Posted by Sam Stern, Developer Programs Engineer

The release of version 11.6.0 of the Google Play services SDK moves a number of popular APIs to a new paradigm for accessing Google APIs on Android. We have reworked the APIs to reduce boilerplate, improve UX, and simplify authentication and authorization.

The primary change in this release is the introduction of new Task and GoogleApi based APIs to replace the GoogleApiClient access pattern.

The following APIs are newly updated to eliminate the use of GoogleApiClient:

Auth - updated the Google Sign In and Credentials APIs.
Drive - updated the Drive and Drive Resource APIs.
Fitness - updated the Ble, Config, Goals, History, Recording, Sensors, and Sessions APIs.
Games - updated the Achievements, Events, Games, Games Metadata, Invitations, Leaderboards, Notifications, Player Stats, Players, Realtime Multiplayer, Snapshots, Turn Based Multiplayer, and Videos APIs.
Nearby - updated the Connections and Messages APIs.

These APIs join others that made the switch in previous releases, such as the Awareness, Cast, Places, Location, and Wallet APIs.

The Past: Using GoogleApiClient

Here is a simple Activity that demonstrates how one would access the Google Drive API using GoogleApiClient using a previous version of the Play services SDK:


public class MyActivity extends AppCompatActivity implements
        GoogleApiClient.OnConnectionFailedListener,
        GoogleApiClient.ConnectionCallbacks {

    private static final int RC_SIGN_IN = 9001;

    private GoogleApiClient mGoogleApiClient;

    @Override
    protected void onCreate(@Nullable Bundle savedInstanceState) {
        super.onCreate(savedInstanceState);

        GoogleSignInOptions options =
               new GoogleSignInOptions.Builder(GoogleSignInOptions.DEFAULT_SIGN_IN)
                        .requestScopes(Drive.SCOPE_FILE)   
                        .build();

        mGoogleApiClient = new GoogleApiClient.Builder(this)
                .enableAutoManage(this, this)
                .addConnectionCallbacks(this)
                .addApi(Auth.GOOGLE_SIGN_IN_API, options)
                .addApi(Drive.API)
                .build();
    }

    // ...
    // Not shown: code to handle sign in flow
    // ...

    @Override
    public void onConnectionFailed(@NonNull ConnectionResult connectionResult) {
        // GoogleApiClient connection failed, most API calls will not work...
    }

    @Override
    public void onConnected(@Nullable Bundle bundle) {
        // GoogleApiClient is connected, API calls should succeed...
    }

    @Override
    public void onConnectionSuspended(int i) {
        // ...
    }

    private void createDriveFile() {
        // If this method is called before "onConnected" then the a   pp will crash,
        // so the developer has to manage multiple callbacks to make this simple
        // Drive API call.
        Drive.DriveApi.newDriveContents(mGoogleApiClient)
            .setResultCallback(new ResultCallback() {
                // ...
            });
    }
}

The code is dominated by the concept of a connection, despite using the simplified "automanage" feature. A GoogleApiClient is only connected when all APIs are available and the user has signed in (when APIs require it).

This model has a number of pitfalls:

Any connection failure prevents use of any of the requested APIs, but using multiple GoogleApiClient objects is unwieldy.
The concept of a "connection" is inappropriately overloaded. Connection failures can be result from Google Play services being missing or from authentication issues.
The developer has to track the connection state, because making some calls before onConnected is called will result in a crash.
Making a simple API call can mean waiting for two callbacks. One to wait until the GoogleApiClient is connected and another for the API call itself.

The Future: Using GoogleApi

Over the years the need to replace GoogleApiClient became apparent, so we set out to completely abstract the "connection" process and make it easier to access individual Google APIs without boilerplate.

Rather than tacking multiple APIs onto a single API client, each API now has a purpose-built client object class that extends GoogleApi. Unlike with GoogleApiClient there is no performance cost to creating many client objects. Each of these client objects abstracts the connection logic, connections are automatically managed by the SDK in a way that maximizes both speed and efficiency.

Authenticating with GoogleSignInClient

When using GoogleApiClient, authentication was part of the "connection" flow. Now that you no longer need to manage connections, you should use the new GoogleSignInClient class to initiate authentication:


public class MyNewActivity extends AppCompatActivity {

    private static final int RC_SIGN_IN = 9001;

    private GoogleSignInClient mSignInClient;

    @Override
    protected void onCreate(@Nullable Bundle savedInstanceState) {
        super.onCreate(savedInstanceState);

        GoogleSignInOptions options =
               new GoogleSignInOptions.Builder(GoogleSignInOptions.DEFAULT_SIGN_IN)
                        .requestScopes(Drive.SCOPE_FILE)
                        .build();

        mSignInClient = GoogleSignIn.getClient(this, options);
    }

    private void signIn() {
        // Launches the sign in flow, the result is returned in onActivityResult
        Intent intent = mSignInClient.getSignInIntent();
        startActivityForResult(intent, RC_SIGN_IN);
    }

    @Override
    protected void onActivityResult(int requestCode, int resultCode, Intent data) {
           super.onActivityResult(requestCode, resultCode, data);

        if (requestCode == RC_SIGN_IN) {
            Task task = 
                    GoogleSignIn.getSignedInAccountFromIntent(data);
            if (task.isSuccessful()) {
                // Sign in succeeded, proceed with account
                GoogleSignInAccount acct = task.getResult();
            } else {
                // Sign in failed, handle failure and update UI
                // ...
            }
        }
    }
}

Making Authenticated API Calls

Making API calls to authenticated APIs is now much simpler and does not require waiting for multiple callbacks.


    private void createDriveFile() {
        // Get currently signed in account (or null)
        GoogleSignInAccount account = GoogleSignIn.getLastSignedInAccount(this);

        // Synchronously check for    necessary permissions
        if (!GoogleSignIn.hasPermissions(account, Drive.SCOPE_FILE)) {
            // Note: this launches a sign-in flow, however the code to detect
            // the result of the sign-in flow and retry the API call is not
            // shown here.
            GoogleSignIn.requestPermissions(this, RC_DRIVE_PERMS, 
                    account, Drive.SCOPE_FILE);
            return;
        }

        DriveResourceClient client = Drive.getDriveResourceClient(this, account);
        client.createContents()
                .addOnCompleteListener(new OnCompleteListener() {
                    @Override
                    public void onComplete(@NonNull Task task) {
                        // ...
                    }
                });
    }

Before making the API call we add an inline check to make sure that we have signed in and that the sign in process gran ted the scopes we require.

The call to createContents() is simple, but it's actually taking care of a lot of complex behavior. If the connection to Play services has not yet been established, the call is queued until there is a connection. This is in contrast to the old behavior where calls would fail or crash if made before connecting.

In general, the new GoogleApi-based APIs have the following benefits:

No connection logic, calls that require a connection are queued until a connection is available. Connections are pooled when appropriate and torn down when not in use, saving battery and preventing memory leaks.
Sign in is completely separated from APIs that consume GoogleSignInAccount which makes it easier to use authenticated APIs throughout your app.
Asynchronous API calls use the new Task API rather than PendingResult, which allows for easier management and chaining.

These new APIs will improve your development process and enable you to make better apps.