As noted in my comment, you cannot modify the method set of a type in the return from a function call, so directly extending the behavior of the myStruct
struct is not possible in your "decorator" function as written.
I am actually using this decorator to wrap fmt.Sprintf
, Sprintln
, Sprint
method to get Logger.Debugln
, Logger.Debugf
, Logger.Warning
, ... methods, where Logger
is a self-defined struct, since these functions shares almost the same behavior.
You also will not be able to pass around the fmt.Sprintf
, fmt.Sprintln
and fmt.Sprint
functions in a generic way, as they share different signatures:
func Print(a ...interface{}) (n int, err error)
func Printf(format string, a ...interface{}) (n int, err error)
func Println(a ...interface{}) (n int, err error)
so you will need to wrap these methods separately.
It isn't entirely clear to me how you are seeking to wrap those methods to produce your Logger
type.
If you require each of those methods at the top level of your logger, they will need to be explicitly declared in your logging package or on your Logger
type's interface. You cannot dynamically bind each level function at runtime, despite their obvious similarities; go is a statically-typed language so your types must be declared upfront.
Defining the types explicitly will be the clearest solution, which is also in keeping with a key go proverb: "Clear is better than clever."
If necessary, you can declare each method as a simple wrapper of an internal method on your type which actually performs the printing logic, keeping repetition to a minimum.
Here's a quick example I threw together to show the logic for some Debug
, Debugf
and Debugln
endpoints. You can see the core logic is handled by a generic set of methods on the Logger
type, and you can imagine how endpoints for other log levels could be trivially implemented.
If repetition is really a problem, you could trivially write a code generator to automatically generate the log-level specific methods, calling on the core functionality explicitly declared in the various print functions. Remember you can declare receivers for a type in multiple source files in the same package, allowing some methods to be generated in some source files while others are implemented explicitly elsewhere.
package main
import (
"fmt"
"io"
"os"
)
type Level string
const (
Debug Level = "DEBUG"
Error = "ERROR"
// etc.
)
type Logger struct {
writer io.Writer
prependLevel bool
}
func (l *Logger) print(level Level, msg string) {
var s string
if l.prependLevel {
s = string(level) + ": "
}
s += msg
fmt.Fprint(l.writer, s)
}
func (l *Logger) printf(level Level, format string, a ...interface{}) {
l.print(level, fmt.Sprintf(format, a...))
}
func (l *Logger) println(level Level, msg string) {
l.printf(level, "%s
", msg)
}
func (l *Logger) Debug(msg string) {
l.print(Debug, msg)
}
func (l *Logger) Debugf(format string, a ...interface{}) {
l.printf(Debug, format, a...)
}
func (l *Logger) Debugln(msg string) {
l.println(Debug, msg)
}
func main() {
logger := Logger{os.Stderr, true}
logger.Debugln("A plain message with a new line")
logger.Debugf(
("The most customizable log entry with some parameters: " +
"%s %v
"),
"myStr", false,
)
logger.Debug("A bare print statement")
}
Playground link.
Third-party packages
In case you haven't yet done so, there are many third-party logging packages in Go which I would recommend considering, or at least assessing their interfaces to determine how others have approached similar tasks.
Logrus is a Go logger with levels, for example. You can see in their source files how they explicitly declare their interface for each function at each log level, while using a more comprehensive version of the method above to reduce repetition in the core logging implementation: Interface Formatter implementations