Advanced error handling for Go
Go to file
Tordarus 8ee50a788a migration notice 2022-08-03 21:44:17 +02:00
README.md migration notice 2022-08-03 21:44:17 +02:00
calltrace.go removed useless println 2020-09-10 15:05:28 +02:00
doc.go initial commit 2020-09-09 11:48:46 +02:00
error.go chained errors introduced 2021-09-09 16:35:48 +02:00
error_test.go chained errors introduced 2021-09-09 16:35:48 +02:00
error_tmpl.go chained errors introduced 2021-09-09 16:35:48 +02:00
error_tmpl_test.go chained errors introduced 2021-09-09 16:35:48 +02:00
globals.go fixed default value for DisableTrace 2020-09-10 14:39:18 +02:00
go.mod moved to gitlab 2021-12-13 12:04:31 +01:00
utils.go initial commit 2020-09-09 11:48:46 +02:00

README.md

adverr

Package adverr implements errors with call stack traces as well as error templates for error equality

Migration notice

This project migrated to git.milar.in!

Future changes and versions will be made in that new repository!

Installation

go get git.tordarus.net/Tordarus/adverr

Usage examples

Import

import "git.tordarus.net/Tordarus/adverr"

Create error templates

var (
    ErrDoStuffFailed = adverr.NewErrTmpl("ErrDoStuffFailed", "Could'nt do stuff because of %s")
)

Create independent error (without error template)

func doStuffWithIndependentErr() error {
    return adverr.New("Could'nt do stuff")
}

Create error based on template

func doStuff() error {
    return ErrDoStuffFailed.New("reasons")
}

Print errors on stderr convieniently

adverr.Print(myErr)
adverr.Println(myErr)

Print errors on stderr and exit with exit code

adverr.Fatal(myErr, 1)
adverr.Fatalln(myErr, 1)

Advantages of error templates

two errors made by the same template will return true when called with errors.Is()

func doStuffAndFailWithMsg(msg string) error {
    return ErrDoStuffFailed.New(msg)
}

err1 := doStuffAndFailWithMsg("err1")
err2 := doStuffAndFailWithMsg("err2")

fmt.Println(errors.Is(err1, err2))          // true
fmt.Println(err1 == err2)                   // false
fmt.Println(err1.Error() == err2.Error())   // false

Wrap errors (Causality of errors)

By wrapping errors, you can provide an error that is caused by another error. A 'Caused by' section will be printed in the stack trace showing the original error. You can also retrieve the original error by using errors.Unwrap()

func doStuffWrapped() error {
    err := doStuff()
    if err != nil {
        return adverr.Wrap("doStuffWrapped failed", err)
    }

    return nil
}

Chain errors (Errors caused in succession)

By chaining errors, you can provide an error that represents multiple errors caused in succession inside the same function. For each chained error a 'Previously thrown' section will be printed in the stack trace. You can programmatically check chained errors using the following methods:

// Get Returns the first error in the chain for which errors.Is(target) returns true
Get(target error) error

// GetByIndex returns the i'th error in the chain
GetByIndex(i int) error

// Chain returns a slice of all chained errors
Chain() []error

// Contains is a shorthand for Get(target) != nil.
// Can be considered as an errors.Is function but for chains instead of causes
Contains(target error) bool

Be aware that the standard library calls wrapped errors chains as well! But these chains are something different. Here is an example use case:

You have a list of files from which you only want to read the first one you have read permissions for. This is most likely done in a loop inside the same function. A chained error can keep all previously failed read errors and show them in a debuggable way. Wrapping by causality would be ambiguous because they might already have been wrapped multiple times and their causes can therefore not be distinguished from previously failed errors (chained errors).

Retrieve call stack trace (for debugging purposes)

fmt.Println(adverr.Trace())

Example of a printed error

Code:

package main

import (
	"adverr"
	"errors"
)

var (
	ErrDoStuffFailed = adverr.NewErrTmpl("ErrDoStuffFailed", "Could'nt do stuff because of %s")
)

func main() {
	err := doStuffInAnotherGoroutine()
	if err != nil {
		adverr.Fatalln(err, 1)
	}
}

func doStuff() error {
	err := doGoNativeStuff()
	if err != nil {
		return ErrDoStuffFailed.Wrap(err, "reasons")
	}

	return nil
}

func doStuffInAnotherGoroutine() error {
	ch := make(chan error, 1)

	go func() {
		ch <- doStuff()
		close(ch)
	}()

	err := <-ch
	if err != nil {
		return adverr.Wrap("Goroutine failed because of errors", err)
	}

	return nil
}

func doGoNativeStuff() error {
	return errors.New("some go error")
}

Output:

adverr.Error: Goroutine failed because of errors
        at main.doStuffInAnotherGoroutine (/home/user/go/src/test/main.go:38)
        at main.main (/home/user/go/src/test/main.go:13)
        at runtime.main (/usr/local/go/src/runtime/proc.go:204)
Caused by ErrDoStuffFailed: Could'nt do stuff because of reasons
        at main.doStuff (/home/user/go/src/test/main.go:22)
        at main.doStuffInAnotherGoroutine.func1 (/home/user/go/src/test/main.go:32)
Caused by errors.errorString: some go error
        (Unknown source)

Globals

You can set the maximum limit of the call stack trace via

adverr.CallStackLength = 50 // default value: 100

If you are in a productive environment, consider disabling call traces completely for performance reasons:

adverr.DisableTrace = true // default value: false

Change log

v0.1.2

Introduced error chaining

v0.1.1

Improved errors.Is behavior so that ErrTmpl's are considered as targets as well. Example:

err := ErrDoStuffFailed.New("some error")
fmt.Println(errors.Is(err, ErrDoStuffFailed)) // returns true since v0.1.1

v0.1.0

initial release