It’s great that they identified this (incredibly common) pain point and introduced a way to solve it, but I can’t help being disappointed.
Reading the examples I found myself thinking, “that looks like a really useful pattern, I should bookmark this so I can adopt it whenever I write code like that.”
The fact that I’m considering bookmarking a blog post about complex boilerplate that I would want to use 100% of the times when it’s applicable is a huge red flag and is exactly why people complain about Go.
It feels like you’re constantly fighting the language: having to add error handling boilerplate everywhere and having to pass contexts everywhere (more boilerplate). This is the intersection of those two annoyances so it feels especially annoying (particularly given the nuances/footguns the author describes).
They say the point is that Go forces you to handle errors but 99% of the time that means just returning the error after possibly wrapping it. After a decade of writing Go I still don’t have a good rule of thumb for when I should wrap an error with more info or return it as-is.
I hope someday they make another attempt at a Go 2.0.
I agree go’s error handling feels a bit clunky, though I prefer the local error handling and passing up the chain (if it were a bit more ergonomic) to exceptions, which IMO have a lot of other problems.
The main problems seem to me to be boilerplate and error types being so simplistic (interface just has a method returning a string). Boilerplate definitely seems solvable and a proper error interface too. I tend to use my own error type where I want more info (as in networking errors) but wish Go had an interface with at least error codes that everyone used and was used in the stdlib.
My rule of thumb on annotation is default to no, and add it at the top level. You’ll soon realise if you need more.
Yes I guess I do annotation in two places - initial error deep in libraries is annotated, this is passed back up to the initial handlers who log and respond and decide what to show users. Obviously that’s just a rule of thumb and doesn’t always apply.
Depends if it can be handled lower (with a retry or default data for example), if it can be it won’t be passed all the way up.
Generally though I haven’t personally found it useful to always annotate at every point in the call chain. So my default is not to annotate and if err return err.
What I like about errors instead of exceptions is they are boring and predictable and in the call signature so I wouldn’t want to lose that.
Not in my experience. Usually your call chain has forks. Usually the DoThing function will internally do 3 things and any one of those three things failed and you need a different error message to disambiguate. And four methods call DoThing. The 12 error paths need 12 uniquely rendered error messages. Some people say "that is just stack traces," and they are close. It is a concise stack trace with the exact context that focuses on your code under control.
If you have both the start of the call chain and the end of the call chain mapped you will get a different error response almost every time and it is usually more than enough, so say your chain is:
Do1:...Do10, which then DoX,DoY,DoZ and one of those last 3 failed.
Do you really need Do1 to Do10 to be annotated to know that DoY failed when called from Do1? I find:
Do1:DoZ failed for reason bar
Just as useful and a lot shorter than:
Do1: failed:Do2:failed...Do9 failed:Do10:failed:DoZ failed for reason bar
It is effectively a stack trace stored in strings, why not just embed a proper stack trace to all your errors if that is what you want?
Your concern with having a stack trace of calls seems a hypothetical concern to me but perhaps we just work on different kinds of software. I think though you should allow that for some people annotating each error just isn't that useful, even if it is useful for you.
Your example shows an ideal case w/o repetition. If every layer just wraps error without inspecting, then there will be duplication in the error string.
I have never seen that. I have shipped multiple dozens of services at half a dozen companies. Big code bases. Large teams. Large volumes of calls and data. Complicated distributed systems.
I am unable to imagine a case where an error string repeated itself. On a loop, an error could repeat, but those show as a numerical count value or as separate logs.
I close my nose and always wrap errors with a sentinel error for public functions/methods so that callers can check with `errors.Is`. And you can always identify the place in the call-stack where the error occurred.
I need to start getting used to context with cancel cause - muscle memory hasn't changed yet.
It was a great piece and I learned a lot, thanks for writing it. I hope you didn’t think that it was you I was disappointed with rather than the language designers :)
It’s ironic how context cancellation has the opposite problem as error handling.
With errors they force you to handle every error explicitly which results in people adding unnecessary contextual information: it can be tempting to keep adding layer upon layer of wrapping resulting in an unwieldy error string that’s practically a hand-rolled stacktrace.
With context cancellation OTOH you have to go out of your way to add contextual info at all, and even then it’s not as simple as just using the new machinery because as your piece demonstrates it doesn’t all work well together so you have to go even further out of your way and roll your own timeout-based cancellation. Absurd.
No worries. Your intent was clear. I don't mind the boilerplates if they were footgun free. Context requires you write a bunch of boilerplate where it's still really easy to make mistakes.
Go is seen as too boiler plate-ish, and no one likes that. But one of the biggest Go's biggest assets is its simplicity. And it might not be possible to have both simplicity and low boiler plate.
I quite enjoy C# and F# and while they are low boiler plate, you can really learn them in a week or two the way you can learn Go.
And even you don't know anything about Go, you can literally jump into the code base and understand and follow the flow with ease - which quite amazes me.
So unfortunately, every language has trade offs and Go is not an exception.
I can't say I enjoy Go as a language but I find it very, very useful.
And since many people are using LLMs for coding these days, the boiler plate is not as much an issue since it be automated away. And I rather read code generated in Go than some C++ cryptic code.
There are two things I think you could have as implict in Go - error values, and contexts.
Just pass along two hidden variables for both in parameters and returns, and would anything really change that the compiler wouldn't be able to follow?
i.e. most functions return errors, so there should always be an implicit error return possible even if I don't use it. Let the compiler figure out if it needs to generate code for it.
And same story for contexts: why shouldn't a Go program be a giant context tree? If a branch genuinely doesn't ever use it, the compiler should be able to just knock the code out.
What's the difference between an implicit error and exceptions? Being explicit about errors is good. Go's syntactical implementation, coupled with its unexpressive type system, is the problem.
I will freely go on the record as saying that there's nothing wrong with exceptions for this exact reason: errors are so common that a function being "pure" is the exception, and that errors-as-value handling invariable turns into an endless chain of something like "if err; return (nil/zero, err)" in every language which tries it.
The same would apply to anytime you have Result types - ultimately its still just syntactic sugar over "if err then...".
What's far more common in real programs is that an error can occur somewhere where you do not have enough context to handle or resolve it, or you're unaware it can happen. In which case the concept of exceptions is much more valid: "if <bad thing here> what do I want to do?" usually only has a couple of places you care about the answer (i.e. "bad thing happened during business process, so start unwinding that process" and many more where the answer is either "crash" or "log it and move on to the next item".
Exceptions can be bad if done the wrong way. But the solution isn’t to not deal with it and put it on the programmer. That’s laziness.
The problems are that the signature of functions doesn’t say anything about what values it might throw, and that sometimes the control flow is obscured — an innocuous call throws.
Sure but that also feels like a compiler problem. The compiler knows everywhere my function can go. So rather then having it just throw an exception - i.e. arbitrary data - on the stack, surely what's really happening is I'm creating a big union of "result | error[type,type,type,type]" which only gets culled when I add my "exception" handling.
My argument here would be, that all of this though doesn't need to be seen unless its relevant - it seems reasonable that the programmer should be able to write code for the happy path, implicitly understanding there's an error path they should be aware of because errors always happen (I mean, you can straight up run out of memory almost anywhere, for example).
I think this post needs better examples to show case the issue, because right now the issue is not clear. Ideally you would need an example that uses the context.Cause function, see below
The contexts and errors communicate information in different directions. Errors let upstream function know what happened within the call, context lets downstream functions know what happened elsewhere in the system. As a consequence there isn't much point to cancel the context and return the error right away if there isn't anybody else listening to it.
Also, context can be chained by definition. If you need to be able to cancel the context with a cause or cancel it with a timeout, you can just make two context and use them.
Haskell is the king of cancellation. Using asynchronous exceptions, you can cancel anything, anytime, with user
-defined exception types so you know what the cancellation reason is.
Example:
maybeVal <— timeout 1000000 myFunction
Some people think that async exceptions are a pain because you nerd to be prepared that your code can be interrupted any time, but I think it's absolutely worth it because in all the other languages I encounter progress bars that keep running when I click the cancel button, or CLI programs that don't react to CTRL+C.
In Haskell, cancellability is the default and carries no syntax overhead.
This is one of the reasons why I think Haskell is currently the best language for writing IO programs.
ZIO in Scala tracks this sort of thing except you don't have to remember to pass around or select on the ctx (it's just part of the fibre/"goroutine"); if it's cancelled, the fibre and its children just stops the next time it yields (so e.g. if it "selects" on anything or does any kind of IO).
Python async tasks can be cancelled. But, I don't think you can attach must context to the cancel (I think you can pass a text message), so it would seem the argument of what go suffered from would apply.
(I also think there's some wonkiness with and barriers to understanding Python's implementation that I don't think plagues Go to quite the same extent.)
All mainstream languages have it in one or more forms (either direct task I/O cancellation, or cancellation tokens or I/O polling that can include synthetic events) since otherwise several I/O patterns are impossible
Golang returning tuples but not having pattern matching is something I'll never get. I really feel it's a too dumbed down version of erlang/elixir, especially with this context passing business
Reading the examples I found myself thinking, “that looks like a really useful pattern, I should bookmark this so I can adopt it whenever I write code like that.”
The fact that I’m considering bookmarking a blog post about complex boilerplate that I would want to use 100% of the times when it’s applicable is a huge red flag and is exactly why people complain about Go.
It feels like you’re constantly fighting the language: having to add error handling boilerplate everywhere and having to pass contexts everywhere (more boilerplate). This is the intersection of those two annoyances so it feels especially annoying (particularly given the nuances/footguns the author describes).
They say the point is that Go forces you to handle errors but 99% of the time that means just returning the error after possibly wrapping it. After a decade of writing Go I still don’t have a good rule of thumb for when I should wrap an error with more info or return it as-is.
I hope someday they make another attempt at a Go 2.0.
reply