Reverse indexing should be 0-based

[jridgewell]

I want make a parallel to regular forward indexing, and the easiest example is array iteration:

for (let i = 0; i < array.length; i++) {
  console.log(array[i]);
}

This form of = 0, < array.length, i++ is soooo common it's burned into my mind. Importantly, forward indexing is 0-based.

Now, compare that to the 1-based reversed indexing currently proposed:

for (let i = 1; i <= array.length; i++) {
  console.log(array[^i]);
}

That = 1 and <= is going to drive me crazy. It seems like if make reverse indexing 0-based instead, we can mirror C++'s std::rbegin (which is a 0-based iterator, it immediately points to the last item):

for (let i = 0; i < array.length; i++) {
  console.log(array[^i]);
}

[ljharb]

While i agree with the comparison, having ^0 be the -1st item will be very confusing.

[hax]

@jridgewell This question is well answered here: Why doesn't the new hat-operator index from the C# 8 array-slicing feature start at 0?

The main motivation of index from end syntax is not for iteration (which I believe reverse iterable would be a feature to satisfy such use cases in a better form), but for random accessing from end, and slice notation, in those cases, make ^0 represent "the end" (position after the last element) would be much convenient.

Note, of coz we could consider using C++ rbegin model, but:

1. As current proposal, ^i could be think as a better -i (no -0 footgun), which make it very easy to get for the programmers have python, ruby experience (and JS already have slice/splice using -i). On the other side, as jordan said, "having ^0 be the -1st item will be very confusing" for them.
2. Actually if follow rbegin, we can't use ^i syntax anymore because make same syntax do very same thing but with a significant semantic difference with another mainstream language like C# will cause very big confusion and discord. So we have to invent some other syntax, and it's really really hard.

[bmeck]

@jridgewell this would directly go against my being able to do arr[^0] = newTail;, having arr[^-1] would be wildly confusing

[hax]

As previous discussions, I think as the goal and constraint of this proposal and related proposal, it's very unlikely we could make a syntax using 0 to access the last (-1st) item. So I close this issue.

I add a FAQ item in README to link back to this issue, so if people have similar concerns they can continue the topic here, and we can reopen if needed.