Define string size about infra

[domenic]

I think I agree the latter is better.

Nit: should probably be "length" instead of "size" as is traditional for strings.

[annevk]

I think maybe just as with size, we should use for to distinguish. So we can have for="JavaScript string" (code units), for=string (code points), and for="byte sequence" (bytes).

We can do the same for sorting, which we also need for byte sequences (see Fetch).

I wanted to start working on this but I'm blocking myself on #73 since I don't want to deal with merge conflicts. Suggestions for how to formally define sorting would be welcome by the way.

[domenic]

Hmm. I guess if people are being very good about distinguishing their string types, then just using "length" and linking to the appropriate definition (with for="") should work OK. The benefit of having explicit separate length concepts (e.g. code-unit length and code-point length) is that it makes you re-emphasize the difference at the point at which you determine the length. But if people are careful with their types that's not necessary.

Formally defining sorting: based on https://en.wikipedia.org/wiki/Sorting_algorithm, the key is to define an ordering relation (x <= y) and then state that the sorted version is the unique permutation of the list such that it's in non-decreasing order. This permutation will not be unique if it's possible to have two things that are equal (x <= y && y <= x) but are not truly identical, e.g. if two strings have identity (which I don't believe they should).

Maybe it'd be best to just define ordering relations actually, and leave sorting with its English/computer science definition. Once we have well-defined ordering relations it's pretty obvious IMO.

[annevk]

For lists we use "size" rather than "length", contradicting ECMAScript. Using size consistently might be easier (since you just need to remember the correct for value). WDYT?

[domenic]

ECMAScript is a bit weird. Most of its data structures (Maps and Sets) use size. I think if arrays were designed today they might use size. Spec "List" type does use length though, it's true.

I think it's best to not blur the line between data structures and strings/byte sequences, and using different words (size vs. length) helps enforce this I think.

[annevk]

Given what List does I we should also change lists while we can (or maybe accept both for a while). Having one exception is going to be annoying.

[domenic]

I don't understand. What is the exception? What would we change about lists?

[annevk]

If we align with ECMAScript on size/length except for lists, lists would be an exception. Therefore we'd change it to use length as well.

[annevk]

(Or we consistently use size.)

[domenic]

I think it's OK for infra to set its own rules here. For data structures, infra uses size. For strings, it uses length. What ECMAScript does can be separate.

I don't think we should make an exception for lists and say they use length even though all other data structures use size.