Error-Handling and Panicking in a User-Defined Package
This lesson provides an implementation and a detailed explanation about catching errors in custom packages and recovering programs in case of a panic.
We'll cover the following...
Here are a couple of best practices which every writer of custom packages should apply:
- Always recover from panic in your package: no explicit
panic()
should be allowed to cross a package boundary. - Return errors as error values to the callers of your package.
This is nicely illustrated in the following code:
package parse import ( "fmt" "strings" "strconv" ) // A ParseError indicates an error in converting a word into an integer. type ParseError struct { Index int // The index into the space-separated list of words. Word string // The word that generated the parse error. Err error // The raw error that precipitated this error, if any. } // String returns a human-readable error message. func (e *ParseError) String() string { return fmt.Sprintf("pkg parse: error parsing %q as int", e.Word) } // Parse parses the space-separated words in in put as integers. func Parse(input string) (numbers []int, err error) { defer func() { if r := recover(); r != nil { var ok bool err, ok = r.(error) if !ok { err = fmt.Errorf("pkg: %v", r) } } }() fields := strings.Fields(input) numbers = fields2numbers(fields) return } func fields2numbers(fields []string) (numbers []int) { if len(fields) == 0 { panic("no words to parse") } for idx, field := range fields { num, err := strconv.Atoi(field) if err != nil { panic(&ParseError{idx, field, err}) } numbers = append(numbers, num) } return }
In parse.go, we implement a simple version of a parse
package. From line 9 to line 13, we define a ParseError
type (see the comments in the code for more info). Then, we have a String()
...