Spine Logging

Spine Logging is a versatile library designed for Kotlin and Java projects, with a potential for multi-platform use.

As for now, only JVM target is supported, with a JavaScript implementation being our priority for future development.

API and implementation are largely inspired by Google Flogger, and the introduction of fluent logging API in SLF4J in v2.0.0.

Current status: Experimental

Please note that this library is still in the experimental phase of development, and hence, its API may undergo significant changes. As such, we advise using this library cautiously in your projects until it has reached a stable release stage.

Simple example

To start logging, only a dependency on spine-logging is needed:

dependencies {
    implementation("io.spine:spine-logging:$version")
}

In this case, the default logging backend will be used. It outputs log records to the console using java.util.logging (JUL) framework.

All logging operations are done with an instance of io.spine.logging.Logger. To get a logger, one can use the following:

1. Make a logging class implement WithLogging interface.
2. Get a logger from LoggingFactory.

The interface provides a default property logger that returns a logger for the implementing class or object:

import io.spine.logging.WithLogging

class Example : WithLogging {
    fun doSomething() {
        logger.atWarning() // Call to the default property of `WithLogging`.
            .log { "..." }
    }
}

LoggingFactory has two methods that return a logger for the enclosing class and for the given KClass:

import io.spine.logging.LoggingFactory

class App {
    private val logger1 = LoggingFactory.forEnclosingClass()
    private val logger2 = LoggingFactory.loggerFor(this::class)

    fun doSomething() {
        check(logger1 === logger2) // There is always one logger per class.
        logger1.atWarning()
            .log { "..." }
    }
}

Logging backends

A logging backend handles an actual output of the logged statement. LogRecord may be printed to the console, be written to a file or sent by the network to some log-aggregating service. It all is up to a chosen backend and its configuration.

The following backends are available:

• io.spine:spine-logging-jul-backend – the default JUL-based backend.
• io.spine:spine-logging-log4j2-backend – Log4j2 backend.

The default backend is supplied along the logging library itself. To use it, one just needs to not supply any other backend. Then JUL backend will be used automatically.

For other backends, put a chosen one to runtimeOnly configuration, and the logging library will discover it in the runtime.

An example usage of Log4j2 backend:

dependencies {
    implementation("io.spine:spine-logging:$version")
    runtimeOnly("io.spine:spine-logging-log4j2-backend:$version")
}

Please note, only one backend implementation should be present in the runtime. Two or more backends will cause an exception because the logging framework will not be able to understand, which one should be used. The default backend doesn't count here.

Logging contexts

A logging context refers to a set of attributes that are attached to all log records while a context is installed. For example, rate limit counters are always attached to the context.

Here is an example of rate limiter context metadata:

import io.spine.logging.WithLogging

class Example : WithLogging {
    fun action() = repeat(12) {
        logger.atInfo()
            .every(7) // Should be emitted once per N calls.
            .log { "Call #$it" } // Rate limiter metadata will be included here.
    }
}

// Produces the following output (without timestamps):
// INFO: Call #0 [CONTEXT ratelimit_count=7 ]
// INFO: Call #7 [CONTEXT ratelimit_count=7 skipped=6 ]

Also, a user can attach its own metadata. For instance, you can attach a user ID to all log records for the current request, or force logging level if the requested URL contains a debug parameter.

Here is an example of how to attach a user ID:

import io.spine.logging.LoggingFactory.singleMetadataKey
import io.spine.logging.WithLogging
import io.spine.logging.context.ScopedLoggingContext

typealias Action = String
typealias User = String

data class Request(
    val action: Action,
    val user: User,
)

class RequestHandler : WithLogging {

    companion object {
        // Metadata is represented with key-value entries.
        private val USER_KEY = singleMetadataKey("user", User::class)
    }

    fun handle(request: Request) = withinUserContext(request.user) {
        logger.atInfo().log { "Handling `${request.action}` request." }
        // All log statements within this block will include `user` info.
    }

    private fun withinUserContext(user: User, action: () -> Unit) = ScopedLoggingContext.newContext()
        .withMetadata(USER_KEY, user)
        .execute(action)
}

Let's send several requests to the created handler:

val handler = RequestHandler()
with(handler) {
    handle(Request(action = "create", user = "Jack"))
    handle(Request(action = "update", user = "Jack"))
    handle(Request(action = "remove", user = "Bill"))
}

// Produces the following output (without timestamps):
// INFO: Handling `create` request. [CONTEXT user="Jack" ]
// INFO: Handling `update` request. [CONTEXT user="Jack" ]
// INFO: Handling `remove` request. [CONTEXT user="Bill" ]

Gradle configuration

The default implementation provides a no-op context. Context metadata is ignored in this case.

To use a logging context, a runtimeOnly dependency for a context implementation should be added.

If your project does not use gRPC, use the following dependency:

dependencies {
    implementation("io.spine:spine-logging:$version")
    runtimeOnly("io.spine:spine-logging-std-context:$version")
}

If your project does use gRPC, add the following dependency:

dependencies {
    implementation("io.spine:spine-logging:$version")
    runtimeOnly("io.spine:spine-logging-grpc-context:$version")
}