Yet another logging package for Go.
The main thing about this package is that it's good for usage in libraries and doesn't involve itself with policy. Essentially, if, as a library, you want to log something, you write this:
var log, Log = xlog.NewQuiet("my.logger.name")
func Foo() {
log.Debugf("Bar")
}
The log
variable is what you use to log, and the Log
variable is exported
from the package and provides methods for controlling the log site. (These are
actually two interfaces to the same object which enforce this pattern.)
The idea is that consuming code can call somepkg.Log to control where it logs to, at what verbosity, etc.
You should instantiate with NewQuiet
if you are a library, as this suppresses
most log messages by default. Loggers created with New
don't suppress
any log messages by default.
xlog uses a traditional Print/Printf interface. It also has the following conveniences:
-
Methods ending in
e
, such asDebuge
, take an error as their first argument and are no-ops if it is nil. -
Fatal
andPanic
call os.Exit(1) and Panic, respectively. Thee
variants of these are no-ops if the error is nil, providing a simple assertion mechanism.
xlog uses syslog severities (Emergency, Alert, Critical, Error, Warning, Notice, Info, Debug) and also provides a Trace severity which is even less severe than Debug. You should generally not emit Alert or Emergency severities from your code, as these are of system-level significance.
You can visit all registered log sites in order to configure loggers programmatically.
Loggers should be named via a dot-separated hierarchy with names in lowercase.
If you have a repository called foo
and a subpackage baz
, naming the logger
for that subpackage foo.baz
might be reasonable.
Loggers are arranged in a hierarchy terminating in the root logger, which is
configured to log to stderr by default. You can create a logger under another
logger using NewUnder
.
Licence
© 2014—2016 Hugo Landau <[email protected]> MIT License