Skip to content

Latest commit

 

History

History
611 lines (479 loc) · 21.7 KB

DEVELOPMENT.md

File metadata and controls

611 lines (479 loc) · 21.7 KB

Development guide

See the CHEATSHEET for day-to-day development.

For detailed information about the implementation of this application, read this guide:

Configuration

This application can be configured through environment variables. You can find a list of all available variables and examples in the .env.sample file.

In the code, all this configuration is centralized in the config/index.js file. All environment variables should be retrieved there, and any code which needs configuration should require this file to obtain it.

Dotenv

The dotenv library is used to load environment variables from a file on application startup (when not in production).

Which files are loaded depends on the environment, as determined by the $NODE_ENV environment variable. For example, for the development environment (the default), the following files are loaded by dotenv if they exist:

  • .env.development (highest precedence)
  • .env.development.defaults
  • .env (lowest precedence)

The first value found is used for a given variable. For example, if a variable is defined in .env.development, it takes precedence over the same variable in .env.development.defaults and .env. Note that if an environment variable is already set in your environment, it will not be overwritten.

File structure

The main directories in this project are:

Directory Description
app Express.js application.
app/api API routes, controllers and documentation.
app/models Sequelize models.
app/services External services (e.g. mailing).
app/utils Utility functions used throughout the application.
app/express.js Factory function for the Express.js application and its middlewares.
app/server.js Function to start the application's HTTP server.
bin/www Executable that starts the application.
collada2gltf The COLLADA2GLTF binary.
commands Node.js scripts meant to be run from the command line.
config Application configuration.
coverage Test coverage reports (only exists if you have run the tests with coverage enabled).
docker Files used by Docker or Docker Compose.
locales Translation files.
spec Utilities for automated tests (the tests themselves are in the app directory along with the code).
spec/assertions Custom Chai assertions.
spec/expectations More complex expectations usually composed of multiple assertions.
spec/fixtures Data population functions to set up the initial state of tests.
spec/utils Utility functions used throughout test code.
spec/config.js Test-related configuration.
utils Utilities shared by different types of code (e.g. scripts and tests).

You will find subdirectories under app/api which group together files related to an API resource. They generally contain:

File(s) Description
foo.routes.js Express.js route definitions.
foo.controller.js Route implementations.
foo.openapi.yml OpenAPI documentation for these routes.
schemas/*.schema.json Related JSON schemas.
foo.*.spec.js Automated tests for these routes.

In general, prefer the <what>.<type...>.js naming convention:

  • foo.controller.js
  • foo.req.schema.js

As opposed to:

  • fooController.js
  • foo_Controller.js
  • foo_controller.js

When a filename or part of it is composed of multiple words, prefer hyphenating:

  • foo.cool-thing.js
  • lemon-chicken.schema.json
  • awesome-stuff.js

As opposed to:

  • foo.coolThing.js
  • lemonChiken.schema.json
  • awesomeStuff.js

Linting

This project's code is linted with ESLint. You can find the linting configuration in the .eslintrc.json file.

Linting is triggered automatically when developing with npm run dev, using the --fix option to automatically fix simple problems. You can also trigger linting manually with npm run lint.

Database migrations

The evolution of the database is managed through Sequelize migrations. Migrations files are in the db/migrations directory of this repository.

You can use the Sequelize Command-Line Interface to manage migrations. Here's a few useful commands when developing with Docker Compose:

Command Description
npm run compose:sequelize-cli -- db:migrate Run pending migrations.
npm run compose:migrate Alias for the previous command.
npm run compose:sequelize-cli -- db:migrate:status List the status of all migrations.
npm run compose:sequelize-cli -- db:migrate:undo Reverts a migration.
npm run compose:sequelize-cli -- db:migrate:undo:all Reverts all migrations ran.
npm run compose:sequelize-cli -- migration:generate --name add-new-stuff Generates a new migration file.

When developing locally, you can use sequelize-cli directly using npx, for example:

Command Description
npx sequelize-cli db:migrate Run pending migrations.
npm run migrate Alias for the previous command.
npm run migrate:wait Wait for the database to be reachable, then run pending migrations.
npx sequelize-cli db:migrate:status List the status of all migrations.

Migrations can be written in two ways:

Logging

This application uses the winston library for logging. It is configured in config/logger.js.

Note that our Express application creates a logger for each request which will include a randomly generated request ID in log lines. In application code, you should use the getLogger utility function from app/utils/express.js in order to retrieve the correct logger:

const { getLogger } = require('path/to/utils/express');

exports.someControllerFunction = function(req, res) {
  getLogger(req).info('Something cool happened');
  // ...
};

If you are not in the context of a request, you may import the main application logger directly from config/logger.js instead:

const logger = require('path/to/config/logger');

logger.info('Something cool happened');

You may log error objects, which will serialize their stack trace:

try {
  // Do something dangerous.
} catch (err) {
  logger.error(err);
}

Logger methods such as .info(msg) and .error(msg) in the examples above correspond to the available log levels which are, from most to least important:

  • error - Unrecoverable errors which will cause the current request to fail or the application to crash. Note that "errors" which occur during normal operation, such as validation errors, should not use this log level.
  • warn - Warnings about conditions that are not critical but are not expected to happen during normal operation and may warrant investigation. Again, validation errors should not use this log level.
  • info - Information about important application lifecycle events that occurred, e.g. a new resource was created or an existing resource was updated. All controller functions which modify domain entities should have an information log.
  • http - Used for automatic HTTP request logging. Should generally not be called in application code.
  • verbose - Detailed or additional information normally not shown in the logs. "Errors" that occur during normal operations, such as validation errors, should use this log level.
  • debug - Information intended to help debug issues, normally not shown in the logs.
  • silly - Trace information, when debug is not detailed enough.

You should generally use info (appears at the default log level) and verbose (does not appear at the default log level) in application code, or warn to indicate a condition that should not happen in normal operation.

Use debug and silly for debugging complex code.

API routes that simply retrieve information do not necessarily need to perform any logging.

Automated tests

This project includes an automated test suite for its API, implemented with the following libraries:

  • Mocha to run the tests.
  • Chai to make assertions.
  • Sinon to create spies, stubs and mocks.
  • [Chance][change] to randomly generate data.
  • SuperTest to test HTTP calls to the API.

If you are running the application with Docker Compose, you can run the test suite with:

Command Description
npm run compose:test Run tests.
npm run compose:test:coverage Run tests with a coverage report (slower).
npm run compose:test:debug Run tests in debug mode (very verbose logs).

If you are running the application locally, you can use these commands instead:

Command Description
npm test Run tests locally.
npm run test:coverage Run tests locally with a coverage report (slower).
npm run test:debug Run tests in debug mode (very verbose logs).
npm run test:watch Automatically run tests every time a change is made.

The test coverage report is saved to coverage/index.html.

Test environment

The test environment has custom settings defined in the .env.test.defaults file (which is under version control). Since this file takes precedence over your .env file, you can create a .env.test file if you need to override these settings. This file will be local to your machine and not under version control.

Random seed

Mocha runs the tests in a deterministic order. However, this project's tests use the Chance library to generate random data. The library is seeded with a random value which is printed to the console every time you run the test suite:

$> npm test
Using random seed 290610
...

To reproduce an exact test run, including the same "random data", you may use the $SEED environment variable:

$> SEED=290610 npm test
Using seed 290610 from environment
...

This also works with the other test commands like npm run compose:test.

Custom assertions

A number of custom assertions have been implemented and added to Chai to improve the readability of this project's tests. You will find their implementation in the spec/assertions directory.

OpenAPI assertions

This project includes assertions which can verify that:

  • An HTTP request you are making is valid according to the OpenAPI documentation.

    const req = {
      method: 'POST',
      url: '/auth/local/login',
      body: {
        email: '[email protected]',
        password: 'changeme'
      }
    };
    
    // Check that the POST /auth/local/login operation is documented. The content
    // type of the request body is checked. The documentation must also include a
    // JSON schema, and the body of the request must match that schema.
    expect(req).to.matchRequestDocumentation();
  • An HTTP response you have received is valid according to the OpenAPI documentation.

    const req = {
      method: 'POST',
      url: '/auth/local/login',
      body: {
        email: '[email protected]',
        password: 'changeme'
      }
    };
    
    const res = await testHttpRequest(app, req);
    
    // Check that the received response is one of the documented responses for the
    // POST /auth/local/login operation. The status code and content type are
    // checked. If the response has a body, the documentation of the response must
    // include a JSON schema and the response body must match that schema.
    expect(res).to.matchResponseDocumentation();

This serves as a kind of contract testing to make sure that the application and its documentation are in sync.

These assertions are implemented using a dereferenced version of this API's OpenAPI document, i.e. with all $ref pointers resolved. This dereferenced document is saved to tmp/openapi.dereferenced.json every time the tests are run, in case you need to check its contents to debug the assertions.

A few options are available to customize the documentation assertions:

  • If you are knowingly making a request with an invalid body (e.g. to test validations), set the invalidBody option to true:

    const req = {
      method: 'POST',
      url: '/auth/local/login',
      body: {
        // Invalid login data.
        email: 'foo',
        password: ''
      }
    };
    
    // Check that the POST /auth/local/login operation is documented, and that the
    // body of the request does NOT match the associated schema. (The assertion
    // will fail if the body is valid.)
    expect(req).to.matchRequestDocumentation({ invalidBody: true });

API documentation

This application is a REST API which is documented according to the OpenAPI specification. This documentation is served by the API using Swagger UI:

The base OpenAPI document is app/api/openapi.yml. This is not a full OpenAPI document: it includes other documentation fragments under the app/api directory, using a custom !include YAML tag and a custom $$merge JSON property which are explained below.

The full OpenAPI document is compiled and saved to app/generated/openapi.yml. This is done automatically when developing and testing through commands such as npm start, npm run dev or npm test. In production, it must be done with npm run openapi before starting the server.

A dereferenced version of the OpenAPI document is also saved to app/generated/openapi.dereferenced.yml along with the previous file. This version is used programmatically for schema validations.

Similarly, the application's various JSON schemas must be assembled and saved to app/generated/schemas.json. This is also done automatically when developing and testing. In production, it must be done with npm run schemas before starting the server.

To precompile all the required files for production, you may use npm run precompile. Note that this and the above commands requires the installation of development dependencies (with npm ci or npm install), which can then be removed with npm prune --production.

Custom !include YAML tag

You may use the !include FILE tag in this project's OpenAPI document (or in included YAML fragments):

---
foo: !include ./bar.yml
baz:
  - corge
  - !include ./**/*.grault.yml

It behaves as follows:

  • If FILE is a path to a single file, relative to the file in which the include tag was found, then that file is loaded and its contents injected at that location. An absolute path may also be used.

  • If FILE is a glob pattern which matches one or several files relative to where the include tag was found, those files are all loaded, merged together, and the resulting object or array is injected at that location.

    The matched files must either all represent a YAML sequence (array) or a YAML mapping (object). Sequences and objects cannot be mixed. The merge is shallow (sequences are simply concatenated, mappings are simply merged at the top level).

Including JSON schemas

You may also use the variant !include jsonSchemasById FILE, in this case, files will be found as explained above, but their contents will be transformed. For example, this schema:

{
  "$id": "FooBar",
  "type": "string"
}

Will be transformed to:

{
  "FooBar": {
    "$id": "FooBar",
    "type": "string"
  }
}

This is intended to allow merging multiple JSON schemas together into the /components/schemas object of the OpenAPI document, each identified by its $id property.

For example, if you have 3 JSON schemas like the one above, with $id properties Foo, Bar and Baz, you would normally have to include them like this:

components:
  schemas:
    Foo: !include ./foo.schema.json
    Bar: !include ./bar.schema.json
    Baz: !include ./baz.schema.json

This forces you to have duplication (the schema ID is both in the schema itself and in the OpenAPI document), and is more verbose. Instead, you can include all those schemas in one line like this:

components:
  schemas: !include jsonSchemasById ./**/*.schema.json

The assembled OpenAPI document will contain:

components:
  schemas:
    Foo:
      # Contents of the Foo schema
    Bar:
      # Contents of the Bar schema
    Baz:
      # Contents of the Baz schema

Custom $$merge JSON property

You can merge content from other files into the JSON schemas referenced by this project's OpenAPI document. If a JSON object contains a $$merge key:

{
  "$$merge": {
    "file": "./other-file.json",
    "pointer": ""
  },
  "a": "value",
  "other": {
    "data": 42
  }
}

And assuming the ./other-file.json file referenced in this example contains the following JSON:

{
  "a": "value (overriden)",
  "other": {
    "moreData": 24
  }
}

Then reading the original JSON file will produce:

{
  "a": "value (overriden)",
  "other": {
    "data": 42,
    "moreData": 24
  }
}

The merge algorithm is the one provided by Lodash's merge function (see https://lodash.com/docs/4.17.15#merge).

The "pointer" key is optional. If not specified, the whole JSON document referenced by "file" is merged. If a pointer is specified, it must be a JSON pointer to a subset of the JSON you want to merge. Whether merging either a whole file or a subset, the merged value MUST be a JSON object, not any other kind of JSON value.

The intended purpose of this feature is to merge properties from one JSON schema into another to avoid duplication.

COLLADA2GLTF

This application includes COLLADA2GLTF, a tool to convert COLLADA files to the glTF format. Updating this tool is done manually:

Sample data

Available to smapshot team members:

  • Switch drive - binary database dump

    Restore a binary database dump when using Docker Compose:

    pg_restore -h localhost -U postgres -p 5434 -d smapshot smapshot.backup

    Restore an SQL or binary database dump locally:

    psql -h localhost -U postgres -d smapshot -f smapshot.sql
    pg_restore -h localhost -U postgres -d smapshot smapshot.backup
  • Switch drive - sample images.

    Unzip the contents of the data.zip file into the public/data directory in this repository.