "gzip" :encoder 'gzip :decoder 'gunzip)
...and then all requests made will include "gzip" in the list of accepted encodings to a server.
Additionally, all requests made can have their bodies encoded as well. Simply set the "Content-Encoding" header in your request to the encoding type you'd like to use and the data will be thusly encoded for you before it is sent (the body of the request will not be altered, only the data sent over the wire).
NOTE: the same is true of the "Content-Type", simply setting it in the header of a request will encoded the body of the request properly for you before sending it. And any decoded response will use the "Content-Type" as well.
All of the above is nicely wrapped up for you in the following helper functions:
(http-get url &rest request-initargs)
(http-head url &rest request-initargs)
(http-options url &rest request-initargs)
(http-trace url &rest request-initargs)
(http-delete url &rest request-initargs)
(http-put url &rest request-initargs)
(http-post url &rest request-initargs)
(http-patch url &rest request-initargs)
Each will create a request to the URL provided along with the other keyword initargs. The :read-body initarg will be NIL for http-head. The request will be performed and any redirects will be automatically followed for you (up to 3 times).
CL-USER > (http-get "google.com")
#<RESPONSE 200 "OK">
The available request-initargs are:
:method The HTTP verb. Defaults to "GET".
:protocol The HTTP protocol. Default is "1.0".
:headers An associative list of (header value) pairs. Default is NIL.
:keep-alive If T, do not close the socket. Default is NIL.
:read-body Reads the respone body if T. Default is T.
:body Binary or string data to send with the request. Default is NIL.
The with-response macro can be used when you only care about successful requests (i.e. (<= 200 resp-code 299) is T).
CL-USER > (with-response (resp (http-get "www.apple.com"))
(resp-code resp))
200
T
This can return up to 3 values. If the request was successful, the return value will be the result of the body and T to indicate that the request was successful. If the request failed, the return values will be NIL (no result), NIL (failure), and the response.
CL-USER > (with-response (resp (http-get "www.apple.com/foo"))
(resp-code resp))
NIL
NIL
#<RESPONSE 404 "Not Found">
The with-headers macro is useful if you would like to bind variables to header values and also make them setf-able within a body of code.
(let ((req (http-make-request "www.google.com")))
(with-headers ((dummy "x-my-dummy-header"))
req
(setf dummy "My dummy value")))
NOTE: you can bind variables in with-headers to headers that don't exist yet, but will be added after setf-ing them.
Adhearing to http://www.w3.org/TR/2011/WD-eventsource-20111020/, you can easily open an event stream to process continuous events sent from an HTTP server.
(http-open-event-stream url event-callback &key method headers redirect-limit)
The event-callback should be a function that takes 3 parameters when called: the event type, an id, and a data value. The id and data are both optional and may be NIL.
CL-USER > (flet ((process-event (type id data)
(format t "New event: ~s ~s ~s~%" type id data)))
(http-open-event-stream "url.com" #'process-events))
The http package can be a pretty descent server as well. It supports basic routing, multi-threading, configuration, session management, continuations, and more.
NOTE: All of the examples below require use of the HTML package.
Before we can start the server, we need to define a router:
CL-USER > (define-http-router my-routes
(:get "/" #'(lambda ()
(http-ok "Hello, world!"))))
Once you've defined your router, you can then start the server with it:
CL-USER > (http-start-server 'my-routes)
It takes other, keyword arguments besides a router, but we'll start with just the router for now. To test it, just open your browser to localhost:8000 and see the response.
Each server route is essentially made up of the following:
- a method
- a path
- a function
The method should be a string or keyword for the HTTP method this route responds to (e.g. :GET or :POST).
The path is the request path for the response. It may contain keywords, which will end up being passed to the route's function (e.g. "/user/:user").
The function takes any of the arguments in the path as keyword arguments and uses the dynamic variables for the *server-config*, *session*, and *response* to generate the successful response body or error.
Example:
(define-http-router reddit-clone
(:get "/top" 'handle-top-posts)
(:post "/submit" 'handle-submit-post)
(:get "/user/:username" 'handle-user-lookup))
(defun handle-top-posts ()
(http-ok (render-top-posts)))
(defun handle-submit-post ()
(let* ((req (resp-request *response*))
(post (req-body req)))
(create-post *session* post)
(http-ok)))
(defun handle-user-lookup (&key username)
(let ((user (find-user username)))
(if user
(http-ok (render-user user))
(http-not-found))))This shows some extra use of *session* for the current user session data, as well as the *response* object being constructed (to get the request). It also shows use of a keyword argument in a route function defined by the path.
Each server - when started - requires a configuration object that is a subclass of server-config. If one isn't provided then a default configuration is used. During the handling of routes, the *server-config* variable will be available for use.
Since each request is processed in a separate thread, any "global" data you'd like to share among all routes should be stored in the server config. For example, maybe you're hosting a blog server and need a database connection that all routes can use:
(defclass blog-config (server-config)
((db-conn :initarg :db :accessor blog-db)))
(define-http-router blog-router
(:get "/post/:id" 'render-post))
(defun render-post (&key id)
(let ((db (blog-db *server-config*)))
(http-ok (lookup-post db id))))
(defun start-blog-server (db-connection)
(let ((config (make-instance 'blog-config :db db-connection)))
(http-start-server 'blog-router :config config)))Each unique user connecting to your server is given a session object that is a subclass of http-session. This class can be specified in the server-config object if you wanted anything unique tied to the session (e.g. a user ID).
(defclass my-session (http-session)
((user-id :initform nil :accessor session-user-id)))
(defun start-server ()
(let ((config (make-instance 'server-config :session-class 'my-session))
(http-start-server 'my-router :config config))))NOTE: Keep in mind that when a session is created, you won't be able to pass any arguments to it. So either use :initform to initialize members or create an initialize-instance method for your session class.
Along with the :session-class is also a :session-timeout, which is the number of minutes the session is valid for. The session is refreshed each time the user requests another page.
If you do not want sessions, simply set the :session-class to nil and no session objects will be created.
The HTTP server module is a continuation-based server (see: Seaside). Each session tracks all the continuations available to it. This can be very handy for passing dynamic content between pages through links.
(define-http-router counter-router
(:get "/" 'counter))
(defun counter (&optional (n 0))
(flet ((next-number (n)
(http-make-continuation 'counter n)))
(http-ok
(html-render
(<html>
(<body>
(<h1> n)
(<hr>)
(<code>
(<a> :href (next-number (1+ n)) "++")
#\space
(<a> :href (next-number (1- n)) "--"))))))))Try it out!
You definitely aren't required to use continuations in your server, but they can be incredibly useful. For example, consider a REST API used to query a database and you'd like to support paginated requests. With continuations, you can return both the current results as well as a link that's able to be used to get the next page of results.
NOTE: Using continuations completely by-passes the router! This can be used to great effect. For example, you can create "routes" that are impossible to reach without having a session and gone through a very specific path of code.
- Add support for serving static files
- Add support for SSL server
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent