Re:Upcoming Features in Go 1.18
by silisili on Tuesday November 09, 2021 @12:34PM
(#29164625)
Apologies, to be more precise, my team wraps every error in a filename/line number pair, not an entire stacktrace.
[ Parent
| on HN ]
What is the (moral or performance) difference between wrapping every error you return anywhere with a filename/lineno pair, and a stack trace? If anything, the stacktrace proper probably only needs to allocate once, at the lowest level.
The specifics of it are deep enough in runtime that I've never actually looked, but it's designed as a chain to return a specific caller - it looks like -
_, file, line, _ := runtime.Caller(2)
locTxt := fmt.Sprintf(" (at %s:%d)", filepath.Base(file), line)
I think if the above on errors became a performance concern, I'd heavily rethink what I considered an error. But, it is also possible the systems we've designed are miles apart in function and philosophy.Yeah, this is fine for like "unable to contact database server" or "kafka producer timeout" where you're gonna terminate a large portion (or all) of the application at a much higher, removed level. "Lifecycle" stuff.
But if you make the short syntax do something like this, it's gonna get used for hundreds of library's equivalents of `Atoi` or `ParseQuery`. I don't need line annotations on my `http.ResponseWriter#Write` timeouts, they're slow enough already.
(And if you are going to do it - really, just attach the whole stack trace then and there, it will probably be faster.)