SDK methods now return 1 error. To differentiate between Client errors and Server errors I've also implemented the AsServerError function which returns an *api.Error if it is an error originating from the remote server or nil otherwise.
59 lines
1.8 KiB
Go
59 lines
1.8 KiB
Go
package api
|
|
|
|
import (
|
|
"errors"
|
|
"fmt"
|
|
"net/http"
|
|
"strings"
|
|
|
|
"google.golang.org/grpc/codes"
|
|
)
|
|
|
|
var (
|
|
ErrNotFound = &Error{Status: http.StatusNotFound, Code: codes.NotFound.String()}
|
|
ErrInvalidArgument = &Error{Status: http.StatusBadRequest, Code: codes.InvalidArgument.String()}
|
|
ErrPermissionDenied = &Error{Status: http.StatusForbidden, Code: codes.PermissionDenied.String()}
|
|
ErrUnauthorized = &Error{Status: http.StatusUnauthorized, Code: codes.Unauthenticated.String()}
|
|
)
|
|
|
|
// AsServerError returns an api *Error from the provided error. If the provided error
|
|
// is not an api Error nil is returned instead.
|
|
func AsServerError(in error) *Error {
|
|
var serverErr *Error
|
|
if !errors.As(in, &serverErr) {
|
|
return nil
|
|
}
|
|
return serverErr
|
|
}
|
|
|
|
// Error satisfies the error interface.
|
|
func (e *Error) Error() string {
|
|
res := fmt.Sprintf("Status: %d, Code: %q, Error: %q", e.Status, e.Code, e.Message)
|
|
var dets []string
|
|
if e.Details != nil {
|
|
if e.Details.ErrorId != "" {
|
|
dets = append(dets, fmt.Sprintf("error_id: %q", e.Details.ErrorId))
|
|
}
|
|
if e.Details.RequestId != "" {
|
|
dets = append(dets, fmt.Sprintf("request_id: %q", e.Details.RequestId))
|
|
}
|
|
if e.Details.TraceId != "" {
|
|
dets = append(dets, fmt.Sprintf("TraceId: %q", e.Details.TraceId))
|
|
}
|
|
for _, rf := range e.Details.RequestFields {
|
|
dets = append(dets, fmt.Sprintf("request_field: {name: %q, desc: %q}", rf.Name, rf.Description))
|
|
}
|
|
}
|
|
if len(dets) > 0 {
|
|
det := strings.Join(dets, ", ")
|
|
res = fmt.Sprintf("%s, Details: {%s}", res, det)
|
|
}
|
|
return res
|
|
}
|
|
|
|
// Errors are considered the same iff they are both api.Errors and their statuses are the same.
|
|
func (e *Error) Is(target error) bool {
|
|
tApiErr := AsServerError(target)
|
|
return tApiErr != nil && tApiErr.Code == e.Code && tApiErr.Status == e.Status
|
|
}
|