No, panic/recover is a major hack.

Rob Pike has written ideas here on how to clean up error handling if-blocks: http://blog.golang.org/errors-are-values

No, it's not a "major hack." That blog post is excellent, but it just doesn't always apply.

The panic/recover idiom is extremely useful in parsers. For example, Go's JSON decoder: http://golang.org/src/encoding/json/decode.go

The difference is that in parsers, errors are almost certainly fatal. As opposed to a runtime SQL error, where it could be a user error, value error, or other database error, which can be recovered from.

Panic/recover are usually meant to catch "coding" and "assertion"-type errors (null pointers, array overflows, etc), not runtime errors. The latter is what the "error" values are used for.

I'm pretty sure you've misunderstood. Did you read the code I linked you to? Look at how it handles errors internally. Go's standard JSON decoder isn't going to crash your program if it fails to parse JSON---that would be disastrous. It exposes regular errors to the user. But internally, it uses panic/recover because it elides a lot of manual error checking.

Panic/recover is no problem as long as it does not traverse API boundaries, i.e. affects the user of your package, unless it really is a "fatal user error" in a function that wouldn't otherwise return an error.

Interesting. That's basically the Maybe monad.

I haven't written go in a while, but I recall being rather dissatisfied with the so-called 'if-hell'. I've been wondering lately - is there any reason why you couldn't/shouldn't use monads for this? Nothing obvious popping out of Google.

"gonads" would be my choice of name for this library.

Love the name! But there's no way Go's feeble type system supports anything remotely like a Monad.

Here's what GIlad Bracha might say about that:

http://www.infoq.com/presentations/functional-pros-cons?utm_...

Since they're wrapping things in transactions anyways, they could just defer the error check to the end (or for that matter rework their interface to the database to stop after the first error is detected).

This is a simple way of dealing with repetitive error handling.

I've written a tiny package that does just that in a controlled way, using named returns to bubble the error out.

https://github.com/dagoof/failure

[Docs] https://godoc.org/github.com/dagoof/failure

I have never messed with panic, I read that code and I think I'd make a function that takes functions - the error handling response is the same every time. Maybe even put the list of calls in an actual list and invoke them with an iterator.

Explicit error recovery is not hell, it's a blessing. Much easier to understand and to maintain than exception handling.

There is a third way. See Haskell's Either l r, or Rust's Result<T, E> types. See http://lucumr.pocoo.org/2014/10/16/on-error-handling/ and http://lucumr.pocoo.org/2014/11/6/error-handling-in-rust/

I don't know why so often, the immediate reaction to legitimate criticism of this wart in Go is to argue against exceptions, even when nobody has even brought up exceptions as an alternative.

In fact, Either/Maybe can function like safer versions of both paradigms: Functors allow you to build a program of many functions that will terminate with 'Nothing' or 'Either l' as long as any one of them experiences an error, without thinking about errors, and it still lets you explicitly handle any errors if you want to, this time on the sum type Maybe or Either, meaning that you usually handle both the error and non-error case, or receive a compiler warning.

I agree that, probably universally, the people who fervently argue for either exceptions or explicit Go-style error handling don't know about Maybe/Either/Result. Then again, all this is pretty much impossible to replicate in Go.

Thank you for mentioning this. It's frustrating that every time someone complains about Go's error-handling style, the first assumption is that the complainer then necessarily has to prefer exceptions. In truth, errors-as-values are very elegant in languages like Haskell IMO. And errors-as-values are not at all something that Go has managed to put back on the roadmap in a modern language against the grain of everyone else (the "exception people") as many seem to imply, including Rob Pike in his "errors as values" article.

Option types are nice. The fourth way (and I think the best) is supervision trees, though this is more applicable when you have multiple units of execution... which Go does, so not having something similar out of the box seems strange.

That said, even something as simple as errors being literal values that you can pattern match on is surprisingly versatile. Exceptions don't necessarily have to be awful weights that spill down all over the call stack, either. They can be distilled to a more basic catch/throw mechanism for shuffling around error properties and states.

Personally I love option types but working on a large Scala project at the moment it sometimes results in non ideal behaviour from developers especially those coming from the Java world.

What I've seen a lot of is swallowing of errors. For example readFromFile().orElse("") where there is no indication from an outsider that anything wrong happened. A cool feature would be logging any instance of when orElse was the result.

Would be curious how these work in other languages.

The trick is to make it very ergonomic to return the error. I thought Scala had do notation to make this the case. In Rust, the try! macro (which returns on error) is shorter than almost any other way to handle an error, so people usually use that.

I'm confused what makes exception handling so troublesome ?

Personally I find them essential in particular for larger codebases. They allow for centralised, consistent error handling behaviour. For example paging application support teams only for certain types of errors.

And you can always emulate explicit error recovery with exceptions. So what is so bad about having the flexibilty to accomodate different styles of error handling.

Yeah, I don't understand the objection to exceptions either. I _think_ the reason is because people treat them much the same was as errors in C/Go/etc., i.e., "I have to catch them here and recover right away." If you're doing it that way, you're doing it wrong. Design the system such that exceptions can be rethrown (bubbled up) and defer the handling as long as possible.

Nothing. This is one of those "Go doesn't have it, so it must be bad" things. The way Go handles errors is a throwback to C and was abandoned by literally every single imperative language for good reasons.

Honestly I wish that golang forced you to acknowledge all return values from methods even if you simply ignore them ( _ = Method() ). When I read the docs for any method calls in golang I like being able to see that something returns an error just by looking at the function signature. Certainly I might not know how to handle all the error variations, but not having to decide if I need to wrap something in try/catch... I dunno it just resonates with me.