While [the Apple "goto fail bug"] is pretty interesting, it's anecdotal evidence. I understand that it's still possible to produce bugs like this. One could also reasonably argue that it could have been avoided if they used else if instead of if or didn't use goto.

Aviation safety is built, pretty much completely, on that kind of "anecdotal evidence"; and, on the whole, it's pretty successful (it's part of why virtually every airliner accident makes headlines, but very few car accidents do, even though in aggregate, far more people die in car accidents than in airliner accidents). So are many other safety regulations. Basically, when something goes wrong, a bunch of very smart people sit down and try to figure out what went wrong, what allowed it to go wrong, and what change to the existing regulations or practices would have prevented it from going wrong.

I can't help but notice that, in your example, you suggest that

while(arr[index] != 0) {
    index++;
}

should be shortened to

while(arr[index] != 0)
    index++;

because the compound statement delimeters are "unnecessary". But you don't take the obvious next step and suggest to instead write

while(arr[index++] != 0);

which will save you yet another line. It also saves you from, completely unnecessarily, naming the iterator variable separately on that second line. Why not?

I'll hazard a guess: because that will trip people up. It's perfectly valid code. It'll accomplish the same thing. It'll almost certainly have the same side effects in all of C, C++ and Java. And because people aren't used to that kind of construct, at a minimum, it will require expending additional effort to parse. Just because it's valid doesn't make it good.

Yes, in an ideal world, people wouldn't make mistakes, and everyone would do exactly what they intended to do and nothing else. Unfortunately, we don't live in that world, nor do I think we are likely to inhabit any world like it any time soon. If we accept this, it stands to reason that we should look at what can be done to reduce the likelihood of mistakes.

And let's face it, it's a lot harder to accidentally add a statement to the wrong level if every level is laid out with curly braces, begin/end, or whichever other construct the particular language uses to combine statements into compound statements. It's far easier to accidentally (and without noticing it afterward) go from

if (x)
    y();
z();

to

if (x)
    y();
    y2();
z();

than from

if (x)
{
    y();
}
z();

to

if (x)
{
    y();
}
    y2();
z();

when you meant to make both the y() and y2() calls conditioned on x. No matter what tooling you're using, that last example immediately stands out as "it just looks plain wrong", whereas the variant without the curly braces doesn't. Likewise, if you were to write

if (x)
{
    y();
}
y2();
z();

(correct indentation, but the y2() call placed in the wrong spot) it is also immediately clear that the call to y2() is not conditioned on x, even though that was the original intent.

If we can make syntactically valid but semantically wrong code actually look wrong, it greatly reduces the risk of semantically wrong code slipping through the cracks. The compiler will already catch things that are syntactically wrong, but it can't catch those things where what the programmer wrote is valid, just not what was intended.

Indeed, this "costs" a few extra lines for the braces (or equivalent keywords in languages that don't derive basic syntactical elements from B, such as for example Python or SQL). But if your code is so deeply nested, with so many separate short snippets, that this is actually a problem, maybe you should consider refactoring it anyway. Individual preferences aside, it's not like we're still using VT52 terminals as our primary display devices.

posted over 3 years ago

CC BY-SA 4.0

3y ago

Canina‭

1479 reputation 2 23 149 31