phediverse/mastodon-rest

Mastodon API REST client

dev-master 2017-04-21 19:44 UTC

This package is not auto-updated.

Last update: 2024-11-10 04:27:39 UTC


README

This is a PHP HTTP client, using the Guzzle HTTP library, to the Mastodon federated social network REST API.

Installation

You'll need PHP 7.0+ and Composer to get this running. Got 'em? Run:

composer require phediverse/mastodon-rest

to import the library into your project. Then, wherever you use it, make sure you've included Composer's autoloader, via

require __DIR__ . 'vendor/autoload.php'; // assuming you're in the same directory as the vendor dir Composer created

Getting Authenticated

With a few exceptions (see Unauthenticated Endpoints below), you need an access token, generated for a valid user from a valid registered application on the instance you're pointing at, to do anything useful with the API. Both the application and access tokens are long-lived at this point; once you get one you don't need to keep going back for a given user. As such, different classes handle auth-related tasks than actual endpoint activity.

Registering an App

First, you need to register an application. You'll need a name and a redirect URI, and you'll need to decide what scopes the application should default to. Information on scopes is available at https://github.com/tootsuite/mastodon/blob/master/docs/Using-the-API/OAuth-details.md.

Once you have your app, you can json_encode() the object to store its credentials for later use.

use Phediverse\MastodonRest\{Auth\AppRegisterClient, Auth\Scope, Resource\Application};

$registerClient = AppRegisterClient::forInstance('social.targaryen.house');
$app = $registerClient->createApp('My Phediverse Reader', 'https://example.com/oauth_landing', [Scope::READ]);

If you're planning to use the application with logins directly (rather than the proper OAuth 2 way via an Authorization Code grant), or you're okay with folks copy-pasting an Auth Code into your app from their browser, you can use the Resource\Application::REDIRECT_NONE URI rather than a standard HTTP URL. Additionally, if you don't specify a scope on app create, the client will automatically ask for all three scopes for you, aka Scope::ALL. Or, to put it another way:

$anotherApp = $registerClient->createApp('No Redirect, All The Power', Application::REDIRECT_NONE);
file_put_contents('instance_creds.json', json_encode($anotherApp));

Logging In Via Authorization Code Grant

NOTE: The examples below use built-in PHP functions for things like request handling; if you're using this library within a (micro-)framework, use that framework's request/response handling methods instead.

Starting The Process

Now that you've got your app, you can go through the OAuth2 Authorization Code grant process. To start, let's pull our app out of its configuration JSON, and drop it into an AuthClient instance.

$app = \Phediverse\MastodonRest\Resource\Application::fromJsonConfig(file_get_contents('instance_creds.json'));
$authClient = \Phediverse\MastodonRest\Auth\AuthClient::forApplication($app);

Then let's figure out where to redirect our users, adding in a random state value to avoid cross-site request forgery (CSRF) attacks. Mastodon will hand us back state as a query string parameter when it redirects back to us.

$url = $authClient->getAuthCodeUrl($state = bin2hex(random_bytes(12)));
// record $state to the user's session or similar
header('Location', $url); // redirect the user to the OAuth2 Auth Code URL

Note that, while state is recommended by the OAuth2 RFC, it's not necessary, and it's pointless if you don't have a real redirect in place. In that case, you can omit that parameter in the call, or set it to null if you need to use the second parameter: an array of requested scopes for the access token. If you leave off the second parameter, as we did above, you'll get whatever scopes you asked for when you originally set up the app. This is a bit different than the raw API behavior when you leave off scopes, which gives you back a token with only read access.

Finishing The Process

Once the user has signed into the Mastodon instance and allowed your app access, they'll be redirected back to you with code in the query string, as well as state if you provided it above. If your app used the not-a-redirect, the user will get the authorization code in heir browser window instead.

Assuming that you have a redirect landing, there's one more step to get the access token once they land on your page:

// since we're in a different request than where we started this process...
$app = \Phediverse\MastodonRest\Resource\Application::fromJsonConfig(file_get_contents('instance_creds.json'));
$authClient = \Phediverse\MastodonRest\Auth\AuthClient::forApplication($app);

// verify state here; it'll be in $_GET['state']

$accessToken = $authClient->finishAuthCodeRequest($_GET['code']); // string

Congratulations! You now have an access token for that user for that Mastodon instance, which you can use with the main API client class...or any other Mastodon API client, for that matter.

Logging In Via Password Grant

If you're using this library for personal use, and as such don't mind user credentials passing through, or being stored in, your system, the flow's a bit simpler, and doesn't have to happen across multiple requests:

// still need your app and an AuthClient
$app = \Phediverse\MastodonRest\Resource\Application::fromJsonConfig(file_get_contents('instance_creds.json'));
$authClient = \Phediverse\MastodonRest\Auth\AuthClient::forApplication($app);

$accessToken = $authClient->login('your@email.address', 'SuperSecretP4$$w0rd'/*, override scopes here */);

As with the Authorization Code flow, if you don't specify any scopes, you'll get a token that uses the default scopes that you specified when you set up your app.

Methods And Resources

Now that you've got your access token, you can set up the main API client instance:

$client = \Phediverse\MastodonRest\Client::build('social.targaryen.house', $accessToken);

Let's start by getting your own account's display name:

$account = $client->getAccount(/* defaults to your ID; can put someone else's ID in here too */);
echo $account->getDisplayName(); // Your Display Name

Resources can also reference other resources directly, e.g.

$instance = $account->getInstance(); // Instance resource

You can also serialize a resource, to either JSON or PHP (which uses a tweaked version of the JSON representation under the hood), and use the client object to bring the resource back from its serialized form. note that serializing a resource will force it to finish downloading (see The Turbo Button for more info on that). The client will inject itself on deserialization, so pulling related resources will work at that point.

I also try to follow the original API response format pretty closely on the JSON side, so JSON-encoding a resource will get you something very similar (same key names etc.) to what the Mastodon API spits out.

$serializedAccount = json_encode($accout);

$deserializedAccount = $client->deserialize($serializedAccount);
echo $deserializedAccount->getId(); // your ID

Unauthenticated Endpoints

Unauthenticated endpoints don't use an access token, but they're few and far between. The main one of interest is the instances endpoint, which we demo here. It uses the technique described in The Turbo Button to get instance information significantly faster than you would by doing all the calls at once.

$client = \Phediverse\MastodonRest\Client::build('mastodon.network', 'ACCESS_TOKEN');
$hosts = ['mastodon.xyz', 'icosahedron.website', 'sealion.club', 'cybre.space', 'toot.cat', 'toot.cafe'];

/** @var \Phediverse\MastodonRest\Resource\Instance[] $instances */
$instances = array_map(function($host) use ($client) { // requests are started here, in parallel!
    return $client->getInstance($host);
}, array_combine($hosts, $hosts));

foreach ($instances as $hostname => $instance) { // everything in this loop will finish around the same time
    echo $hostname . ($instance->isMastodon() ? (': ' . $instance->getName()) : ' is not a Mastodon instance') . "\n";
}

// we can also pull the name for the host we specified in client setup, as it's the default
echo "Default client: " . $client->getInstance()->getName() . "\n";

The Turbo Button

Parallel Requests

Under the hood, the client tries to block as little as possible, only forcing resolution of an HTTP request (well, HTTPS...that's the default if you don't specify scheme in a hostname) when you ask for something that requires an HTTP response. This may not seem like a big deal, but it means that the client will parallelize requests as much as it can if it knows about them early enough. So ask for resource objects early, and pull info out of those resource objects as late as possible if you have other stuff to do in your app, and things will magically go faster!

I may tweak things further to allow for more direct manipulation of the underlying promises so you can let the system know when a bunch of requests don't necessarily need to resolve in order, for even more speed. But that's for another day.

If you want to force a resource block until fully downloaded, call its resolve() method. That method can be chained.

A few catches:

  1. The login endpoints, whether for completing an Auth Code grant or for logging in via a username and password, don't do anything asynchronous, since you're getting an access token back. If folks want to async-ify that bit, that can be a task for a later date.
  2. Async runs one level deep in most cases at this point; if you request a related resource before the original resource loads, you'll block until the original resource load completes. I (or some other contributor) will fix that bit later.

Caching

The client caches GET requests, including referencing the same resource if multiple requests for the same one are in flight, to avoid naive use from pounding the server on the other end. If you need to clear the cache for any reason, use $client->clearCache() for that; see the code for information on arguments. You can also bypass the cache for a request by setting the $useCache parameter on Client calls to false.

Once I add methods for updating a given resource, updates will automatically refresh or remove the associated resource in the cache, depending on what can be done without making another request to the server.

Contributing

Feel free to fork/PR on this, though I may be nitpicky about either code style (this follows PSR-1, PSR-2, PSR-4, and PSR-7 standards) or performance-related lazy loading concerns (when in doubt, look at how the existing code lazy-loads all the things).

Tests will be built in due time; the use of Guzzle's ClientInterface to construct this library's Client class is largely to aid in mocking up Mastodon's API so we can unit-test the client without instance availability.

Code is licensed MIT. I am not affiliated with the main Mastodon dev team in any meaningful way.

You can drop me an email at iansltx@gmail.com (and that's how you should report security issues...NOT on the public issue tracker...public issue tracker's fine for other library-related issues though), or find me on the Fediverse at @iansltx@social.targaryen.house.