Go checking for the type of a custom Error

Question

Reading the Blog post further exposes a bit of Go like this:

serr, ok := err.(*model.ModelMissingError)

This is the comma ok idiom, clearly I need to re do my go lang tour

I have manged to make an error assertion using the switch statement as follows:

 err := FuncModelMissingError()

 switch t := err.(type) {
 default:
     fmt.Println("not a model missing error")
 case *ModelMissingError:
     fmt.Println("ModelMissingError", t)
}

I hope this helps you out.

Now with Go 1.13 we can use the following from the errors package:

if errors.Is(err, model.ModelMissingError) {...}

See the blog post: https://blog.golang.org/go1.13-errors

To check the TYPE of the error, use errors.As

As finds the first error in err's chain that matches target [...] ​An error matches target if the error's concrete value is assignable to the value pointed to by target

Of course type identity is a condition for assignability.

So it would look like:

target := &model.ModelMissingError{} 
if errors.As(err, &target) {
    fmt.Println(target) // no model found for id
}

Pay attention to the two uses of & in the example above. This is because:

As panics if target is not a non-nil pointer to either a type that implements error, or to any interface type.

In your case, you declared Error() string method on the pointer receiver, therefore "a pointer to the type that implements the error interface" to satisfy As is **ModelMissingError. So you need to address twice.

---

The other method errors.Is checks for value equality.

An error is considered to match a target if it is equal to that target or if it implements a method Is(error) bool such that Is(target) returns true.

This is useful for example in case of fixed error values, e.g. errors declared as var or const like the standard library io.EOF. As an example:

var ErrModelMissing = errors.New("no model found for id")

func foo() {
    err := bar()
    if errors.Is(err, ErrModelMissing) {
        fmt.Println(err) // no model found for id
    }
}

---

Consider that the usefulness of Go 1.13 errors.As and errors.Is lies in error unwrapping. If you inspect the error at the top of an arbitrarily long call stack, you must remember that the original error may become wrapped into other errors while being bubbled up. Then directly checking for equality or type assignability is not enough.

    err1 := fmt.Errorf("wrapped: %w", &ModelMissingError{})
    target := &ModelMissingError{}
    fmt.Println(errors.As(err1, &target)) // true
    
    err2 := fmt.Errorf("wrapped: %w", FixedError)
    fmt.Println(errors.Is(err2, FixedError)) // true
    fmt.Println(err2 == FixedError) // false

Additionally, the package github.com/pkg/errors is compatible with errors.As and errors.Is:

// import pkgerr "github.com/pkg/errors"

err3 := pkgerr.Wrap(pkgerr.Wrap(&ModelMissingError{}, "wrapped 1"), "wrapped 2")
fmt.Println(errors.As(err3, &target)) // true

Playground: https://play.golang.org/p/FEzggdBLCqq

---

Naturally, if you know for sure that the error is not wrapped, a good old type assertion works just as fine:

if myerr, ok := err.(*model.ModelMissingError); ok {
    // handle myerr
}