Skip to main content

LoopBack 4 November 2019 Milestone Update

ยท 9 min read
Janny Hou

Originally published on

The LoopBack team greeted November with the CASCONxEVOKE conference in Toronto. CASCONxEVOKE is one of Canada's largest combined academic, research and developer conferences. As its speakers and attendees, we had a booth with posters to advocate LoopBack, and also delivered a workshop about developing extensible LoopBack applications. You can check this blog for more details.

For Q4 achievements, we finished 3 epics this month: Inclusion of related models, Deployment guide in a cloud native environment and Support partitioned database in Cloudant connector, and significantly progressed in the Migration, Authentication & Authorization epics.

Keep reading to learn about the recently added features!

Running Repository Tests for Cloudantโ€‹

To ensure our relations test suites work against real databases, we've been adding different kinds of databases to our test environment. This month we added a new repository@loopback/test-repository-cloudant to run shared CRUD and relation tests on Cloudant. You can also set up docker instance easily with our setup script to test out your application. See PR#3968 for more details and try it out if you're interested.

Verifying Relation Type in Metadataโ€‹

Besides supporting inclusion in queries, we now also set constraints to CRUD operations with navigational properties to avoid unexpected errors. For example, if you try to create an instance of TodoList with all Todos it has through the hasMany relation such as:

id: 1,
name: 'my list 1',
{id: 1, description: 'todo 1', todoListId: 1},
{id: 2, description: 'todo 1', todoListId: 2}, // incorrect foreign key

Such requests might be problematic because they might contain incorrect primary key or foreign key. Therefore, with such concerns, request contains navigational properties will be rejected. PR#4148 implements the verification for CRUD methods.

Additionally, in order to ensure that the correct metadata type is being using when it is resolved, we've added tests in PR#4046 and simplified our test setup.

Migration from LoopBack 3โ€‹

We keep incrementally building the migration guide for LoopBack 3 users upgrading to LoopBack 4. In November, we added content for migrating model relations. We explained how to convert a relation defined in LoopBack 3 model JSON files into corresponding LoopBack 4 artifacts.

Authentication and Authorizationโ€‹

Adding Authorization Example and Tutorialโ€‹

In loopback4-example-shopping, the /users/{userId}/orders endpoints are now secured by an authorization system. When a request comes in, the authentication module resolves the user profile and passes it to the authorization module. Then an interceptor retrieves the metadata from the decorated endpoint, and invokes registered authorizers to determine whether the user can perform the operation. We also have a documentation PR in progress that explains the usage of authorization module.

Refactoring Customer Credentialsโ€‹

In LoopBack 3.x, we store users' passwords together with user data in the same model (table), which opens a lot of security vulnerabilities to deal with. For example, when returning user data in HTTP response, the password property must be filtered out and when searching for users, the password must be excluded from the query.

Our example Shopping App used to store user credentials together with other profile information too. In November, we refactored the domain model of our example app and extracted the password property into a new model UserCredentials. Beside the immediate benefits in increased security, this new domain model makes it easier to implement additional features in the future. For example: 2-factor authentication, and the password validation rule forbiding repeated use of the same password.

Context Improvementโ€‹


In PR#4193 we've improved the context/binding with an inspect() method for metadata dumping. ctx.inspect() can now be used to print out the context hierarchy in JSON. This is useful for troubleshooting and rendering in UI. An example snippet of calling the new function:

const ctx = new Context();

Choosing the Right Scopeโ€‹

When creating a binding, you can configure the scope as SINGLETON or TRANSIENT. To know more about how to make the right choice based on requests, see the new documentation in Choosing the right scope.

@service Decoratorโ€‹

A new shortcut decorator @service was introduced to inject an instance of a given service from a binding that matches the service interface. You can inject a service as follows:

class MyController {
constructor(@service(MyService) public myService: MyService) {}

More details of its usage and explanation could be found in Service Decorator

Receive Information of Current Bindingโ€‹

PR#4121 allows a class or provider to receive its own binding information as follow:

export class HelloController {
// If the `bindingKey` is not specified,
// the current binding from the resolution session is injected.
@inject.binding() private myBinding: Binding<string>;
async greet() {
return `Hello from ${this.myBinding.key}`;

Application Life Cycleโ€‹

In PR#4145 we improved states and introduced graceful shutdown for LoopBack applications.

Now an application's states are classified as stable or in process. Operations can only be called at a stable state. Calling a different operation in an in-process state will throw an error. See Application states for details.

The shutdown of application is now controllable by specifying an array of signals in the configuration. For example:

const app = new Application({
shutdown: {
signals: ['SIGINT'],
// Schedule some work such as a timer or database connection
await app.start();

When the application is running inside a terminal, it can respond to Ctrl+C, which sends SIGINT to the process. The application calls stop first before it exits with the captured signal. This gives you a better control when the LoopBack 4 application is running inside a managed container, such as a Kubernetes Pod. See graceful-shutdown for more details.

Partitioned Database in Cloudantโ€‹

We've finished the Partitioned Database Epic in loopback-connector-cloudant by supporting partitioned index, find, and property definition.

Creating Partitioned Indexโ€‹

To create partitioned indexes as the secondary optimization for Cloudant query, you can add index entry in your model configuration with {partitioned: true} like:

Product = db.define('Product', {
name: {type: String},
}, {
forceId: false,
indexes: {
// create a partitioned index for frequently queried
// fields like `name`
'product_name_index': {
partitioned: true,
keys: {
name: 1

Learn more about its signature and usage in Adding Partitioned Index

Partitioned Findโ€‹

When the partition key is discovered in the query's options or filter, Model.find() will invoke the underlying partitioned find to optimize the query. Here are two examples:

Specifying the partition key in the options: Product.find({where: {name: 'food'}}, {partitionKey: 'toronto'});

Or defining a model property as the partition key field. So that the find method could infer the partition key from the query filter:

// Define `city` as the partition key field in model `Product`
Product = db.define('Product', {
id: {type: String, id: true},
name: String,
// partition key field
city: {type: String, isPartitionKey: true},

// `Product.find()` will infer the partition key `toronto`
// from the filter
where: {
city: 'toronto',
name: 'food'

You can learn more about the partitioned query in Performing Partitioned Find

Build Improvementsโ€‹

We upgraded the version of TypeScript to 3.7.2. An application created by the new CLI module ^@loopback/cli@1.26.0 is able to use the latest TypeScript features such as optional chaining and nullish coalescing. A list of new features can be found in TypeScript's release blog.

Now, in the new generated application, the json configuration files are renamed from *.datasource.json to *.datasource.config.json to avoid generating declarations. Explanations see TypeScript issue#34761.

Bug Fixesโ€‹

  • lb4 discover should generate the correct type for property definition. Fixed by PR#4143.

  • @param.path.<primitive_type> generated with lb4 relation considers Wrapper datatypes. Also fixed by PR#4143.

Documentation Improvementsโ€‹

  • We've added a series of tutorials to illustrate how LoopBack can be used as an enabler to build large-scale Node.js applications. If you want to have a deeper understanding of LoopBack and/or to build an application with great flexibility and extensibility, don't miss this tutorial series!

  • In loopback-component-explorer, we added a note in the README file to explain why the module is not affected by the security vulnerabilities in swagger-ui.

  • We've added "Boot" and "Advanced Topics" to the core tutorial in Advanced Recipes and Discover and load artifacts by convention.

  • We've updated the steps of creating model relations to use lb4 relation command in the TodoList tutorial.


  • We added the OpenAPI and gRPC connectors to be a part of our available connectors in PR#558 and PR#906. Now, when a user calls lb4 datasource, they will have OpenAPI and gRPC as options for the connector.

  • The lb4 update command runs against a LoopBack 4 project and checks dependencies against the installed @loopback/cli. Optionally, it updates the dependencies in package.json. Details can be found on page Update generator.

  • In spike story#3770 we came up with a plan to support querying with nested filter in the API Explorer by re-designing the @param.query.object() decorator. The follow-up implementation story is tracked in #2208.

  • We fixed a bug in loopback-datasource-juggler where applyDefaultOnWrites was not being applied in nested objects and arrays. You can find the details in PR#1797.

Community Contributionsโ€‹

Here are some of the highlighted contributions from the community:

  • For the model definition created by running lb4 openapi, we fixed the JavaScript type mapping of date from Date to string. Details see PR#142.

  • A flag useDefaultIdType was introduced in loopback-datasource-juggler to preserve the user provided id(primary key) property against the one generated by connectors. Details see PR#1783.

  • We now reject the the promise properly for create and replaceById when model initialization has errors. Details see PR#1790.

What's Next?โ€‹

If you're interested in what we're working on next, you can check out the December Milestone.

Call to Actionโ€‹

LoopBack's success depends on you. We appreciate your continuous support and engagement to make LoopBack even better and meaningful for your API creation experience. Here's how you can join us and help the project: