AitoAI Config parser

Intent

This library consists of a environment variable parser, which has been used in multiple different microservices at Aito.ai. The library has been adapted from the original sources. If something looks strange or out of place, please let us know.

Benefits

Fail fast

The config parsing is done at startup, which allows for a fail-fast sequence at startup. This has proven to reduce the number of hard to parse runtime errors happening only at the time a variable is referenced the first time.

Auto-completion

The other potential benefit of the approach is that the config parser supports auto-completion for config variables. Unlike simply using 'process.env.VARIABLE' the user can get some help with discoverability of variables, and bugs due to typo's in the variable names.

Type checking

The config parser will allow checking the provided variables. Currently 3 different basic types are supported: strings, numbers and booleans. This could easily be extended with a library like runtypes, which could allow for more customisable validation. (Todo for the future)

Parsing modes

The config parser supports two modes for parsing.

1. development
2. production

In development-parsing the variables are fetched from env variables, env-file, and finally from the defaults-file, in this particular order. Once the variable is found, this value is validated against the rule-set. The default default file is ' .env.defaults and is versioned.

In production-parsing the the "production" values will not use the defaults file. Hence, these variables must be set in the application config. The parsing uses firstly the env-variables, secondly the main config file. The defaults file is not used.

These modes allow storing e.g. local passwords in the defaults, but ensuring they are set in the live config. This also ensures we keep the declared type-contract. Missing values would otherwise be either optional of not, depending on the mode.

Note: The default values are still expected to be a valid value

Using and Contributing

Please feel free to use the library, and report back bugs, findings and improvements.

The library and the source is licensed under the Apache-2.0 license.