-
-
Notifications
You must be signed in to change notification settings - Fork 29
Description
Problem
When adding context (attributes, user info, tags, etc.) to an existing OopsError, the current API requires wrapping the error again, which regenerates the stack trace and changes its origin point.
Example
// Create an error with stack trace pointing to this line
originalErr := oops.Code("ERR001").Errorf("database connection failed")
// Add context by wrapping - this regenerates the stack trace!
// Stack trace now points to this line instead of the original error
enrichedErr := oops.With("user_id", 123).Wrap(originalErr)This behavior causes:
- The stack trace no longer points to where the error actually occurred
- Multiple context additions create multiple unnecessary stack traces
- It's impossible to preserve the original error location while adding context
Suggestion
I've come up with about three ideas:
1. Add With*() methods on OopsError
Add methods that allow adding context without regenerating the stack trace:
// Add context
func (o OopsError) WithContext(kv ...any) OopsError
// Add user information
func (o OopsError) WithUser(userID string, userData ...any) OopsError
// Add tags
func (o OopsError) WithTags(tags ...string) OopsError
// Add other metadata
func (o OopsError) WithTenant(tenantID string, tenantData ...any) OopsError
func (o OopsError) WithCode(code string) OopsError
func (o OopsError) WithDomain(domain string) OopsError
func (o OopsError) WithHint(hint string) OopsError
func (o OopsError) WithPublic(public string) OopsError
func (o OopsError) WithOwner(owner string) OopsError
func (o OopsError) WithTrace(trace string) OopsError
func (o OopsError) WithDuration(duration time.Duration) OopsError2. Modify copy() to preserve stacktrace
Currently, OopsErrorBuilder.copy() explicitly does NOT copy the stacktrace field (I don’t understand why it’s implemented this way):
func (o OopsErrorBuilder) copy() OopsErrorBuilder {
return OopsErrorBuilder{
// ...
// stacktrace: o.stacktrace, // Not copied as it's generated per error
}
}If copy() preserved the stacktrace, we could do:
if oe, ok := oops.AsOops(err); ok {
builder := oops.OopsErrorBuilder(oe)
err = oops.OopsError(builder.With("user_id", 123))
// Stack trace would be preserved!
}This would require copying the stacktrace in the copy() method:
func (o OopsErrorBuilder) copy() OopsErrorBuilder {
return OopsErrorBuilder{
// ... existing fields ...
stacktrace: o.stacktrace, // Copy stacktrace
}
}This may be problematic as it alters the current behavior.
3. Add Apply() method to OopsErrorBuilder
Add a method that merges builder metadata into an existing OopsError while preserving its stack trace:
// Apply applies the builder's metadata to an existing OopsError
// without regenerating its stack trace
func (o OopsErrorBuilder) Apply(oe OopsError) OopsError {
o2 := oe
// Merge context
if len(o.context) > 0 {
o2.context = lo.Assign(map[string]any{}, oe.context, o.context)
}
// Merge tags
if len(o.tags) > 0 {
o2.tags = lo.Uniq(append(oe.tags, o.tags...))
}
// Override other fields if set
if o.code != "" {
o2.code = o.code
}
// ... other fields
// Preserve original stacktrace
return o2
}I'm happy to contribute if the solution looks good to you!