snakelet

snakelet

Ultra minimal go config lib based on viper.

Goals:

• simplicity => correctly validate once and return a nice statically typed struct that contains all the config
• testability => dependency to config library only in main package. Everywhere else use pure functions and params

How?:

• define whichever complex type you want for your config and setup which params are required and which are not
• call InitAndLoad only once in main package and unmarshal to a struct that is statically typed against the config type, reusing nice viper default for precedence (env -> file -> default)
• crash fast if required fields hasn't been set or if config file cannot be parsed to the config types (ex: random string to int)
• if no error, forget about viper and snakelet, only use this unmarshaled struct everywhere

Instead of using the lib directly, the source code itself can be used for inspiration as to how to use viper internally to achieve the goals set above.

Usage

TL;DR:

func main() {
	// Get config struct and default variables
	conf := GetDefaultConfig()

	// this will mutate the `conf` variable:
	err := snakelet.InitAndLoad(conf)

	// Example with params usage:
	// err := snakelet.InitAndLoadWithParams(conf, "/opt/company/project.yaml", "company-project")

	// Will throw an error because some param are required
	if err != nil {
		fmt.Println("Unable to init and load config: %w", err)
	}
	// Pass `conf` or a subset of this variable as param throughout the rest of the code so you don't have any dependency on snakelet or viper whatsoever.
	// This also contributes to improving testability of your code (you can code with pure funcitons):
	// someServerComponent := server.NewServer(conf.Server)
	// someDbComponent := database.NewDatabase(conf.Database)
}

See the example directory for a full working example.

Also checkout the test file and the source code.

License

Copyright (c), 2022, CIDgravity.

All this work is licensed under permissive open-source license.

Except written otherwise in particular files, this work is dual-licensed under both Apache v2 and MIT.