Environment Variables


  1. EDN Based Configuration
  2. Managing Environment Variables
  3. Default Environment Variables
  4. The Config Namespace
  5. Environment Specific Code

Luminus aims to facilitate developing 12 factor style applications. The 12 factor approach states that the configuration should be kept separate from the code. The application should not have to be packaged differently for each environment that it's deployed in.

The environment variables are managed by the cprop library. The configuration is represented by a map. The map is constructed by aggregating variables from multiple sources in the environment. The sources include EDN configuration, shell variables, and Java system properties.

EDN Based Configuration

The library will first look for a config.edn file on the resource path. This will be used as the base configuration for the application. An external configuration file can be specified using the conf Java option at runtime: -Dconf=prod-config.edn.

The configuration placed in config.edn should consist of a map such as the following:

{:port 4000}

This configuration will be merged on top of the configuration found on the resource path. The library uses a deep merge strategy, so any nested structures will be merged intelligently.

Managing Environment Variables

The System/getProperties will be merged on top of the configuration found on the resource path and the optional EDN configuration file.

java -Ddatabase-url="jdbc:postgresql://localhost/app?user=app_user&password=secret" -jar app.jar

The variable names are converted into Clojure style keywords. The variables are lowercased and _ characters are used to indicate nesting, while . characters are converted to - characters.

  • -Dport=3000 -> {:port 3000}
  • -Dnrepl-port=7000 -> {:nrepl-port 7000}
  • -Ddatabase-url="jdbc:h2:./guestbook_dev.db" -> {:database-url "jdbc:h2:./guestbook_dev.db"}
  • -Dio_http.max.connections=10 -> {:io {:http-max-connections 10}}

Any environment variables found in System/getenv will be merged last. These variables are parsed using the following strategy:

  • PORT=3000 -> {:port 3000}
  • NREPL_PORT=7000 - {:nrepl-port 7000}
  • DATABASE_URL="jdbc:h2:./guestbook_dev.db" -> {:database-url "jdbc:h2:./guestbook_dev.db"}
  • IO__HTTP_MAX_CONNECTIONS="{:value 10}" -> {:io {:http-max-connections {:value 10}}}

Note that the _ is converted to -, while __ is used to indicate nesting for shell variables. These conventions can be mixed as seen with IO__HTTP_MAX_CONNECTIONS.

See the official documentation for further details.

Default Environment Variables

Luminus projects use the following environment variables by default:

  • PORT - HTTP port that the application will attempt to bind to, defaults to 3000
  • NREPL_PORT - when set the application will run the nREPL server on the specified port, defaults to 7000 for development
  • DATABASE_URL - the URL for the database connection
  • APP_CONTEXT - used to specify an optional context for the routes in the application

The Config Namespace

The variables are populated in the env map that's found in the <app>.config namespace that looks as follows:

(ns <app>.config
  (:require [cprop.core :refer [load-config]]
            [cprop.source :as source]
            [mount.core :refer [args defstate]]))

(defstate env :start (load-config

The configuration will load the environment variables from the known sources and merge it with the command line arguments populated by clojure.tools.cli in the <app>.core/start-app function. The resulting configuration is a map that can be accessed as seen in the example below:

(ns <app>.db.core
  (:require [<app>.config :refer [env]]))

(def database-url
  (-> env :database :url))

Environment Specific Code

Some code, such as development middleware for showing stacktraces in the browser, is dependent on the mode the application runs in. For example, we'd only want to run the above middleware during development and not show stacktraces to the client in production.

Luminus uses env/dev/clj and env/prod/clj source paths for this purpose. By default the source path will contain the <app>.env namespace that has the environment specific configuration. The dev config looks as follows:

(ns <project-ns>.env
  (:require [selmer.parser :as parser]
            [clojure.tools.logging :as log]
            [<project-ns>.dev-middleware :refer [wrap-dev]]))

(def defaults
   (fn []
     (log/info "\n-=[app started successfully using the development profile]=-"))
   :middleware wrap-dev}

The config references the <app>.dev-middleware namespace found in the same source path. Any development specific middleware should be placed there.

Meanwhile, the prod config will not

(ns <project-ns>.env
  (:require [clojure.tools.logging :as log]))

(def defaults
   (fn []
     (log/info "\n-=[app started successfully]=-"))
   :middleware identity})

Only the middleware defined in the <app>.middleware namespace is run during production.