A ready to go, modular, multi transport, streaming friendly, JSON(P) API Server.
Strong competitors I guess.
Express targets web applications providing support for templates, views, and all the facilities the you probably need if you're writing a web app. Restify let you "build "strict" API services" but it's too big and it concentrates on server to server API, that will not be consumed by your browser.
ApiServer is rad. It is a slim, fast, minimal API framework, built to provide you a flexible API consumable both in the browser and from other apps. It ships with JSON, JSONP (GET/POST) transports and a powerful fast routing engine OOTB. The source code is small, heavily tested and decoupled. Your API source will be well organized in context objects, allowing you to keep it in a meaningful maintainable way.
- Streaming JSON(P) transport GET/POST browser friendly
- Fast routing system with cached routes
- Implicit route parameters (rails like)
- Fully configurable custom routing
- API modules as Objects/Classes
- Payload paused by default
- Transports decoupled from the core
- Router decoupled from the core
- Compatible with senchalabs/Connect middleware
⚡ npm install apiserver
var ApiServer = require('apiserver')
The example below is intended to be a small sneak peek of the ApiServer API, modules and routes should be moved to separate files.
var ApiServer = require('apiserver')
var apiServer = new ApiServer({ port: 8080 })
// middleware
apiServer.use(/^\/admin\//, ApiServer.httpAuth({
realm: 'ApiServer Example',
encode: true,
credentials: ['admin:apiserver']
}))
apiServer.use(ApiServer.payloadParser())
// modules
apiServer.addModule('1', 'fooModule', {
// only functions exposed
options: {
opt1: 'opt1',
opt2: 'opt2',
opt3: 'opt3'
},
foo: {
get: function (request, response) {
response.serveJSON({
id: request.querystring.id,
verbose: request.querystring.verbose,
method: 'GET',
options: this.options
})
},
post: function (request, response) {
request.resume()
request.once('end', function () {
response.serveJSON({
id: request.querystring.id,
verbose: request.querystring.verbose,
method: 'POST',
payload: request.body // thanks to payloadParser
})
})
}
},
bar: function (request, response) {
response.serveJSON({ foo: 'bar', pow: this._pow(5), method: '*/' + request.method })
},
// never exposed due to the initial underscore
_pow: function (n) {
return n * n
}
})
// custom routing
apiServer.router.addRoutes([
['/foo', '1/fooModule#foo'],
['/foo/:id/:verbose', '1/fooModule#foo'],
['/foo_verbose/:id', '1/fooModule#foo', { 'verbose': true }],
['/bar', '1/fooModule#bar', {}, true] // will keep default routing too
])
// events
apiServer.on('requestStart', function (pathname, time) {
console.info(' ☉ :: start :: %s', pathname)
}).on('requestEnd', function (pathname, time) {
console.info(' ☺ :: end :: %s in %dms', pathname, time)
}).on('error', function (pathname, err) {
console.info(' ☹ :: error :: %s (%s)', pathname, err.message)
}).on('timeout', function (pathname) {
console.info(' ☂ :: timedout :: %s', pathname)
})
apiServer.listen()
Server will respond to
GET, POST http://localhost:8080/foo
GET, POST http://localhost:8080/foo/5/true
GET, POST http://localhost:8080/foo_verbose/5
* http://localhost:8080/bar
* http://localhost:8080/1/foo_module/bar
For full and detailed examples look at the examples folder
All options will be also passed to the the default transport (JSONTransport) constructor, then add here your transport configuration.
new ApiServer([options])
- port - (
Number|String
: defaults to 8080) the server binding port - server - (
http(s).Server
: defaults http.Server) - timeout - (
Number
: defaults to 15000) milliseconds to wait before arbitrary closing the response - router - (
Object
: defaults to the standard router) the routes manager conforms to the router interface - standardHeaders - (
Object
: below the default) response headers defaults, can be overwritten by the transport
{
'cache-control': 'max-age=0, no-cache, no-store, must-revalidate',
'expires': 0,
'pragma': 'no-cache',
'x-server': 'ApiServer v' + ApiServer.version + ' raging on nodejs ' + process.version
}
var https = require('https'),
ApiServer = require('apiserver')
apiserver = new ApiServer({
port: 80,
server: https.createServer(),
standardHeaders: {
'cache-control': 'max-age=0, no-cache, no-store, must-revalidate',
'x-awesome-field': 'awezing value'
},
timeout: 2000,
indent: ' ', // transport
domain: '.myservice.com', // transport
defaultRoute: '/:version/:module/:method' // router
})
Adds a new module to to the current API set. It triggers the router.update
method.
ApiServer.prototype.addModule(apiVersion, moduleName, apiModule)
- apiVersion - (
String
) the version of the API you want to add your module to, it will be the part of the url - moduleName - (
String
) the name of the module, this will be the second part of your derived routes, after a case conversion - apiModule - (
Object
) the module object conform to the modules interface
var apiserver = new ApiServer()
apiserver.addModule('v1', 'user', userModule)
apiserver.addModule('v1', 'pages', pageModule)
apiserver.addModule('v2', 'user', userModule2)
Adds a middleware object to the middleware chain. It triggers the router.update
method.
Each middleware is associated to a RegExp
used to test the API end-point route. If the route matches the RegExp
the middleware will be a part of the chain and will be executed.
Read more about middleware here.
ApiServer.prototype.use([route], middleware)
- route - (
RegExp
: defaults to/./
) regular expression that the route should match - middleware - (
Object
) the middleware object conforms to the middleware interface
var apiserver = new ApiServer()
apiserver.use(new MyMiddleWare({ foo: 'bar', bar: true }))
apiserver.use(/(signin|signup)/, ApiServer.payloadParser())
apiserver.use(/^\/v1\/files\/upload$/, ApiServer.multipartParser())
Bind the server to a port
ApiServer.prototype.listen([port], [callback])
- port - (
Number|String
) overwrite the constructor port parameter - callback - (
Function
) called when the port is actually bound to the server
From this point on, all the examples will take the require statements as assumption
var apiserver = new ApiServer()
apiserver.listen(80, function (err) {
if (err) {
console.error('Something terrible happened: %s', err.message)
} else {
console.log('Successful bound to port %s', this.port)
}
})
Unbind the server from the current port
ApiServer.prototype.close([callback])
- callback - (
Function
) called when the port is actually unbound from the server
var apiserver = new ApiServer()
apiserver.listen(80, onListen)
function onListen(err) {
if (err) {
console.error('Something terrible happened: %s', err.message)
} else {
setTimeout(function () {
apiserver.close(onClose)
}, 5000)
}
}
function onClose() {
console.log('port unbound correctly')
}
Emitted when an API endpoint got hit.
apiserver.on('requestStart', function (url, requestTime) {
})
- url (
String
) - therequest.url
- requestTime (
Number
) - when the API method was requested
Emitted when an API method closes the response, even with response.end
.
apiserver.on('requestEnd', function (url, responseTime) {
})
- url (
String
) - therequest.url
- responseTime (
Number
) - how log the API method took for closing the response
Emitted when an API method exceed the maximum allowed time (see timeout
option), before closing the response.
apiserver.on('timeout', function (url) {
})
- url (
String
) - therequest.url
Emitted when a sync error is triggered during the middleware chain execution, can be both your API, a transport or a simple middleware.
You still have to deal with async errors
apiserver.on('error', function (url, err) {
})
- url (
String
) - therequest.url
- err (
Error
) - the error which triggered the event
A module is a set of API end-points grouped in the same context:
- context: a simple object
- end-point: function/method accessible by the object and scoped within the object
Each module method (API end-point) must implement this interface and expect request and response parameters
function (request, response)
The request object is "extendend" ootb with the following members (aliases in round brackets):
- requestedAt (at): timestamp of the request
- parsedUrl: a parsed version of the request url with
url.parse
- pathname (path): the pathname that corresponds to the end-point route
- querystring (qs): the querystring object parsed with visionmedia/node-querystring
As you can see, there is no callback to call, you have to deal directly with the response.
Take a look at your transport documentation and use the right method that ships within the response object. You can also roughly close and write to the response stream in an edge case.
var apiserver = new ApiServer()
var userModule = {
signin: function (request, response) {
// rough approach
response.writeHead(200)
response.end('ok')
},
signout: function (request, response) {
// JSON transport
response.serveJSON({ foo: 'bar' })
}
}
apiserver.addModule('v1', 'user', userModule)
var apiserver = new ApiServer()
function UserModule(options) {
this.database = options.database
this.serviceName = options.serviceName
}
UserModule.prototype.signin = function (request, response) {
var self = this
self.database.searchUser(request.querystring.username, function (err) {
if (err) {
response.serveJSON({ success: false, err: err.message })
} else {
response.serveJSON({ success: true, message: 'welcome to ' + self.serviceName })
}
})
}
UserModule.prototype.signout = function (request, response) {
// you can use the response as usual
// a redirect for example
response.writeHead(302, {
'location': 'http://example.org/logout_suceesful'
})
response.end()
}
var database = /* your db object*/
apiserver.addModule('v1', 'user', new UserModule(database, 'My Awesome Service'))
The concept of middleware is not new at all, you can find the same pattern in senchalabs/Connect in mcavage/node-restify and in many others. A middleware is a piece of software that adds (or patches) a feature into another software. Usually there is a common interface to implement, because the caller software, in this case our ApiServer, should know how to interact with the middleware.
You should check out the source code for a large understanding, middleware is relatively easy to code.
The ApiServer uses kilianc/node-fnchain to execute all the active middleware and reach the API method (that actually is the last ring of the chain). This means that the order of the execution depends on the order you activated the middleware.
Each middleware can both exit with an error or explicitly stop the chain (not reaching your API method). This is useful in case of a precondition check (auth, sessions, DoS attack filter...), or just because you packed some shared code as middleware which must be executed before your API method.
At the middleware execution level, the response object is already patched with the default transport methods, so you can use these methods to write and close the response. Is a good practice to leave at the top of the chain the extra transports middleware.
// constructor adds the default transport automatically
var apiserver = new ApiServer()
// let's ad first our custom transports
apiserver.use(/\.xml$/, myXMLTransport())
apiserver.use(/\.csv$/, myCSVTransport())
apiserver.use(/\.yml$/, myYAMLTransport())
// now activate our middleware
apiserver.use(/form/, ApiServer.payloadParser())
apiserver.use(/upload/, ApiServer.multipartParser())
...
The request payload (the data
event) is paused by default and can be resumed calling request.resume()
at any level of execution: middleware, module, transport. Why? Because you should explicitly accept or refuse a payload, this way you will save memory not buffering useless data.
Take a look at both the pause and resume official docs.
ApiServer is using this patch to provide a robust buffered pause resume method, so you don't have do deal with the flying chunks after the pause call
Each middleware must implement this interface.
module.exports = function (options) {
return function (request, response, next) {
// do sometihng async and when you're done call the callback
options.count++
next()
}
}
A middleware basically, is a function that returns another function, this one must declare 3 paramaters:
- request: the server request already extended by the server
- response: the server response already extended by the transports
- next: a callback in the following form
function (err, stop)
The next
callback expects 2 parameters:
- err - (
Error
) an error object that will throw a server error event and will close the response - stop - (
Boolean
) a flag that stops the internal chain, that means that your API method will never be called and your middleware should be able to correctly close the response. At this point you already have all the transports available, and you can freely use them.
A transport is a particular middleware that "extends" the response object. It can provide new methods that allow you to serve your data to the client in different ways.
Usually this is how you send data back to the client:
function (request, response) {
response.writeHead(200, {
'content-type': 'application/json'
})
response.end(JSON.stringify({ foo: 'bar' }))
})
This is for example how the default JSONTransport simplify the process
function (request, response) {
response.serveJSON({ foo: 'bar' })
})
Basically what a transport does, is to wrap your data around a meaningful format (JSON, JSONP, HTML, XML, CSV, ...) understandable by your clients. It takes care of all the small things that the raw response needs (headers, status codes, buffering, ...)
Transports must be at the top of the middleware chain, in order to allow other middleware to use them.
JSONTransport is the default one, is attached before the middleware chain execution and then is available at every level of execution. You don't need to allocate it directly, the server itself will allocate the transport passing as options the ApiServer constructor options object.
module.exports = function (options) {
function serve<FORMAT>(request, response, data, options) {
response.writeHead(200, {
'content-type': 'application/<FORMAT>'
})
response.end(<FORMAT>.stringify(data))
}
return function (request, response) {
// attach some new method to the response
response.serve<FORMAT> = serve<FORMAT>.bind(this, request, response)
}
}
where <FORMAT>
is the formatting method of your data.
Apiserver uses apiserver-router as default router, a fast routing system with integrated caching. It basically translates your API methods names in routes, doing some convenient case conversion. Also, it supports (rails like) custom routes and implicit route parameters.
You can change the default behavior passing a custom router as router
option in the ApiServer constructor.
function UserModule(options) {
this.options = options
}
// will be translated into /1/random_photo_module/create_album
UserModule.prototype.createAlbum = function (request, response) { ... }
// will be translated into /1/random_photo_module/upload_photo
// in this case we will overwrite the default path with a custom one
UserModule.prototype.uploadPhoto = {
post: function (request, response) { ... }
}
// private method, skipped by the router
UserModule.prototype._checkFileExtension = function (request, response) { ... }
apiserver.addModule('1', 'randomPhotoModule', new UserModule())
apiserver.router.addRoute('/photo/:caption', '1/randomPhotoModule#uploadPhoto')
N.B. the moduleName
also will be translated
Your custom router must implement the following interface.
function Router () {
...
}
Router.prototype.update = function (modules, middlewareList) {
...
}
Router.prototype.get = function (request) {
...
}
The get
method must return the the middleware chain associated with the request
parameter, and eventually extend the request
with new data (ex. implicit route parameters).
JSONTransport is the default transport bundled with ApiServer and we can call it the real killer feature.
It provides JSON and JSONP that work with both GET / POST methods.
// decontextualized API method
function (request, response) {
response.serveJSON({ foo: 'bar' })
})
// decontextualized API method
function (request, response) {
response.serveJSON(['foo','bar', ...], {
httpStatusCode: 404,
httpStatusMessage: 'maybe.. you\'re lost',
headers: {
'x-value': 'foo'
}
})
})
// decontextualized API method
function (request, response) {
var count = 3
var interval = setInterval(function () {
if (count === 0) {
clearInterval(interval)
response.streamJSON()
} else {
count--
response.streamJSON({ foo: 'bar' })
}
}, 200)
})
yields
[
{ "foo": "bar" },
{ "foo": "bar" },
{ "foo": "bar" }
]
Read the full docs here
The payload parser automatically buffers the payload and parse it. It only works with PUT POST OPTIONS http methods, because they are the only that can carryout a payload by specs definition.
Two kinds of payload can be parsed:
application/x-www-form-urlencoded
application/json
The following attributes will be attached to the request object:
- body: an object containing the parsed data
- rawBody: the raw payload as binary buffer
- parseError: can be
null
orError
in case of parse error
ApiServer.payloadParser()
var apiserver = new ApiServer()
apiserver.use(/1\/my_module\/my_method_api$/, ApiServer.payloadParser())
apiserver.addModule('1', 'myModule', {
'my_method_api': function (request, response) {
request.resume()
request.once('end', function () {
if (request.parseError) {
// :(
console.error(request.parseError.message)
} else {
request.body // an object containing the parsed data
request.rawBody // contains a binary buffer with your payload
}
})
}
})
The multipart-parser the attach the payload to a felixge/node-formidable IncomingForm
object. It only works with PUT POST OPTIONS http methods, because they are the only that can carryout a payload by specs definition.
Only a multipart/form-data
payload is parsed and the following attribute will be attached to the request object:
- form an IncomingForm object, read how to deal with it
The following attributes will be attached to the request object, after the IncomingForm end event:
- body: an object containing the parsed data
- files: array of uploaded files instaceof formidable.File
- parseError: can be
null
orError
in case of parse error
ApiServer.multipartParser()
var apiserver = new ApiServer()
apiserver.use(/1\/my_module\/my_method_api$/, ApiServer.multipartParser())
apiserver.addModule('1', 'myModule', {
'my_method_api': function (request, response) {
var fields = Object.create(null)
request.resume()
request.form.on('field', function (name, value) {
fields[name] = value
})
request.form.on('file', function (name, file) {
fields[name] = fs.readFileSync(file.path, 'utf8')
})
request.form.once('end', function () {
// do something with your data
})
},
'my_smarter_api': function (request, response) {
request.resume()
request.form.once('end', function () {
// do something with your data
request.body // fields
request.files // files
request.parseError // error
})
}
})
The httpauth middleware acts as an auth precondition, checking the authorization
headers sent with the request.
If the request doesn't pass the authorization check, httpAuth will close the response using the standard JSONTransport:
response.serveJSON(null, {
httpStatusCode: 401,
headers: { 'www-authenticate': 'Basic realm=\'' + realm + '\'' }
})
This will trigger a user/password prompt in your browser
ApiServer.httpAuth([options])
- realm: (
String
) the name of your service, this is used by the browser when it prompts for username and password - credentials - (
Array
) a list of strings (credentials), if your client is a browser you must use the form username:password - encode: (
Boolean
: defaults to false) set to true if your client is a browser (will base64 encode)
var apiserver = new ApiServer()
apiserver.use(/1\/admin\/.+/, ApiServer.httpAuth({
realm: 'signin please',
credentials: ['foo:password','bar:password', ...],
encode: true // we suppose that at the other end of the wire we have a browser
}))
apiserver.addModule('1', 'admin', {
'protectedApi': function (request, response) {
// this will executed only if you provide valid credentials
}
})
ApiServer follows the awesome Vincent Driessen branching model.
- You must add a new feature on his own topic branch
- You must contribute to hot-fixing directly into the master branch (and pull-request to it)
ApiServer follows (more or less) the Felix's Node.js Style Guide, your contribution must be consistent with this style.
The test suite is written on top of visionmedia/mocha and it took hours of hard work. Please use the tests to check if your contribution is breaking some part of the library and add new tests for each new feature.
⚡ npm test
and for your test coverage
⚡ make test-cov
This software is released under the MIT license cited below.
Copyright (c) 2010 Kilian Ciuffolo, me@nailik.org. All Rights Reserved.
Permission is hereby granted, free of charge, to any person
obtaining a copy of this software and associated documentation
files (the 'Software'), to deal in the Software without
restriction, including without limitation the rights to use,
copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the
Software is furnished to do so, subject to the following
conditions:
The above copyright notice and this permission notice shall be
included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED 'AS IS', WITHOUT WARRANTY OF ANY KIND,
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
OTHER DEALINGS IN THE SOFTWARE.