log acts as drop-in replacement for std/log and uses the power of
Uber's Zap logging library internally.
It implements the following Zap cores:
A development setup is used by default. Similar to std/log.
All logs are written to stderr.
import "github.com/mattes/log"
func main() {
defer log.Sync()
log.Info("Hello world")
}In production the setup depends on where you want to ship your logs to. Here is an example that ships all logs to Google Stackdriver, as well as Google Error Reporting.
import (
"go.uber.org/zap"
"go.uber.org/zap/zapcore"
"github.com/mattes/log"
gerr "github.com/mattes/log/googleErrorReporting"
gsdr "github.com/mattes/log/googleStackdriver"
)
func init() {
cores := []zapcore.Core{}
{
// Stackdriver core
c := gsdr.NewConfig()
c.LogID = "my-service.v2"
core, err := c.Build()
cores = append(cores, core)
}
{
// Error reporting core
c := gerr.NewConfig()
c.ServiceName = "my-service"
c.ServiceVersion = "v2"
core, err := c.Build()
cores = append(cores, core)
}
// Build Zap logger with options
logger := zap.New(zapcore.NewTee(cores...)).WithOptions(
zap.AddCaller(),
zap.AddCallerSkip(1),
zap.AddStacktrace(zapcore.ErrorLevel),
log.ErrorOutput("stderr"), // for internal errors
log.Sampling(100, 100),
)
// Set global logger
log.Use(logger)
}
func main() {
defer log.Sync()
defer log.CapturePanic() // optional, it logs unhandled panics before crashing
log.Error("Hello Production!")
}Update the config to use a reference of zap#AtomicLevel that you control. It can serve as HTTP handler, too.
Sometimes third-party libraries log on their own with no way of disabling it. With the help of Go modules we can replace third-party logging libraries and instruct them to log through our logging infrastructure.
Some tests require credentials and configration that can't be commited to the repo.
I recommend putting a .env file in each directory with contents like:
export SLACK_URL='https://hooks.slack.com/services/xxx'
Then run test with source .env && go test -v