xdrm-framework/notice/api/3.0.md

17 KiB

module: api
version: 3.0
requires:
  - http:  1.0
  - error: 2.0

For developers

I. Overview

II. Usage

III. Configuration

IV. Implementation

V. Type Checker

VI. Documentation

For clients

I. Simple request

II. Usage

I. Overview

1 Introduction & features

The api package (v3.0) allows you to easily create and manage an API for your applications. It can be used as an HTTP API (REST, or other kind), and you can use it as an internal core for your system.

The aim of this package is to make your life easier working with APIs and internal delegation. The only things you have to do is to implement your processes and write the configuration, the package will do the rest.

Things you have to do :

  • write the configuration file (cf. configuration)
  • implement your authentication system (cf. AuthSystem)
  • implement your processes (obviously)

Things you don't have to do :

  • input type check (cf. Checker)
  • multiple permission management
  • optional or required input
  • multipart and file input
  • URL input
  • before and after scripts
  • and a lot more ...

2 Basic knowledge

The api is made of paths that binds to a php class, each path can manage multiple HTTP METHODS each will correspond to a method of the bound class.

So each of your functionalities must have a dedicated HTTP method and a corresponding a path.

Example:

  • the module article contains methods:
    • GET to get article data
    • POST to post a new article
    • PUT to edit an existing article
    • DELETE to delete an exisint article

Note that each method must be a valid HTTP METHOD.

II. Usage

1 REST API

In order for the API to catch URI, you must use a router. It will allow the API to get the requested URI.

Important: You must use $_SERVER['REQUEST_URI'] because it will (instead of other methods) keep some useful format.

<?php

require_once '../autoloader.php';


/* (1) Create the request */
$request = \api\core\Loader::remote($_SERVER['REQUEST_URI']);

/* (2) Process the response (execute your implementation) */
$response = $request->dispatch();

/* (3) Serialize the response */
echo $response->serialize();

Note: Request or Response errors will propagate to serialize() method into the JSON output.

2 Internal use

You can also use the API from within your code (not from URI).

<?php

require_once '../autoloader.php';


/* (1) Emulate API data */
$emu_uri    = '/some/target/uri';
$emu_params = [ 'p1' => 'param1', 'URL0' => 'uri param 0' ];
$emu_method = 'DELETE';


/* (2) Create the request */
$request = new \api\core\Request($emu_uri, $emu_params, $emu_method);

/* (3) Process the response (execute your implementation) */
$response = $request->dispatch();

/* (4) [OPTIONAL] Check for errors */
if( $response->error->get() != \error\core\Err::Success )
	die('encountered error: '.$response->error->explicit());

/* (5) Fetch response data */
$all_response_fields     = $response->getAll();
$specific_response_field = $response->get('specific_field_name');

III. Configuration

The documentation consists of a chain of urls each one can contain several HTTP method specifications. Note that each url can be chained apart the method specifications.

1 - configuration format

The configuration is a set of uri paths that can contain up to 4 method types: POST, DELETE, PUT, GET or uri subpaths.

For instance the 4 methods directly inside uri1 will be triggered when the calling URI is /uri1, the ones directly inside uri2 will be triggered when calling /uri1/uri2 and so on..

You can also set full paths if you don't need transitional methods, for instance the path uri5/uri6/uri7 will be triggered by the url /uri5/uri6/uri7.

Example: The example file loaded with the default configuration can be found here.

{
	"uri1" : {
		"GET":    method.definition,
		"POST":   method.definition,
		"PUT":    method.definition,
		"DELETE": method.definition,

		"uri2": {
			"GET":    method.definition,
			"POST":   method.definition,
			"PUT":    method.definition,
			"DELETE": method.definition,

			"uri3": {}
		}
	},

	"uri5/uri6/uri7": {
		"GET":    method.definition,
		"POST":   method.definition,
		"PUT":    method.definition,
		"DELETE": method.definition
	}
}

Note: It is possible to trigger the root uri (/), so you can set methods directly at the root of the JSON file.

2 - method.definition

{
	"des": method.description,
	"per": method.permissions,
	"par": method.parameters,
	"out": method.output_format,
	"opt": method.options
}

2.1 - method.description

The description field must be a string containing the human-readable description of what the method does.

2.2 - method.permissions

The permissions field must be an array. You can manage OR and AND permission combinations.

  • OR is applied between each 0-depth array
  • AND is applied between each 1-depth array

For instance the following permission [ ["a","b"], ["c"] ] means you need the permissions a and b combined or only the permission c.

2.3 - method.parameters

The parameters field must be an object containing each required or optional parameter needed for the implementation.

"parameter.name": {
	"des": parameter.description,
	"typ": parameter.checker_type,
	"opt": parameter.is_optional,
	"ren": parameter.rename,
	"def": parameter.default_value,
}

parameter.name

The name field must be a string containing variable name that will be asked for the caller.

Note that you can set any string for body parameters, but GET parameters must be named URL#, where # is the index within the URI, beginning with 0.

parameter.description

The description field must be a string containing the human-readable description of what the parameter is or must be.

parameter.checker_type

The checker_type field must be a string corresponding to a \api\core\Checker type that will be checked before calling the implementation.

parameter.is_optional

The is_optional field must be a boolean set to true if the parameter can be ommited. By default, the field is set to false so each parameter is required.

parameter.rename

The rename field must be a string corresponding to the variable name given to the implementation. It is useful for GET parameters because they need to be called URL#, where # is the position in the URI. (cf. paramter.name))

If ommited, by default, parameter.name will be used.

parameter.default_value

The default_value field must be of any type according to the checker_type field, it will be used only for optional parameters when ommited by the caller

By default, each optional parameter will exist and will be set to null to the implementation.

2.4 - method.output_format

The output_format field must have the same format as method.parameters but will only be used to generate a documentation or to tell other developers what the method returns if no error occurs.

2.5 - method.options

The options field must be an object containing the available options.

The only option available for now is:

"download": true

Your implementation must return 2 fields:

  • body a string containing the file body
  • headers an array containing as an associative array file headers

If the API is called with HTTP directly, it will not print the json response (only on error), but will instead directly return the created file.

AJAX: If called with ajax, you must give the header HTTP_X_REQUESTED_WITH set to XMLHttpRequest. In that case only, it will return a normal JSON response with the field link containing the link to call for downloading the created file. You must take care of deleting not used files - there is no such mechanism.

IV. Implementation

1 - permissions : AuthSystem

In order to implement your Authentification System you have to implement the interface AuthSystem located in /build/api/core/AuthSystem.

You must register your custom authentification system before each api call.

For instance, lets suppose your implementation is called myAuthSystem.

\api\core\Request::setAuthSystem(new myAuthSystem);

Note: By default the API will try to find \api\core\AuthSystemDefault.

2 - core implementation

classes

Each module's implementation is represented as a file so as a class located in /build/api/module/. In order for the autoloader to work, you must name the file the same name as the class. Also the namespace must match, it corresponds to the path (starting at /api).

For instance if you have in your configuration a path called /uri1/uri2/uri3, you will create the file /build/api/module/uri1/uri2/uri3.php.

Specific: For the root (/) path, it will trigger the class \api\module\root, so you have to create the file /build/api/module/root.php. (cf. configuration format)

methods

Each method is represented as a class' method with the same name as the associated HTTP method.

method arguments

Arguments are passed to the method as a single argument which an associative array according to the configuration.

Notes:

return statement

You MUST return an associative array containing at least the field error containing an instance of /api/core/Error, then you can add whatever you want to return in the array.

The list of available errors are set in the class \error\core\Err.

If you don't return the 'error' field, by default to

[ 'error' => new \error\core\Error(\error\core\Err::Success) ]

Before and After scripts

Each time a method is called, the api creates an instance from the class, and after the execution, the class is destroyed. So you can implement the methods __construct and __destruct to add before and after scripts.

example

For instance here, we manage the call GET /article/2 where 2 is the argument URL0 renamed to id_article.

use \error\core\Error;
use \error\core\Err;


public function GET($parameters){

	/* (1) Set parameters available in the method scope */
	extract($parameters);

	/* (2) You can now use the variable `id_article` */
	$article_data = get_article_from_database($id_article);

	/* (3) Return error if error in process */
	if( has_error($article_data) )
		return [ 'error' => new Error(Err::ModuleError, 'Cannot fetch data from db.') ];


	/* (4) Return data on success */
	return [
		'id_article' => $id_article,
		'article' => $article_data
	];

}

Note: Functions get_article_from_database() and has_error() do not exist, it was in order for all to understand the example

V. Type Checker

\api\core\Checker checks the input values according to the type given in the configuration.

The default types below are available in the default package. To add a new type, just open the file /build/api/Checker.php and add an entry in the switch statement.

1 - Default types

Type Example Description
mixed [9,"a"], "a" Any content (can be simple or complex)
id 10, "23" Positive integer number between 0 and 2147483647
numeric -10.2, "23" Any number, null and the string "null"
text "Hello!" String that can be of any length (even empty)
hash "4612473aa81f93a878..." String with a length of 128, containing only hexadecimal characters
alphanumeric "abc029.-sd9" String containing only alphanumeric, _, -, and . characters
letters "abc -sd" String containing only letters, -, and space characters
mail "a.b@c.def" Valid email address
array [1, 3] Non-empty array
object works only within php Non-empty object
boolean true, false Boolean
varchar(a,b) "Hello!" String with a length between a and b (included)
varchar(a,b,c) "abc" String with a length between a and b (included) and matching the c type

2 - Complex types

Type Sub-Type Description
array<a> a Array containing only entries matching the type a
FILE _a raw file send in multipart/form-data A raw file sent by multipart/form-data

Note: It is possible to chain array type as many as needed.

Ex.: array<array<id>> - Will only match an array containing arrays that only contains id entries.

VI. Documentation

With the all-in-config method, we can generate a consistent documentation or other documents from the /config/modules.json file.

1 - API accessible documentation

You can request the API for information about the current URI by using the OPTIONS HTTP method.

====

I. Simple request

1 - URL

format

The uri format is as defined: {base}/{path}/{GET_parameters}, where

  • {base} is the server's API base uri (ex: https://example.com/api/v1 or https://api.exampl.com)
  • {path} is the effective path you want to access (ex: article/author)
  • {GET_parameters} is a set of slash-separated values (ex: val0/val1/val2//val4)

Note: GET parameters are not used as usual (?var1=val1&var2=val2...), instead the position in the URL gives them an implicit name which is URL#, where # is the index in the uri (beginning with 0).

example 1

If you want to edit an article with the server's REST API, it could be defined as following:

http: PUT
path: article/{id_article}
input:
  body: new content of the article
output:
  updated: the updated article data

Let's take the example where you want to update the article which id is 23 and set its body to "blabla new content"

HTTP REQUEST

PUT article/23 HTTP/1.0

body=blabla+new+content

HTTP RESPONSE

HTTP/1.0 200 OK
Content-Type: application/json

{
	"error": 0,
	"ErrorDescription": "all right",

	"updated": {
		"id_article": 23,
		"title": "article 23",
		"body":  "blabla new content"
	}

}

example 2

If you want to get a specific article line, the request could be defined as following

http: GET
path: article/line/{id_article}/{no_line}
input: -
output:
  articles: the list of matching lines

Let's take the example where you want to get all articles because id_article is set to optional, but you only want the first line of each so you have to give only the second parameter set to 1 (first line).

Solution: The position in the uri where id_article must be, have to be left empty: it will result of 2 slashes (//).

HTTP REQUEST

GET article/line//1 HTTP/1.0

HTTP RESPONSE

HTTP/1.0 200 OK
Content-Type: application/json

{
	"error": 0,
	"ErrorDescription": "all right",

	"articles": [
		{
			"id_article": 23,
			"line": [ 0: "blabla new content"
		},{
			"id_article": 25,
			"line": [ 0: "some article"
		}
	]

}