A type-safe, FP, Scala config library.
libraryDependencies += "com.github.japgolly.clearconfig" %%% "core" % "<ver>"
There are plenty of config libraries out there, right? This library is pure FP, type-safe, super-simple to use, highly composable and powerful, yada yada yada... All true but it's biggest and most unique feature is actually:
CLARITY.
Haven't we all had enough of crap like:
-
changing an environment variable setting, pushing all the way though to an environment, testing and then discovering that your expected change didn't occur. Was the new setting picked up? What setting did it use? Where did it come from?
-
after hours of frustration: "That setting isn't even used any more?! Why the hell is it still all over our deployment config?! Why didn't person X magically know to remove this specific piece of text in this big blob of text in this completely separate deployment repo at the same time they made their code change?"
This library endeavours to provide clarity. When you get an instance of your config, you also get a report that describes:
- where config comes from
- how config sources override other sources
- what values each config source provided
- what config keys are in use
- what the total, resulting config is
- which config is still hanging around but is actually stale and no longer in use
(sample report below)
Here's a pretty common scenario that will serve as a decent introduction.
We have an app which has the following config:
import java.net.URL
final case class DatabaseConfig(
port : Int,
url : URL,
username : String,
password : String,
schema : Option[String])
Let's define how we populate our config from the outside world...
import japgolly.clearconfig._
import cats.implicits._
object DatabaseConfig {
def config: ConfigDef[DatabaseConfig] =
(
ConfigDef.getOrUse("PORT", 8080),
ConfigDef.need[URL]("URL"),
ConfigDef.need[String]("USERNAME"),
ConfigDef.need[String]("PASSWORD"),
ConfigDef.get[String]("SCHEMA")
).mapN(apply)
}
Great, now let's define where we want to read config from.
At this point you also need to decide which effect type to use.
You'd typically use something like IO
but for simplicity,
we'll just use Id
and opt-out of safe FP.
import cats.Id
def configSources: ConfigSources[Id] =
ConfigSource.environment[Id] > // Highest priority
ConfigSource.propFileOnClasspath[Id]("/database.props", optional = true) > //
ConfigSource.system[Id] // Lowest priority
Now we're ready to create a real instance based on the real environment.
val dbCfg: DatabaseConfig =
DatabaseConfig.config
.run(configSources)
.getOrDie() // Just throw an exception if necessary config is missing
Done! But so far we're not using the most important feature of the library: the report.
Let's get and print out a report at the end.
- We'll remove env & system from unused keys to keep the report small seeing as it's just a demo.
- We'll also prefix all keys by
POSTGRES_
to make it look a bit more realistic.
val (dbCfg, report) =
DatabaseConfig.config
.withPrefix("POSTGRES_")
.withReport
.run(configSources)
.getOrDie() // Just throw an exception if necessary config is missing
println(report
// Remove env & system columns from the unused section of the report
.mapUnused(_.withoutSources(ConfigSourceName.environment, ConfigSourceName.system))
// Show the full report
.full
)
Sample output:
4 sources (highest to lowest priority):
- Env
- cp:/database.properties
- System
- Default
Used keys (5):
+-------------------+------+-------------------------+---------+
| Key | Env | cp:/database.properties | Default |
+-------------------+------+-------------------------+---------+
| POSTGRES_PASSWORD | | Obfuscated (1C02B9F6) | |
| POSTGRES_PORT | 4000 | | 8080 |
| POSTGRES_SCHEMA | | | |
| POSTGRES_URL | | http://localhost/blah | |
| POSTGRES_USERNAME | | demo | |
+-------------------+------+-------------------------+---------+
Unused keys (1):
+----------------+-------------------------+
| Key | cp:/database.properties |
+----------------+-------------------------+
| POSTGRES_SCHMA | public |
+----------------+-------------------------+
From the above report we can immediately observe the following:
- Which sources override other sources; the report columns (left-to-right) respect this
- We'll be running at port 4000 and the reason for that is there's an override set by the environment
- There's a typo in our
database.properties
;POSTGRES_SCHMA
should bePOSTGRES_SCHEMA
- The password value has been hashed for the report. This still allows you to compare the hash between envs or time to determine change without compromising the value.
-
Simplest and most common methods:
ConfigDef.get [A](key: String) // provides an Option[A] - optional config ConfigDef.getOrUse[A](key: String, default: A) // provides an A - optional config with default ConfigDef.need [A](key: String) // provides an A - mandatory config; error if not provided
-
To define your own type of config value, create an implicit
ConfigValueParser
. Example:sealed trait MyBool case object Yes extends MyBool case object No extends MyBool implicit val myBoolParser: ConfigValueParser[MyBool] = ConfigValueParser.oneOf[MyBool]("yes" -> Yes, "no" -> No) .preprocessValue(_.toLowerCase) // Make it case-insensitive
-
Call
.secret
on yourConfigDef
to force it to be obfuscated in the report. The default (implicit) report settings already obfuscate keys that contain substrings likepassword
,credential
, andsecret
. OverrideconfigReportSettings
if required. -
Shell-style Comments (beginning with
#
) are automatically removed from config values. Create your own implicitConfigValuePreprocessor
to customise this behaviour. -
Keys can be modified after the fact. Eg.
ConfigDef.get("A").withPrefix("P_").withKeyMod(_.replace('_', '.'))
is equivalent toConfigDef.get("P.A")
. This also works when aConfigDef
is a composition of more than one key, in which case they'll all be modified. -
There is special DSL to create
A => Unit
functions to configure a mutable object (which you typically use when working with a Java library)ConfigDef.consumerFn[A](...)
. There is an example below: -
More... (explore the source)
You typically compose using Applicative
, give the composite a prefix,
then use (nest) it in some higher-level config.
For example, this Scala code...
import cats.syntax.apply._
import japgolly.clearconfig._
import java.net.{URI, URL}
import redis.clients.jedis.JedisPoolConfig
case class AppConfig(postgres: PostgresConfig, redis: RedisConfig, logLevel: LogLevel)
object AppConfig {
def config: ConfigDef[AppConfig] =
( PostgresConfig.config,
RedisConfig.config,
ConfigDef.getOrUse("log_level", LogLevel.Info)
).mapN(apply)
.withPrefix("myapp.")
}
case class PostgresConfig(url: URL, credential: Credential, schema: Option[String])
object PostgresConfig {
def config: ConfigDef[PostgresConfig] =
( ConfigDef.need[URL]("url"),
Credential.config,
ConfigDef.get[String]("schema"),
).mapN(apply)
.withPrefix("postgres.")
}
case class Credential(username: String, password: String)
object Credential {
def config: ConfigDef[Credential] =
( ConfigDef.need[String]("username"),
ConfigDef.need[String]("password"),
).mapN(apply)
}
case class RedisConfig(uri: URI, credential: Credential, configurePool: JedisPoolConfig => Unit)
object RedisConfig {
def poolConfig: ConfigDef[JedisPoolConfig => Unit] =
ConfigDef.consumerFn[JedisPoolConfig](
_.get("block_when_exhausted", _.setBlockWhenExhausted),
_.get("eviction_policy_class_name", _.setEvictionPolicyClassName),
_.getOrUse("fairness", _.setFairness)(true),
_.get("jmx_enabled", _.setJmxEnabled),
_.get("jmx_name_base", _.setJmxNameBase),
_.get("jmx_name_prefix", _.setJmxNamePrefix),
_.get("lifo", _.setLifo),
_.get("max_idle", _.setMaxIdle),
_.get("max_total", _.setMaxTotal),
_.get("max_wait_millis", _.setMaxWaitMillis),
_.get("min_evictable_idle_time_millis", _.setMinEvictableIdleTimeMillis),
_.getOrUse("min_idle", _.setMinIdle)(2),
_.get("num_tests_per_eviction_run", _.setNumTestsPerEvictionRun),
_.get("soft_min_evictable_idle_time_millis", _.setSoftMinEvictableIdleTimeMillis),
_.get("test_on_borrow", _.setTestOnBorrow),
_.get("test_on_create", _.setTestOnCreate),
_.get("test_on_return", _.setTestOnReturn),
_.get("test_while_idle", _.setTestWhileIdle),
_.get("time_between_eviction_runs_millis", _.setTimeBetweenEvictionRunsMillis)
)
def config: ConfigDef[RedisConfig] =
( ConfigDef.need[URI]("uri"),
Credential.config,
poolConfig.withPrefix("pool."),
).mapN(apply)
.withPrefix("redis.")
}
sealed trait LogLevel
object LogLevel {
case object Debug extends LogLevel
case object Info extends LogLevel
case object Warn extends LogLevel
implicit def configValueParser: ConfigValueParser[LogLevel] =
ConfigValueParser.oneOf[LogLevel]("debug" -> Debug, "info" -> Info, "warn" -> Warn)
.preprocessValue(_.toLowerCase)
}
will read the following properties:
myapp.postgres.password
myapp.postgres.schema
myapp.postgres.url
myapp.postgres.username
myapp.redis.password
myapp.redis.uri
myapp.redis.username
myapp.redis.pool.block_when_exhausted
myapp.redis.pool.eviction_policy_class_name
myapp.redis.pool.fairness
myapp.redis.pool.jmx_enabled
myapp.redis.pool.jmx_name_base
myapp.redis.pool.jmx_name_prefix
myapp.redis.pool.lifo
myapp.redis.pool.max_idle
myapp.redis.pool.max_total
myapp.redis.pool.max_wait_millis
myapp.redis.pool.min_evictable_idle_time_millis
myapp.redis.pool.min_idle
myapp.redis.pool.num_tests_per_eviction_run
myapp.redis.pool.soft_min_evictable_idle_time_millis
myapp.redis.pool.test_on_borrow
myapp.redis.pool.test_on_create
myapp.redis.pool.test_on_return
myapp.redis.pool.test_while_idle
myapp.redis.pool.time_between_eviction_runs_millis
myapp.log_level
If you like what I do —my OSS libraries, my contributions to other OSS libs, my programming blog— and you'd like to support me, more content, more lib maintenance, please become a patron! I do all my OSS work unpaid so showing your support will make a big difference.