Onelog

Dead simple, super fast, zero allocation logger for Golang

Description

Onelog is a dead simple but very efficient JSON logger. It is one of the fastest JSON logger out there. Also, it is one of the logger with the lowest allocation.

It gives more control over log levels enabled by using bitwise operation for setting levels on a logger.

It is also modular as you can add a custom hook, define level text values, level and message keys.

Go 1.9 is required as it uses a type alias over gojay.Encoder.

It is named onelog as a reference to zerolog and because it sounds like One Love song from Bob Marley :)

Programming language: Go

License: MIT License

Tags: Logging     Structured Logging     Performance    

Latest version: v0.14.0

README

Onelog

Onelog is a dead simple but very efficient JSON logger. It is one of the fastest JSON logger out there. Also, it is one of the logger with the lowest allocation.

It gives more control over log levels enabled by using bitwise operation for setting levels on a logger.

It is also modular as you can add a custom hook, define level text values, level and message keys.

Go 1.9 is required as it uses a type alias over gojay.Encoder.

It is named onelog as a reference to zerolog and because it sounds like One Love song from Bob Marley :)

Get Started

go get github.com/francoispqt/onelog

Basic usage:

import "github.com/francoispqt/onelog"

func main() {
    // create a new Logger
    // first argument is an io.Writer
    // second argument is the level, which is an integer
    logger := onelog.New(
        os.Stdout, 
        onelog.ALL, // shortcut for onelog.DEBUG|onelog.INFO|onelog.WARN|onelog.ERROR|onelog.FATAL,
    )
    logger.Info("hello world !") // {"level":"info","message":"hello world"}
}

Levels

Levels are ints mapped to a string. The logger will check if level is enabled with an efficient bitwise &(AND), if disabled, it returns right away which makes onelog the fastest when running disabled logging with 0 allocs and less than 1ns/op. See benchmarks

When creating a logger you must use the | operator with different levels to toggle bytes.

Example if you want levels INFO and WARN:

logger := onelog.New(
    os.Stdout, 
    onelog.INFO|onelog.WARN,
)

This allows you to have a logger with different levels, for example you can do:

var logger *onelog.Logger

func init() {
    // if we are in debug mode, enable DEBUG lvl
    if os.Getenv("DEBUG") != "" {
        logger = onelog.New(
            os.Stdout, 
            onelog.ALL, // shortcut for onelog.DEBUG|onelog.INFO|onelog.WARN|onelog.ERROR|onelog.FATAL
        )
        return
    }
    logger = onelog.New(
        os.Stdout, 
        onelog.INFO|onelog.WARN|onelog.ERROR|onelog.FATAL,
    )
}

Available levels:

• onelog.DEBUG
• onelog.INFO
• onelog.WARN
• onelog.ERROR
• onelog.FATAL

You can change their textual values by doing, do this only once at runtime as it is not thread safe:

onelog.LevelText(onelog.INFO, "INFO")

Hook

You can define a hook which will be run for every log message.

Example:

logger := onelog.New(
    os.Stdout, 
    onelog.ALL,
)
logger.Hook(func(e onelog.Entry) {
    e.String("time", time.Now().Format(time.RFC3339))
})
logger.Info("hello world !") // {"level":"info","message":"hello world","time":"2018-05-06T02:21:01+08:00"}

Context

Context allows enforcing a grouping format where all logs fields key-values pairs from all logging methods (With, Info, Debug, InfoWith, InfoWithEntry, ...etc) except for values from using logger.Hook, will be enclosed in giving context name provided as it's key. For example using a context key "params" as below

logger := onelog.NewContext(
    os.Stdout, 
    onelog.INFO|onelog.WARN,
    "params"
)

logger.InfoWithFields("breaking news !", func(e onelog.Entry) {
    e.String("userID", "123455")
}) 

// {"level":"info","message":"breaking news !", "params":{"userID":"123456"}}

This principle also applies when inheriting from a previous created logger as below

parentLogger := onelog.New(
    os.Stdout, 
    onelog.INFO|onelog.WARN,
)


logger := parentLogger.WithContext("params")
logger.InfoWithFields("breaking news !", func(e onelog.Entry) {
    e.String("userID", "123455")
}) 

// {"level":"info","message":"breaking news !", "params":{"userID":"123456"}}

You can always reset the context by calling WithContext("") to create a no-context logger from a context logger parent.

Logging

Without extra fields

Logging without extra fields is easy as:

logger := onelog.New(
    os.Stdout, 
    onelog.ALL,
)
logger.Debug("i'm not sure what's going on") // {"level":"debug","message":"i'm not sure what's going on"}
logger.Info("breaking news !") // {"level":"info","message":"breaking news !"}
logger.Warn("beware !") // {"level":"warn","message":"beware !"}
logger.Error("my printer is on fire") // {"level":"error","message":"my printer is on fire"}
logger.Fatal("oh my...") // {"level":"fatal","message":"oh my..."}

With extra fields

Logging with extra fields is quite simple, specially if you have used gojay:

logger := onelog.New(
    os.Stdout, 
    onelog.ALL,
)

logger.DebugWithFields("i'm not sure what's going on", func(e onelog.Entry) {
    e.String("string", "foobar")
    e.Int("int", 12345)
    e.Int64("int64", 12345)
    e.Float("float64", 0.15)
    e.Bool("bool", true)
    e.Err("err", errors.New("someError"))
    e.ObjectFunc("user", func(e Entry) {
        e.String("name", "somename")
    })
}) 
// {"level":"debug","message":"i'm not sure what's going on","string":"foobar","int":12345,"int64":12345,"float64":0.15,"bool":true,"err":"someError","user":{"name":"somename"}}

logger.InfoWithFields("breaking news !", func(e onelog.Entry) {
    e.String("userID", "123455")
}) 
// {"level":"info","message":"breaking news !","userID":"123456"}

logger.WarnWithFields("beware !", func(e onelog.Entry) {
    e.String("userID", "123455")
}) 
// {"level":"warn","message":"beware !","userID":"123456"}

logger.ErrorWithFields("my printer is on fire", func(e onelog.Entry) {
    e.String("userID", "123455")
}) 
// {"level":"error","message":"my printer is on fire","userID":"123456"}

logger.FatalWithFields("oh my...", func(e onelog.Entry) {
    e.String("userID", "123455")
}) 
// {"level":"fatal","message":"oh my...","userID":"123456"}

Alternatively, you can use the chain syntax:

logger.InfoWith("foo bar").
    Int("testInt", 1).
    Int64("testInt64", 2).
    Float("testFloat", 1.15234).
    String("testString", "string").
    Bool("testBool", true).
    ObjectFunc("testObj", func(e Entry) {
        e.Int("testInt", 100)
    }).
    Object("testObj2", testObj). // implementation of gojay.MarshalerJSONObject
    Array("testArr", testArr). // implementation of gojay.MarshalerJSONArray
    Err("testErr", errors.New("my printer is on fire !")).
    Write() // don't forget to call this method! 

Accumulate context

You can create get a logger with some accumulated context that will be included on all logs created by this logger.

To do that, you must call the With method on a logger. Internally it creates a copy of the current logger and returns it.

Example:

logger := onelog.New(
    os.Stdout, 
    onelog.ALL,
).With(func(e onelog.Entry) {
    e.String("userID", "123456")
})

logger.Info("user logged in") // {"level":"info","message":"user logged in","userID":"123456"}

logger.Debug("wtf?") // {"level":"debug","message":"wtf?","userID":"123456"}

logger.ErrorWithFields("Oops", func(e onelog.Entry) {
    e.String("error_code", "ROFL")
}) // {"level":"error","message":"oops","userID":"123456","error_code":"ROFL"}

Change levels txt values, message and/or level keys

You can change globally the levels values by calling the function:

onelog.LevelText(onelog.INFO, "INFO")

You can change the key of the message by calling the function:

onelog.MsgKey("msg")

You can change the key of the level by calling the function:

onelog.LevelKey("lvl")

Beware, these changes are global (affects all instances of the logger). Also, these function should be called only once at runtime to avoid any data race issue.

Benchmarks

For thorough benchmarks please see the results in the bench suite created by the author of zerolog here: https://github.com/rs/logbench

The benchmarks data presented below is the one from Uber's benchmark suite where we added onelog.

Benchmarks are here: https://github.com/francoispqt/zap/tree/onelog-bench/benchmarks

Disabled Logging

	ns/op	bytes/op	allocs/op
Zap	8.73	0	0
zerolog	2.45	0	0
logrus	12.1	16	1
onelog	0.74	0	0

Disabled with fields

	ns/op	bytes/op	allocs/op
Zap	208	768	5
zerolog	68.7	128	4
logrus	721	1493	12
onelog	1.31	0	0
onelog-chain	68.2	0	0

Logging basic message

	ns/op	bytes/op	allocs/op
Zap	205	0	0
zerolog	135	0	0
logrus	1256	1554	24
onelog	84.8	0	0

Logging basic message and accumulated context

	ns/op	bytes/op	allocs/op
Zap	276	0	0
zerolog	141	0	0
logrus	1256	1554	24
onelog	82.4	0	0

Logging message with extra fields

	ns/op	bytes/op	allocs/op
Zap	1764	770	5
zerolog	1210	128	4
logrus	13211	13584	129
onelog	971	128	4
onelog-chain	1030	128	4