A no-code (not only) RESTfull API
Start directly:
npx @norest/cli
-or-
Download the CLI via npm:
npm i -g @norest/cli
And start it by running the command
norest
The default port is 3030 and the base path is api. You can start by POSTing an fragment to that endpoint:
POST http://localhost:3030/api
{
"name": "foo product",
"price": 10,
"#_product": {}
}
That product is then accessible via the endpoint http://localhost:3030/api/product and readable by any user.
Authentication is by default disabled. Use
norest --auth.enabled=trueto enable authentication (See the authentication and authorization chapter and read about all flags in the @norest/cli readme).
The idea behind noREST is as simple as the idea behind noSQL, it should provide data that is representational but without the boundaries of a fixed schema. Therefore so called index fragments are added to the data itself, which allow the data to be requested, created and updated via the REST paradigm. Index fragments belongs to the data and can be added at any time (they start with #_).
As an example we will re-use the given product example from getting started. Imagine an E-Commerce API which holds thousands of products. After a while the project manager decides that they want to add featured products, that should be shown on the start page. Traditional some sort of link or flag would be introduced in the schema:
{
"_id": 1,
"name": "bar product",
"price": 10,
"featured": true
}
In a traditional approach a new endpoint would be implemented and deployed and then the frontend engineers start using it. However with noREST a more top-down approach is possible by simply adding a new index fragment to each featured product:
PATCH /product/1
{
"#_featured": {}
}
Afterwards all featured products can be requested with a simple GET request to the featured endpoint:
GET /featured
[{
"_id": 1,
"name": "bar product",
"price": 10,
"#_featured": {},
"#_product": {}
}]
The frontend engineers can start right away implementing the feature and align the back-office application (to allow adding the fragment) and the store front website to show the featured products. All without any backend deployment.
The index fragment is called index fragment, as database connector automatically apply a index for fast resolving the data. Also the fragment is only visible if you are authenticated (read more about visibility in the Concept -> authentication chapter).
Now you might questioning: Why this might be useful? The idea developed after one of that endless discussions about how an API should be shaped. The noREST paradigm tries to eliminate that discussion to speedup the develop process by using an top-down approach. The user interface can be designed first and the API schema can be shaped afterwards.
Beside the development speed argument, noREST could be useful in other scenarios as well:
- JAM stack applications as private payloads can be used in server side rendering approach (See Concept -> Private and public payload)
- headless system where the content creator is trusted and knows what it does
- programs that needs flexibility as they consist of many different data structures (e.g. IoT applications)
Overall it's an approach for data, where the data producer is trustable. It's not useful for REST endpoint that allow create, update and delete operations for non authenticated users.
Following the core concepts of the noREST paradigm that allow to store and request data.
An index fragment is a simple fragment that gets indexed on the database. At the same time it allows to request that data via the route. E.g. all entities of an index fragment called #_product: {} can be requested via {{url}}/product.
References allow to manage relationship between data-sets. Therefore you can add a reference to any data by simply adding a new fragment which contains the @_ prefix and the name of the index fragment which should be referenced. Extending the above e-commerce example to allow to add your products to categories then is quite simple:
POST /api/categories
{
"id": 2
"name": "Category",
"#_categories": {}
"@_products": [1]
}
Afterwards all products belonging to an given category can simply be requested by callin GET /api/categories/2 will resolve the product with the id 1.
You can either reference multiple other resources in array for a 1:n link or directly for a 1:1 reference.
References have the benefit that they are updated automatically, when the status of the referenced data changes. When for example the product with the id 1 is removed, the category with the id 2 will return no reference any more.
Values that are readonly start with an _ (underscore). They cannot be changed with an PATCH request. There are two special underscore keys: _id is the unique identifier. It must be unique over all index fragments. You can leave it empty to generate a unique one. Additional there is the "_": {} metadata entry which is returned by the API to give additional information about the response. It can also not be changed.
Each create and update request also creates or updates the metadata of an entity. The metadata ar always stored in the _ fragment and containing information about the person who created (owner) and who changed it last. The owner/changedBy are taken from the JWT sub entry (if not configured otherwise).
{
"_": {
"owner": "john",
"created": "2020-08-27T18:32:46.223Z",
"changedBy": "doo",
"changed": "2020-08-28T18:32:46.223Z"
}
}If authentication is disabled (default behavior), the owner and changer will always be
anonymous.
A list request additional contains the information about the pagesize:
{
"_": {
"total": 2059,
"skip": 0,
"limit": 100
}
}Skip and limit can be used for pagination. Every list request supports a limit, skip and orderBy URL parameter. For example the following request will return the 2nd most highest rating:
GET /api/ratings?limit=1&skip=1%orderBy&orderBy=rating desc
Authentication is not handled by the API itself. Instead noREST just have two states:
- unauthenticated: If no JWT is set you can read all endpoints
- authenticated: If a JWT is set you can CRUD all endpoints.
The API just takes any JWT passed as cookie, search query parameter or authentication bearer header and uses the sub as the current user. There is no validation done on the JWT. The API expects that it is valid and therefore the stack needs to ensure that it is validated. This can be done via a middleware, an API gateway or an service like Auth0.
By default all endpoints are readable. Sometimes you want to make data private. Therefore you can simply add the private data to the index fragment. The data inside the index fragment is only exposed if the user is authenticated (JWT is set) and authorized (read or owner rights of crudo). The following data can be seen by an authorized user only:
"#_product": {
"overpriced": true
}To give certain users certain access to certain endpoints a crudo entry can be added to the JWT. The term crudo stands for create, read, update, delete and own. "Own" is a special kind which allows full access to entries you are owning. An example crudo entry looks the following:
{
"sub": "foo",
"crudo": {
"user": "o",
"order": "cru",
}
}Note:
- By default any authenticated user has full CRUD rights to all endpoints and you must opt-out to revoke the rights. So if there is another endpoint called products, the user foo is allow to create, read, update and delete items on this endpoint. It is a good strategy to change that strategy in the finalizing step of the API.
- Read in this case allows to read even the private payload of a response.
To get realtime events, each noREST implementation implement websockets. These websocket are just read streams (updates, delete and additions can only made via http). Following an example on how to connect to a socket:
let socket = new WebSocket('ws://localhost:3030/api/products');
socket.onopen = function() {
socket.onmessage = function(msg) {
console.log(JSON.parse(msg.data));
};
};The returned message then is equal to the http result and gets pushed on each change to the index fragment collection. The metadata is enhanced by information about the change that triggers the new message to be send. Create operations are appended at the end of the list. Authentication needs to be send via header or url parameter.
As mentioned before, noREST is intended to be used for rapid development. However, noREST supports a fixed mode which disallow to add new index fragments to the API.
A GraphQL support is planed, but not yet done.
noREST is more meant as an paradigm, which could be implemented in any language. However to proof the paradigm we introduced a reference implementation in this repository and some tooling to help to get you started:
- @norest/nestjs: A noREST reference implementation for node.js written with the help of the nestjs framework.
- @norest/cli: A small tool which allows to configure and run the application quickly from your command line.
- @norest/playground: An interactive playground to try out the capabilities of noREST. Visit notonly.rest to try it out.