Unified-Env

An lightweight, zero dependency package to unify node environment variables using strong typings

Table of Contents

• Basic Usage
• Advanced Usage
  • Advanced Env Options
    • Parsing process.env using .env()
    • Parsing process.argv using .argv()
      • Argv Casing and Common Mistakes
    • Parsing an Env File using .file(options)
    • Generate Final Env Object with .generate()
    • Order Matters
• Use Cases
  • Real Life Example
  • Heroku Deployments
  • Use a Validation Script
• Samples
• Credits
• Coming Soon (TODO)

Concept

Unified-Env aims to provide a way to ensure required, valid environment variables using TypeScript for a type-safe API. Problems it solves:

• Adding new env variables locally and forgetting to add them on the server/hosting environment
• Not having required env variables set causing errors at runtime (these can now be caught at start up or compile time)
• Having invalid env variables set
• Not having a central API where all env variables are located and strongly typed

Basic Usage

First, install from npm:

1npm insall unified-env
2# or yarn
3yarn add unified-env

Second, create a central file to use UnifiedEnv (for example, src/environment.ts) and create your env.

1import { UnifiedEnv } from 'unified-env';
2
3const environment = new UnifiedEnv({
4  APP_VAR: true, // `true` = a required, string
5  DB_USER: true,
6  DB_PASSWORD: true,
7  DB_HOST: true,
8  DB_NAME: true,
9  DB_PORT: { required: true, type: Number, acceptableValues: [2000, 3000, 4000] }, // a required number of 2000, 3000, or 4000
10  APP_PROD: { required: true, type: Boolean }, // a required boolean
11  APP_DEFAULT: { required: true, defaultValue: 'app default' } // required with a defaultt value
12})
13  .env() // parse `process.env`
14  .argv() // parse `process.argv`
15  .file({ filePath: './.env' }) // parse an `.env` file (relative path)
16  .generate(); // generate the environment object
17
18export default environment;

The above UnifedEnv will parse the process.env, then process.argv, and then an .env file looking for those variables; and generate a final environment object. It will throw an error if 1) any required variable is missing, 2) there was an error parsing a Boolean or Number value, or 3) a value was not in the listed acceptableValues array. The exported environment constant will be strongly typed to the passed in configuration.

Third, import your environment into other files that need env variables (for example, src/database.ts)

1import { Client } from 'pg';
2import environment from './environment';
3
4/* `environment` will be strongly typed */
5
6const client = new Client({
7  user: environment.DB_USER,
8  host: environment.DB_HOST,
9  database: environment.DB_NAME,
10  password: environment.DB_PASSWORD,
11  port: environment.DB_PORT,
12});
13
14export default client;

See Key Notes under Advanced Usage

*See Use Cases for more pracical examples

Advanced Usage

All keys listed in your UnifiedEnv constructor are the variables that will be typed.

Key Notes

• All keys listed in your UnifiedEnv constructor are the variables that will be typed. See Advanced Env Options
• Order matters when calling .env(), .argv(), and/or .file(), order will matter. See Order Matters
• .file() filepath option is relative to __dirname (ie. where you are calling your root node project from). See Parsing an Env File
• .generate() must be called to generate the file env object
• Only the values passed into UnifiedEnv will be checked, any variables in process.env, process.argv, and/or an .env file that were not listed in the configuration, will NOT be in the environment object returned from generate(). See parsing process.env available.
• Keys are CASE SENSATIVE and not parsed differently for each source. process.env, process.argv and .env file keys are all treated the same. Example; if UnifiedEnv has a configuration item of { MY_KEY: true } and process.argv has --my-key='hello', UnifiedEnv will not match against that key.

Advanced Env Options

There are several advanced configuration options for desired variables.

1const env = new UnifiedEnv({
2  /* acceptable configuration options */
3  MY_VAR: true | {
4    required?: boolean,
5    type?: String | Number | Boolean,
6    defaultValue?: string | boolean | number,
7    acceptableValues?: (string | boolean | number)[],
8    tieBreaker?: 'env' | 'argv' | 'file'
9  }
10}, {
11  logLevel?: 'log' | 'debug' | 'info' | 'warn' | 'error',
12  logger?: ILogger
13});

Each key must be of the following type:

• 1st Argument (expectedEnvVariables - required)
  • true: will default to { required: true, type: String }
  • EnvOption object: all options are optional. note: a blank object will be treated as true
    • required: boolean: If true, an error will be throw when .generate() is called if the variable is not set. If false, no error will be thrown
    • type: String | Number | Boolean: If String (default), the variable will be returned as a string. If Number, the variable will be parsed to a number (an error will be throw if parsing fails). If Boolean, the variable will be parsed to a boolean (an error will be throw if parsing fails)
    • defaultValue: string | boolean | number: If the key has not been set, this default value will be used. Ensure the defaultValue matches the typeof the type option (string is default)
    • acceptableValues: (string | boolean | number)[]: The variable value must be a value found in this array. Ensure the defaultValue matches the typeof the type option (string is default)
    • tieBreaker: 'env' | 'argv' | 'file': The value from the listed tieBreaker will always be used in the event of the same value coming from different sources. Example, process.env has MY_VAR=hello and process.argv has --MY_VAR=goodbye; if MY_VAR has a tieBreaker = 'argv', the value from process.argv will always be used -- even if .env() was called before .argv() (See Order Matters for more details). In this example, MY_VAR will equal 'goodbye'.
  • 2nd argument (configOptions - optional)
    • Object
      • logLevel: 'log' | 'debug' | 'info' | 'warn' | 'error': (default: 'warn') will control what kind of logs are displayed
      • logger: ILogger: (default console) any object that implements an ILogger interface

        • 1interface ILogger {
          2  log (...args: any[]): void;
          3  debug (...args: any[]): void;
          4  info (...args: any[]): void;
          5  warn (...args: any[]): void;
          6  error (...args: any[]): void;
          7};

Parsing process.env using .env()

UnifiedEnv will check the process.env keys for any matching key in the UnifiedEnv configuration. Only keys that match will be parsed.

For example, if the process.env has the following values:

1ENV=prod LOG_LEVEL=debug ts-node my-unified-env.ts

And the UnifiedEnv configuration looks like:

1const env = new UnifiedEnv({ LOG_LEVEL: true }).env().generate();
2export default env;

The exported env will NOT has an ENV property.

Note: this rule applies to both .argv() and .file()

Parsing process.argv using .argv()

As mentioned in the Key Notes section, UnifiedEnv does not handle casing differently for process.argv keys. The reason being twofold:

1. There are plenty of other libraries out there that parse process.argv uniquely. UnifiedEnv is not trying to replicate those.
2. UnifiedEnv aims to be as simple, and straight forward as possible. Having different naming conventions only complicates using this (or any) library.

For process.argv usage, take the following example:

1const env = new UnifiedEnv({
2  LOG_LEVEL: true,
3  DB_USERNAME: true,
4  DB_PASSWORD: true
5})
6  .argv().generate();
7
8export default env;

process.argv would need to have the same matching keys. An example command line call may look like:

1ts-node my-example-env.ts --LOG_LEVEL=info --DB_USERNAME=user123 --DB_PASSWORD=secrect123

Argv Casing and Common Mistakes

Some rules and common mistakes to help understand how UnifiedEnv will parse process.argv:

1# CASING MATTERS (only matches on exact case)
2# white space is trimmed if not in quotes
3
4# SINGLE VALUE
5--DEV # { DEV: 'true' } <- note, this will always be string unless you specific type: Boolean in the config
6--DEV=true # { DEV: 'true' }
7--DEV=false # { DEV: 'false' }
8--DEV true # { DEV: 'true' }
9--DEV false # { DEV: 'false' }
10--DEV is awesome # { dev: 'is awesome' }
11--DB 'some url ' # { dev: 'some url ' }
12
13# MULTI VALUES
14--DEV=true dat --PIE # { DEV: 'true dat', PIE: 'true' }
15--DEV --PIE apple # { DEV: 'true', PIE: 'apple' }
16--DEV --PIE apple with cherry # { DEV: 'true', PIE: 'apple with cherry' }
17
18# COMMON MISTAKES
19  # args must start with `--`
20mistake --OTHER_VALUE # { OTHER_VALUE: 'true' }
21  # if they do not start with `--`, they will be
22  # treated as a string for the previous value
23--DEV true -DB=mongo # { DEV: 'true -DB=mongo' }
24--DEV true DB=mongo # { DEV: 'true DB=mongo' }
25
26  # keys must be in format `--{key}` otherwise they will be treated
27  # as strings for the previous value
28--DEV false -- DB mysql # { DEV: 'false -- DB mysql' }
29-- DEV true # { } # no output since no initial key was found
30
31  # equals cannot have spaces
32--SECRET = 'top secret' # { SECRET: '= \'top secret\'' }
33
34  # missing or mismatching quotes
35--SECRET 'top secret" # { SECRET: "'top secret\"" }
36--SECRET 'top secret # { SECRET: "'top secret" }

Parsing an Env File using .file(options)

• Options: optional object
  • filePath: string: (default: ./.env) relative file path to the .env file. Relative to the starting node script
  • encoding: string: (default 'utf-8') file encoding
  • failIfNotFound: boolean: (default false) if the specified env file was not found, throw an error stopping all processing

UnifiedEnv follows the standard NAME=VALUE configuration format for .env. Notes about parsing:

• It will look for new KEYs on every newline
• It will split on the = character
• It will trim whitespace (unless wrapped in quotes)

An example, project:

.
├── .env
└── src
    └── env.ts

In .env

LOG_LEVELS=debug
ENV=dev

In src/env.ts

1const env = new UnifiedEnv({
2  LOG_LEVEL: true,
3  DB_USERNAME: true,
4  DB_PASSWORD: true
5})
6  .file({ filePath: '../.env' }) // relative path
7  .generate();
8
9export default env;

From root directory, running:

1ts-node src/env.ts

Generate Final Env Object with .generate()

Important notes about .generate():

• .generate() must be called to compile (or "generate") the final env object
• At least one of .env(), .argv(), or .file() must be called first in order for any config to be generated
• Once .generate() has been called, no other function can/should be called on UnifiedEnv
• .generate() can only be called once

Order Matters

Important notes about order:

• Env variables are parsed in the order they were loaded
• Env variables do not "override" variables that have already been set (with the exception of tieBreaker scenarios)

Take this example. Take an env.ts environment file that will have the UnifiedEnv configuration. If called with the following…

1MY_VAR=hello ts-node env.ts --MY_VAR=goodbye

With the following configuration, the .env() MY_VAR value will be used because it is called first:

1const env = new UnifiedEnv({
2  MY_VAR: true
3})
4  .env()
5  .argv()
6  .generate();
7
8// env.MY_VAR === 'hello'
9
10export default env;

With a tieBreaker set to argv, the .argv() MY_VAR value will be always used:

1const env = new UnifiedEnv({
2  MY_VAR: { required: true, tieBreaker: 'argv' }
3})
4  .env()
5  .argv()
6  .generate();
7
8// env.MY_VAR === 'goodbye'
9
10export default env;

Use Cases

Real Life Example

Given the following app structure:

.
├── .env
├── environment.ts
└── app
    └── main.ts

.env

ENV=prod
LOG_LEVEL=info

environment.ts

1const environment = new UnifiedEnv({
2  ENV: { required: true, acceptableValues: ['dev', 'test', 'prod'], tieBreaker: 'env' },
3  DB_PORT: { required: true, type: Number },
4  LOG_LEVEL: { required: false },
5  REFRESH_DB: { required: true, typ: Boolean, defaultValue: true }
6})
7  .env()
8  .argv()
9  .file() // default is '.env'
10  .generate();

src/main.ts

1import environment from '../environment';
2
3/* mock app setup */
4const app = new MyApp({
5  isProd: environment.ENV === 'prod',
6  logLevel: environment.LOG_LEVEL
7});
8
9const db = new DB({
10  port: environment.DB_PORT,
11  refreshDb: environment.REFRESH_DB
12});

Starting the application with the following will provide the necessary variables to UnifiedEnv:

1ENV=prod ts-node src/main.ts --ENV=dev --DB_PORT=3456

UnifiedEnv will generate the following object:

1{
2  ENV: 'prod', // used 'env' tieBreaker
3  DB_PORT: 3456, // from argv
4  LOG_LEVEL: 'info', // from .env file
5  REFRESH_DB: true // from default value
6}

Heroku Deployments

Heroku was the inspiration for UnifiedEnv. It was easier for me to have an .env file in my local working project, but in the Heroku dashboard most environment variables are stored in process.env commandline variables. I didn't want to have production level .env files stored in my repo so I always use those process.env vars.

The issue I would run into would be I add a variable to my local .env file, finish out the feature I was working on (sometimes would take a week or two, push the code to Heroku, and have runtime errors because I forgot to set the new env vars in my test and/or prod Heroku apps. UnifiedEnv helps to solve that problem by allowing validation of env variables before app start up. See Use a Validation Script for more details on how to do that.

Use a Validation Script

Most of us have been working on a new fature locally, add a new env variable, and then forget to add it to the test/production environment. We don't always catch the mistake until our new feature is running in that environment and starts throwing errors.

UnifiedEnv can easily be configured to pre-check our env variables before our app is started. For example, Heroku has a Release Phase where tasks can be configured to run before the application is released. A simple use case for UnifiedEnv to validate env variables before releasing is:

Example app structure:

.
├── src
│   ├── environment.ts
│   └── main.ts
├── package.json
└── Procfile

In src/environment.ts setup our UnifiedEnv configuration:

1const environment = new UnifiedEnv({
2  ENV: { required: true, acceptableValues: ['dev', 'test', 'prod'] },
3  DB_CONN_STR: { required: true },
4  LOG_LEVEL: { required: true, acceptableValues: ['debug', 'info', 'warn', 'error'] }
5})
6  .env()
7  .argv()
8  .file()
9  .generate();
10
11export default environment;

src/main.ts will do application bootstrap, but will import the environment.ts file:

1import environment from './environment';
2// other imports
3
4// bootstrap application, etc

Add a script to package.json. All this script needs to do, is load the src/environment.ts. Heroku will call the script will all the configured variables in the Heroku dashboard.

1{
2  "scripts": {
3    "validate-env": "ts-node src/environment.ts",
4    "start": "ts-node src/main.ts", // example startup script
5    // other scripts
6  }
7}

In Procfile, add a "release" step and the "web" step (See Heroku's Procfile docs):

release: npm run validate-env
web: npm run start

When the package.json validate-env script is run, if any env variables are missing UnifiedEnv will throw an error causing the "release" phase to fail. Heroku will not release the application until that script passes. This is an excellent way to ensure all env variables are present before starting an applicaiton in a server environment.

Samples

There are several samples with corresponding configuration files. To run the samples, clone the repo and install dependencies:

1# clone repo
2git clone https://github.com/djhouseknecht/unified-env.git
3# cd into directory
4cd ./unified-env
5# install dependencies
6npm install # or yarn

Run the following npm scripts:

1# sample using `process.argv`
2npm run sample:argv
3
4# sample using `process.env`
5npm run sample:env
6
7# sample that throws errors
8npm run sample:error
9
10# sample using an `.env` file
11npm run sample:file

Be sure to check out the scripts in package.json and the configuration:

• samples/argv/argv-sample.ts
• samples/env/env-sample.ts
• samples/error/error-sample.ts
• samples/file/file-sample.ts
  • and corresponding .env

Credits

Idea was originally designed to make heroku development and deployments easier. It is loosely based on dotenv and nconf

Coming Soon (TODO)

• create a load() function to push all env variables into the process.env
  • maybe have an exclude list?
  • this will probably be an external function
• IEnvOption
  • add false support for non-required string
  • add altKeys: string[] for alternate keys to look for
• file() -> add .json support
• IEnvOption -> better typings (and validation) for defaultValue and acceptableValues
• utils#validateExpectedVariables() -> write this
• utils#finalTypesMatch() -> write this