Can I create a function that must only be used with defer? - go

Question

For example:
package package
// Dear user, CleanUp must only be used with defer: defer CleanUp()
func CleanUp() {
// some logic to check if call was deferred
// do tear down
}
And in userland code:
func main() {
package.CleanUp() // PANIC, CleanUp must be deferred!
}
But all should be fine if user runs:
func main() {
defer package.CleanUp() // good job, no panic
}
Things I already tried:
func DeferCleanUp() {
defer func() { /* do tear down */ }()
// But then I realized this was exactly the opposite of what I needed
// user doesn't need to call defer CleanUp anymore but...
}
// now if the APi is misused it can cause problems too:
defer DeferCleanUp() // a defer inception xD, question remains.

Answer 1

Alright, per OPs request and just for laughs, I'm posting this hacky approach to solving this by looking at the call stack and applying some heuristics.
DISCLAIMER: Do not use this in real code. I don't think checking deferred is even a good thing.
Also Note: this approach will only work if the executable and the source are on the same machine.
Link to gist: https://gist.github.com/dvirsky/dfdfd4066c70e8391dc5 (this doesn't work in the playground because you can't read the source file there)
package main
import(
"fmt"
"runtime"
"io/ioutil"
"bytes"
"strings"
)
func isDeferred() bool {
// Let's get the caller's name first
var caller string
if fn, _, _, ok := runtime.Caller(1); ok {
caller = function(fn)
} else {
panic("No caller")
}
// Let's peek 2 levels above this - the first level is this function,
// The second is CleanUp()
// The one we want is who called CleanUp()
if _, file, line, ok := runtime.Caller(2); ok {
// now we actually need to read the source file
// This should be cached of course to avoid terrible performance
// I copied this from runtime/debug, so it's a legitimate thing to do :)
data, err := ioutil.ReadFile(file)
if err != nil {
panic("Could not read file")
}
// now let's read the exact line of the caller
lines := bytes.Split(data, []byte{'\n'})
lineText := strings.TrimSpace(string(lines[line-1]))
fmt.Printf("Line text: '%s'\n", lineText)
// Now let's apply some ugly rules of thumb. This is the fragile part
// It can be improved with regex or actual AST parsing, but dude...
return lineText == "}" || // on simple defer this is what we get
!strings.Contains(lineText, caller) || // this handles the case of defer func() { CleanUp() }()
strings.Contains(lineText, "defer ")
} // not ok - means we were not clled from at least 3 levels deep
return false
}
func CleanUp() {
if !isDeferred() {
panic("Not Deferred!")
}
}
// This should not panic
func fine() {
defer CleanUp()
fmt.Println("Fine!")
}
// this should not panic as well
func alsoFine() {
defer func() { CleanUp() }()
fmt.Println("Also Fine!")
}
// this should panic
func notFine() {
CleanUp()
fmt.Println("Not Fine!")
}
// Taken from the std lib's runtime/debug:
// function returns, if possible, the name of the function containing the PC.
func function(pc uintptr) string {
fn := runtime.FuncForPC(pc)
if fn == nil {
return ""
}
name := fn.Name()
if lastslash := strings.LastIndex(name, "/"); lastslash >= 0 {
name = name[lastslash+1:]
}
if period := strings.Index(name, "."); period >= 0 {
name = name[period+1:]
}
name = strings.Replace(name, "·", ".", -1)
return name
}
func main(){
fine()
alsoFine()
notFine()
}