Re: Re: Re: What should be returned in scalar context? (best practice)

[sauoq]

Document that it returns a list. It's not possible to return an array.

No. I would (will and do) document that it returns an array. What you are saying might be true from an internals standpoint, but that's irrelevant. Read the code: return @ret; . . . it says "return" and is followed by an array. More importantly, it behaves like an array. It doesn't behave like a list literal (a slice does, however, as ysth points out.) It certainly doesn't behave like an array reference. Documenting that it returns an array is the only clear way to describe its behavior.

If you document that some functions returns an array, the user can only guess: is it a list or an array reference?

When a funtion returns an array reference, document that it returns an array reference. "Array" and "array reference" are not synonyms. The user doesn't have to guess! He just has to read.

If you sometimes describe a reference to an array as "an array", then you are being sloppy and you should stop.

[Juerd]

Read the code: return @ret; . . . it says "return" and is followed by an array.

print @foo; does not print an array, it prints a list.

More importantly, it behaves like an array.

Only in two ways. It returns a number of elements in scalar context and a list of elements in list context. That's where the comparison ends.

You can't splice what your function returns, push onto it, shift from it, unshift onto it, pop from it, reference it as an array, tie it, assign to it or undef it.

It is not an array. Don't call it an array, because it simply is not an array. You say it behaves like an array, but it does not. The very things that make an array an array do not apply. Just some context sensitivity, but that does not make your list or number an array.

If you don't want to call it a list because in scalar context it returns a number of elements, specify both behaviours in your documentation. That's what perl's documentation also does. See grep, keys, split, etcetera.

"Returns a list" is wrong when the sub is called in scalar context. "Returns an array" is wrong in every possible context. Note that "array context" does exist and your sub cannot be used in it.

Only an lvalue sub can return an array, but that cannot be done using the return keyword.

It doesn't behave like a list literal

Indeed. It does not behave like a list *literal*. But in list context, it does behave like a list, in every single way. It's important to know that there are two kinds of lists, and that list literals are special with syntax; other than that, they're just lists.

Documenting that it returns an array is the only clear way to describe its behavior.

No, no, no, no, NO! It's the most vague way to describe its behaviour. Don't call it clear, because it is the very opposite of clear! If you want to be absolutely clear, you must specify the difference in behaviour for every context, or modify the sub so that it really does behave like an array, or like a list.

"Array" and "array reference" are not synonyms.

Correct. But why do you think that "list" and "array" are synonyms? They're not. My point is not that "returns an array" would imply it being a reference. My point is that it's entirely unclear whether it returns a list or a reference to an array. It's not returning an array, that's for sure (since you use the return keyword, we're not talking about an lvalue sub here).

If you sometimes describe a reference to an array as "an array", then you are being sloppy and you should stop.

I could not agree more.

I'd like to add this one:
If you sometimes describe a list as "an array", then you are being sloppy and you should stop.

[sauoq]

Juerd, I am perfectly aware that functions always return a list. Does that mean I'm going to stop documenting that a function like sub f { return 0 } "returns a scalar?"

No.

Somehow I doubt you'd take issue with that even though it isn't technically true. That function is, afterall, returning a list. The list simply has one element. Your argument, if applied evenly, would require the abolishment of the phrase "returns a scalar" as well. Fortunately, we won't have to do that, because your position is completely untenable.

By saying your function "returns a scalar" or "returns an array" or "returns a list" you are able to describe exactly how code the is written without trying to describe how Perl works.

The reason it makes sense to say those things is that context is imposed before the value is returned. When you return @ret; the list returned is composed of elements generated by evaluating @ret in the required context. If it is in scalar context, you get the number of elements of @ret. If it is in list context, you get all of the elements of @ret. Likewise, if you return (33, 42); the list returned is created by evaluating (33, 42) in scalar or list context as necessary.

In other words, there is always a disconnect between what you return and what Perl returns. If you return (33, 42); and your function is called in scalar context, the function returns (42).

I document what I do. Perl is already well-documented.

[Juerd]

Nevermind. You're not getting it. If you document only what your code does, you might as well just duplicate the source.

You say that my argument is "overly academic with no practical basis". Perhaps it is. It's not something I have considered. I'm just someone who hates documentation that is wrong and vague where it could be correct and clear if the author cared.

Technically and linguistically, you're wrong if you say that a Perl non-lvalue subroutine returns an array. Write one sentence more and your documentation is correct and clear.

Instead of Returns an array of foos. just write Returns a list of foos. (In scalar context, returns the number of foos.) and stop making people guess what you mean.

If you insist on redefining the word 'array', go ahead. But don't forget to document your meaning of the word.

I am perfectly aware that functions always return a list. Does that mean I'm going to stop documenting that a function like sub f { return 0 } "returns a scalar?" Somehow I doubt you'd take issue with that even though it isn't technically true. That function is, afterall, returning a list. The list simply has one element.

No, it is perfectly acceptable to call a one-element list a scalar in this context. However, it is not very customary to mention that the single value you return is a scalar. Usually, we talk about "a number" or "a string". "Returns true", "returns the integer portion", "returns a string", "returns the number of". In fact, no function in perlfunc is documented to return a scalar.

However, it is customary to indicate that a collection of numbers or strings is returned by using the word "list", because such a collection is called a "list" in Perl jargon.

By the way, why do you think my argument has no practical basis? That's nonsense. Consistency with existing jargon and documentation is very practical.

Your argument, if applied evenly, would require the abolishment of the phrase "returns a scalar" as well.

Saying that a function returns a scalar isn't even wrong, even if the function returns 15 scalars, it still also returns a scalar. It would be misleading, though, so I'd avoid it. But as I said, it is not common to say that a function returns any scalar at all. We usually describe what the scalars are, since we assume every Perl coder knows that normal (non-lvalue) subs can only return scalars.

Fortunately, we won't have to do that, because your position is completely untenable.

I'm wasting my time. You're even more stubborn than I am. Even when presented with facts and prior art, you refuse to see that what you do is different from what others have agreed on.

By saying your function "returns a scalar" or "returns an array" or "returns a list" you are able to describe exactly how code the is written without trying to describe how Perl works.

A sub returns a foo or a list of bars. You don't document that a function returns a scalar or just a list. But if you do, that's acceptable. Documenting that a function returns an array is *not acceptable* if it does not return an array. Is that so hard? Don't say your function returns 2 if does not return 2. And don't say your function returns 1+2 if it has return 1+2 in it. It returns 3. Stick to the facts and keep it simple (but never make it inaccurate). This is technical documentation, not prose.

The reason it makes sense to say those things is that context is imposed before the value is returned.

Actually, it depends on how you look at it. Context is propagated to every single return statement, and to the last statement of the sub.

When you return @ret; the list returned is composed of elements generated by evaluating @ret in the required context. If it is in scalar context, you get the number of elements of @ret. If it is in list context, you get all of the elements of @ret. Likewise, if you return (33, 42); the list returned is created by evaluating (33, 42) in scalar or list context as necessary.

That is correct. But even if you have return @foo, @foo, the array, is never returned. It cannot be referenced or mutated. The thing that is returned is not an array. The function does not return an array, ever.

It's very simple. You can call something an array if it starts with @. Because lvalue subs are reasonably new, some work-around syntax is needed to actually use an lvalue sub as an array: @{\foo} instead of just foo, and that too proves my point that only if it starts with @, it is an array.

Note that I'm not saying that everything that does start with @ is an array. It could be a slice.

This rule also means that in $foo[0], $foo is not the array. @foo is the array. Another typical aspect of arrays is that you can use them in array context, like the context imposed on the first argument of push. If you can't push @{\foo()}, "bar";, foo did not return a regular array. If it did return an array and it doesn't work, that is because it was made readonly or magical (tied).

Obey gravity, it's the law.

In other words, there is always a disconnect between what you return and what Perl returns.

Better described as "there can be a disconnect between syntax and semantics". And API documentation is about explaining what happens, not about how you make that happen.

$i++;  # increment $i  # silly.

Comments should describe the "why";
Documentation should describe the "what";
Code describes the "how".

Perl is already well-documented.

And it is time to document your own code well too. Perl set the example, now all you have to do is follow that example. Perl is well known for it's high quality documentation, but wouldn't have been if perlfunc did crazy things like tell people that some function returns an array.

If keys returned an array, not only would it return a list of elements in list context and a number of elements in scalar context, it would also be spliceable. And pushable, and shiftable. And the best of all: it would be referable. Yes, that is silly. But that is what returning an array is.

This is my last node in this discussion. I am unable to convince you. You would defy gravity if you could. I just hope other people who read this thread did learn from it, and don't say that functions return arrays.

[!1]

I would (will and do) document that it returns an array...When a funtion returns an array reference, document that it returns an array reference.

Ummm. Perhaps it just has to do with the way I've been taught to document my code but documenting that a subroutine returns an array is very ambiguous especially considering the fact that you can have a subroutine detect context and return the contents of the array for list context and an arrayref in scalar context (this is the behavior I actually expect when the documentation says array because part of my job involves using mod_perl where you need to watch memory consumption). Of course, you could also have it return the number of elements in scalar context. However, just saying that it returns an array can most certainly be a cause for confusion. Since the purpose of documentation is to make certain that the reader clearly understands how to use your module, avoid using "returns an array".