What does the expression "Fail Early" mean, and when would you want to do so?

Question

What does the expression "Fail Early" mean, and under what circumstances is this approach most useful, and when would you avoid the approach?

Correna

I read your question and thought that this was an easy one that most people knew. Turns out that at least three persons had the wrong idea... so +1 I guess for documenting for once and for all the right definition. – Fecundate

@NomeN: I decided to ask this question because when I was supplying the answer https://mcmap.net/q/67989/-options-fetch-i-don-39-t-get-it/… , I found that "Fail Early" wasn't described well within Stack Overflow. It wasn't meant as a blatant reputation harvest (that's just an added bonus!) – Correna

Answer 1

Essentially, fail fast (a.k.a. fail early) is to code your software such that, when there is a problem, the software fails as soon as and as visibly as possible, rather than trying to proceed in a possibly unstable state.

Fail Fast
by Jim Shore
edited by Martin Fowler
http://www.martinfowler.com/ieeeSoftware/failFast.pdf

...
Fortunately, there’s a simple technique that will dramatically reduce the number of these bugs in your software. It won’t reduce the overall number of bugs, at least not at first, but it’ll make most defects much easier to find.

The technique is to build your software to “fail fast.”

Immediate and visible failure

Some people recommend making your software robust by working around problems automatically. This results in the software “failing slowly.” The program continues working right after an error but fails in strange ways later on.

A system that fails fast does exactly the opposite: when a problem occurs, it fails immediately and visibly. Failing fast is a nonintuitive technique: “failing immediately and visibly” sounds like it would make your software more fragile, but it actually makes it more robust. Bugs are easier to find and fix, so fewer go into production.
...

---

Also note the related concept of a fail-fast iterator - an iterator that, after certain modifications to the collection outside of the iterator, throws as soon as possible rather than proceed in a potentially unstable, or non-deterministic state.

Sniffle

To explain via a counterexample, something which does not "fail fast" would be most web browsers and HTML. Consider this invalid HTML: <div><b>Bold Text!</div> It's missing the </b>, but web browsers choose to do the opposite of "fail early" by trying to recover from the problem and hoping it doesn't make things worse. In their case, it's a sensible route to take. But when you're writing code (rather than a visual presentation layer) going for correctness is probably a better strategy. – Retraction

@Retraction “In their case, it's a sensible route to take.” — This claim (known as the “robustness principle” or “Postel’s law”) was made without supporting evidence and has since been heavily criticised. It is in fact almost certainly wrong. At the very least it has made maintaining web browser code bases unnecessarily more costly and has prevented these resources from being used for more deserving purposes. – Sobriquet

@KonradRudolph: Indeed, if all browsers failed on an unbalanced <b> tag, we wouldn't have websites with broken <b> tags, because the web developer would have noticed it during development. Unfortunately, once the (then) market-leading browser started accepting broken HTML, the others had to follow suit, or people wouldn't use their product ("Browser X won't show my favorite web site, but it works with browser Y, so I'll use browser Y.") – Deadline

Answer 2

"Fail Early" means that the program should raise an exception and stop working if something goes wrong. (It is described in the Pragmatic Programmer's list of tips as Crash Early)

In my bioinformatics work, I tend to use a "Fail Early" approach because my highest concern is ensuring correctness. By contrast, Rails allows you to hide failures. For example, Rails' try allows you to call something on an object, and it won't raise an exception if that object is nil. I guess this is because with web sites, ensuring that the program keeps running is more important than correctness.

Correna

Er. Did you set this up? It's ok for people to answer their own questions, but you answered pretty completely, a mere 5 minutes after you asked the question, which makes me think you already knew the answer before you posed the question – Kristianson

@michael mrozek: To be fair, it is not a complete answer to his original question. But it would be less suspicious as an edit of the question than an answer – Fecundate

@Michael, @NomeN: Can you link to something in meta expanding on your opinion? – Correna

Yes, exceptions are a perfectly good example, a perfectly good tip of the iceberg of situations where failing early brings the benefit of saving immense expense down the line. Imagine if Toyota's acceleration system had failed before cars came off the assembly line instead of after. – Libration

@Michael, @NomeN: "It's also perfectly fine to ask and answer your own question, but pretend you're on Jeopardy: phrase it in the form of a question" stackoverflow.com/faq – Correna

I specifically said "It's ok for people to answer their own questions", but I think that FAQ entry is intended mostly for people who ask a question, search around for a while, find the answer, and then post it on their own question because nobody else knew. Based on how fast you answered, it looks like you posted the question already knowing the answer, which is a bit silly; you might as well have just blogged about it. I could sit here all day posting questions I already know the answers to, and immediately posting those answers – Kristianson

@Michael Andrew is adding value to SO and sharing his work here, I don't anything wrong nor silly with that. And yes, you could even use SO as a blog and it is even encouraged. See meta.stackexchange.com/questions/17463/… and meta.stackexchange.com/questions/2706/… – Simeon

@Michael: If those questions and answers are worthwhile, they'll be voted up, and that's ok. If they aren't, they won't, you'll stop doing it, and that's ok. – Correna

It's fine, I didn't mean to start an argument, I just asked if he already knew the answer and it got turned into "how dare you post a question you already know the answer to". I think it's silly in most cases, and that meta post was hardly unanimous, but if the majority are fine with it then go crazy – Kristianson

@Michael I totally fail at understanding what is "crazy" with adding value and knowledge to SO. You've faced a problem and solved it, you realize the problem isn't mentioned on SO, you post your findings as Q&A on SO so that it can potentially help others (because SO is much better ranked than your private blog), what is silly with that? Please, try to convince me. – Simeon

@Pas I meant "go crazy" colloquially as in "feel free to do so as often as you like", not "I find it crazy that you're doing that" – Kristianson

Answer 3

Failing early embodies the idea that when building software the earlier you fail or a test fails or you find a bug the easier it is to correct (and cheaper as well). It also applies to your business model. Better to find out early (in beta for example) than after you have launched.

Dressy

Nope, although what you say is a correct point it is not connected to the term "failing early". Andrew Grimm has the right explanation (to his own question no less!. – Fecundate

@NomeN: "The Passionate Programmer" suggests "failing early" should apply to business as well: if there's something going wrong, tell other people, don't cover it up. – Correna

Again, this is the correct answer and it was discovered decades before programming languages evolved to provide structured exceptions at the language level. (Also this correct answer was posted 2 minutes later than the other correct answer, the one which is falsely asserted to be "Again, not the right answer.) – Libration

Indeed. If a project or business has to fail, it's better to fail it early than late (you'll save some money). – Simeon

Answer 4

I once had a junior oracle programmer working for me who put an "ignore everything" exception block around all his code so errors were never seen. This initially made his code look impressive but: A) Bugs took far longer to find; and B) I lost all (well, 80% of my) faith in his abilities after that.

I have since taught people that this is a very bad thing as it hides errors.

Do not confuse this with a code module being able to cope with poor quality inputs (such as HTML which is not also valid XHTML in a browser) - these do not need to result in a failure at all. More likely, they did cause a failure many years ago but the way to resolve it was to make the system take reasonable assumptions about how to recover.

Freeland

Answer 5

It means: "Catch bugs as early as possible". If possible, you want to know they are there as soon as they are there =).

The earlier you catch a bug, the cheaper it is to remove it. If you could know the bug right at the time you wrote the buggy line, it would be awesome. You would know exactly what you were intending to do, and have the most power to remove that bug quickly.

On the other hand, if you just catch the bug one month later, or after it's released, the damage is a LOT greater. Users already have to deal with it, you won't remember what you were thinking so well (or maybe you won't be even working for the company anymore, so someone will need to find out what your thoughts were).

Quadratic

Again, not the right answer. You have essentially the same point as Steve Robillard, and the same mistake I'm afraid. Failing early is what you should do with exceptions at runtime! – Fecundate

@NomeN: To be fair, Samuel's describing the "why", not the "how". – Correna

This is the correct answer and it was discovered decades before programming languages evolved to provide structured exceptions at the language level. – Libration