Skip to content

Latest commit

 

History

History
271 lines (182 loc) · 10.1 KB

README.md

File metadata and controls

271 lines (182 loc) · 10.1 KB

Common Hosted Email Service

Table of Contents

  1. Application Dependencies
  2. General Code Layout
  3. Installation
  4. Configuration
  5. API Usage
  6. General Design
  7. Templating

Application Dependencies

The CHES API uses the following dependencies from NPM:

Authentication & Password Management

  • keycloak-connect - Keycloak Node.js adapter (npm)

Networking

  • api-problem - RFC 7807 problem details (npm)
  • express - Server middleware (npm)
  • nodemailer - SMTP mail library (npm)

Logging

  • express-winston - HTTP request logger (npm)
  • winston - General log framework (npm)

Templating

  • nunjucks - Jinja2 style templating language (npm)

Queueing and Persisting

  • bull - Redis-based queue for Node (npm)
  • knex - Multi-dialect query builder (npm)

General Code Layout

The codebase is separated into a few discrete layers:

  • components - Business logic layer - the majority of useful functionality resides here
  • docs - Contains OpenAPI 3.0 Yaml specification and ReDoc renderer
  • routes - Express middleware routing

Quickstart Guide

In order for the application to run correctly, you will need to ensure that the following have been addressed:

  1. All node dependencies have been installed and resolved
  2. You have a Redis-compatible memory store available to connect to.
  3. Environment configurations have been set up
  4. You have a 'ches' Postgres database to connect to

Installation

Node Application

As this is a Node application, please ensure that you have all dependencies installed as needed. This can be done by running npm install.

Redis

In order to run this microservice locally, you must have a Redis 3.2 compatible memory store available to connect to. This can be achieved in many ways depending on your platform, such as through Docker, or installing directly onto your machine. Visit https://redis.io/download to get a copy of the binaries if you are on a Unix machine or to acquire the Docker image to run locally.

For Windows users who wish to install Redis directly onto your machine, there is currently no recent Windows binary officially available from Redis Labs. In lieu of that, we can leverage Memurai instead, which is a Redis 5.0 compatible distribution for Windows platforms. You can acquire the binaries for that at https://www.memurai.com/get-memurai.

In order to view and manipulate Redis, you can either acquire a Redis compatible CLI, or get a GUI tool. We suggest installing the GUI tool redis-commander for managing Redis.

One-off execution:

npx redis-commander -p 8888

Global installation:

npm i redis-commander -g
redis-commander -p 8888

Visit http://localhost:8888 for access. Take note of how you access your Redis as you will need that information for later configuration steps.

Configuration

Configuration management is done using the config library. There are two ways to configure:

  1. Look at custom-environment-variables.json and ensure you have the environment variables locally set.
  2. Create a local.json file in the config folder. This file should never be added to source control.
  3. Consider creating a local-test.json file in the config folder if you want to use different configurations while running unit tests. This file will be necessary because local.json takes precedence over test.json.

For more details, please consult the config library documentation.

Environment Variables

Environment Variable Description
DB_DATABASE Database name
DB_HOST Database hostname
DB_USERNAME Database username
DB_PASSWORD Database password
KC_CLIENTID Keycloak Client username
KC_CLIENTSECRET Keycloak Client password
KC_PUBLICKEY Keycloak Public key signature for JWT validation
KC_REALM Associated Keycloak realm
KC_SERVERURL Base authentication url for Keycloak
REDIS_CLUSTERMODE Run Redis in Cluster mode. Options: yes, no
REDIS_HOST URL to access Redis
REDIS_PASSWORD The Redis password
SERVER_ATTACHMENTLIMIT Maximum attachment size the API will accept
SERVER_BODYLIMIT Maximum body length the API will accept
SERVER_LOGFILE Writes logs to specific file location if defined
SERVER_LOGLEVEL Server log verbosity. Options: silly, verbose, debug, info, warn, error
SERVER_PORT Port server is listening to
SERVER_SMTPHOST The SMTP server this app will leverage

Database Connection

The CHES API requires a postgres database First create an empty database named 'ches' (your db connection parameters go in your local.json config file) Then can create the db schema by running fron the /app directory:

npm run migrate

Commands

After addressing the prerequisites, the following are common commands that are used for this application.

Run the server with hot-reloads for development

npm run serve

Run the server

npm run start

Run your tests

npm run test

Lints files

npm run lint

API Usage

This API is defined and described in OpenAPI 3.0 specification. When the API is running, you should be able to view the specification through ReDoc at http://localhost:3000/api/v1/docs (assuming you are running this microservice locally). A hosted instance of the API can be found at: https://ches.nrs.gov.bc.ca/api/v1/docs

General Design

The standard /email endpoint is relatively straightforward due to effectively being a passthrough to NodeMailer. However, the merging endpoints /email/merge and /email/merge/preview are a bit more involved.

Concepts

In order to provide unique templating results to multiple email destinations, we have the concept of a Context. A Context is a freeform JSON object which consists of key-value pairs. Its sole purpose is to provide a key-value mapping repository between an inline templated variable on a template string and what is the intended output after the values are replaced.

The email merge API has a One to Many relationship between a template string and a context. While there can be many contexts that exist, the API expects to only have one template string. This relationship is modeled this way because typically users will want to have a standard template for batch emails, but will want to replace certain parts of text with their own variable content based on whom it is getting issued to.

In order for a template to be successfully populated, it requires a context object which should contain the variables which will be replaced. For the most part, Nunjucks is intended to behave as a glorified string-replacement engine. In the event the Context object has extra variables that are not used by a Template, nothing happens. You can expect to see blank spots where the templated value should be at.

IMPORTANT: All keys in the Context object (variable names in the template) MUST contain only alphanumeric or underscore.

{
  "this_is_fine": {
    "thisIsGood": "good key/variable",
    "thisIs_Good_2": "fine key/variable"
  },
  "this is $%&*$!": "bad key/variable"
}

Templating

We currently leverage the Nunjucks library for templated variable replacement. Its syntax is similar to the well-used Jinja2 library from Python. We will outline the most common use cases and examples for the templating engine below. For full details on templating, refer to the Nunjucks documentation at https://mozilla.github.io/nunjucks/templating.html.

In general, the Nunjucks templating engine allows variables to be in-line displayed through the use of double curly braces. Suppose you wanted a variable foo to be displayed. You can do so by adding the following into a template:

{{ foo }}

Nunjucks also supports complex nested objects in the Context as well. You can lookup properties that have dots in them just like you would in Javascript. Suppose for example you have the following context object and template string:

Context

{
  "something": {
    "greeting": "Hello",
    "target": "World"
  },
  "someone": "user"
}

Template String

"{{ something.greeting }} {{ someone }} content {{ target }}"

You can expect the template engine to yield the following:

"Hello user content World"

Finally, if a value resolves to either undefined or null, nothing is rendered. Suppose you have the following context object and template string:

Context

{
  "void": "abyss"
}

Template String

"{{ verb }} into the {{ void }} and the {{ void }} will {{ verb }} back at you."

You can expect the template engine to yield the following:

" into the abyss and the abyss will  back at you."

You also have the ability to do simple transformations onto variables before they are rendered. These filters may be invoked by the use of a pipe operator (|). These filters may be able to take parameters, and can be chained. For a more comprehensive list of potential filters, check https://mozilla.github.io/nunjucks/templating.html#builtin-filters.

Suppose you have the following context object and template string:

Context

{
  "foo": "bar"
}

Template String

"{{ foo | upper }} everything"

You can expect the template engine to yield the following:

"BAR everything"