Single-element lists should require a trailing comma (maybe)

[GrigorenkoPV]

From what I can tell,

for 0..=10 do (i) {
	std::cout << i << std::endl;
}

iterates over a range of ints, 1 through 10, while

for (0..=10) do (i) {
	std::cout << i << std::endl;
}

iterates over a list of cpp2::range<int>s, which is probably due to the fact that (0..=10) gets treated as a single-element least.

I, personally, find it a bit confusing. I would not expect that adding a pair of parenthesis around an expression fundamentally changes it. And while I know that cpp2 currently uses (...) instead of [...] for lists, intuitively I do expect such behavior from [...], but not from (...).

For example, in Python, there are essentially two types of lists commonly used. The [...] "lists" (variable-length, a-la std::vector¹) and the (...) "tuples" (fixed length, a-la std::array¹). Python allows trailing commas (thankfully), but for single-element tuples it requires a trailing comma for disambiguation.

def show(x):
    print(type(x).__name__, x)

show(  1     ) # int 1
show( (1)    ) # int 1
show( (1,)   ) # tuple (1,)
show( (1,2)  ) # tuple (1, 2)
show( (1,2,) ) # tuple (1, 2)
show( ()     ) # tuple ()
show( (,)    ) # SyntaxError: invalid syntax
show( [1]    ) # list [1]
show( [1,]   ) # list [1]
show( [1,2]  ) # list [1, 2]
show( [1,2,] ) # list [1, 2]
show( []     ) # list []
show( [,]    ) # SyntaxError: invalid syntax

So, should cpp2 do something similar?

Footnotes

1. Except of course Python allows storing elements of different types, because it is dynamically typed. Essentially it stores a vector/array of objects. ↩ ↩²

[hsutter]

From what I can tell,

for 0..=10 do (i) {
	std::cout << i << std::endl;
}

iterates over a range of ints, 1 through 10, while

Yes.

for (0..=10) do (i) {
	std::cout << i << std::endl;
}

iterates over a list of cpp2::range<int>s, which is probably due to the fact that (0..=10) gets treated as a single-element least.

Yes, the ( ) denote a list of values, and that's a list of one value which is a range.

You can find the same effect in non-range examples, such as for r vs for (r) where r is some container or range:

for  r  do (e) { } // ok, e is an element of r
for (r) do (e) { } // ok, e is r

Consider that we do want syntax like ( ) to make a list of things right there in the for statement:

for (1, 2, 3) do (i) { } // ok
for (1, 2)    do (i) { } // ok
for (1)       do (i) { } // ok
for  1        do (i) { } // error, 1 is not a range

So the status quo for a for loop is that when you write a list of individual things you use ( ), and when you write a range you use ... or ..=, and here's how they look side by side:

for (1, 2, 3) do (i) { } // ok
for  1 ..= 3  do (i) { } // ok, same meaning
for  1 ... 4  do (i) { } // ok, same meaning

Python allows trailing commas (thankfully),

(BTW so does Cpp2 now, thanks again to everyone who motivated me to finally add that.)

but for single-element tuples it requires a trailing comma for disambiguation.

That's the kind of thing I would worry about.

[svew]

Remind again, was there a reason not to use [ ] to denote lists instead of ( ), or just a choice made?

[sookach]

Just a choice made by Herb

[SebastianTroy]

It was a well reasoned choice though 😂 so yes, there were reasons, I was on the side of using [] initially, but I was converted, though it was a while ago so I don't remember specifics On 2 August 2024 01:48:15 Andrew Sukach ***@***.***> wrote: Just a choice made by Herb — Reply to this email directly, view it on GitHub<#1179 (reply in thread)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AALUZQLGKPW5W7YLTXF76GDZPLJMZAVCNFSM6AAAAABLMZ3LFKVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTAMRRG44DEMQ>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>