Name CancelToken.any

[domenic]

Possible names:

• CancelToken.any
• CancelToken.some
• CancelToken.linked (meh)

[ljharb]

Can you elaborate on the semantic/meaning/purpose of this method, to help contextualize the naming?

[domenic]

https://docs.google.com/presentation/d/1kSY7X1ymw5f2oatDrZaMJh4Z_wpd0ynZIimhUyqhbrY/edit?usp=sharing

[ljharb]

Which slide in particular?

[domenic]

I guess my attempts at deep-linking didn't work... slides 20 through 22.

[ljharb]

It seems like either Promise.race or Array#some is the analog here - i think either "race" (for symmetry with Promise) or "some" (for conceptual symmetry with Array.prototype) is probably the best approach? I think "any" might be a better name, but then I think that's a better name for "Array#some" also :-p

[domenic]

The discussions we had for promise are coming back to me now. I think they went something like this...

• "some" should generally be reserved for things with a boolean predicate. E.g., for promises, it would make sense as something that returns a promise fulfilled with true if any of the given promises fulfill with true. (Or rejects if any of the given promises reject---like rethrowing exceptions in sync Array.prototype.some.) It is not appropriate for working with the fulfilled/rejected state, which is not about boolean true/false. Similarly, "every" is also a boolean thing.
• Thus, we invent the pair "all" / "any" to be a counterpart that operates on the fulfilled/rejected axis. These are not about booleanness, but instead about success or failure. Their results are not a single boolean, but something more specialized. But they are biased toward fulfillment: "all" = "all of them fulfilled"; "any" = "any of them fulfilled". That is, like some/every, they treat one class of outcome as "good".
• "race" is in a separate category. It's not biased toward fulfillment; any outcome at all is acceptable. Thus it gets a name that doesn't remind us of quantifiers or boolean logic or anything like that.

Now, how do we apply this to cancel tokens? Hmm.

• They don't have potentially-boolean values, so if we used "some", there'd be no potential for confusion (like there would be with promises). However, since there's no boolean values, maybe we shouldn't use the boolean word "some".
• They don't have a fulfillment/rejection dichotomy; it's more "hasn't been canceled yet" / "has been canceled". Is that enough that "all" / "any" is a good enough pair? Or should we treat it as "any outcome" and go with "race"?
• If we chose "race", what would the counterpart be called for the "all"/"every" case? We won't spec that, since the committee could not find any use cases and it has no precedent in other languages. But it seems nice to have a consistent set of words. Meh, maybe this isn't important; when I think about it, I just come up with a bunch of silly words like "tandem" or "together" or something.

I guess this has me leaning toward either "any" or "race".

[ljharb]

Ah, so you're saying that Promise.any is another possible static method that could exist in the future, that would reject only if all of the promises rejected, and would fulfill if any of them resolved? That makes tons of sense to me.

The counterpart to race could be "together", or "complete", or "finished", or "fufilled", or "settled" - but yeah it's a tricky one.

Given that promises have three states (pending, resolved, rejected), and "race" says "the first one that's not pending wins" - altho cancel tokens have two (pending, cancelled) - "race" might apply best.

[domenic]

Ah, so you're saying that Promise.any is another possible static method that could exist in the future, that would reject only if all of the promises rejected, and would fulfill if any of them resolved? That makes tons of sense to me.

Yeah. I really want it; the problem is designing the "AggregateError" type that you use to signal error if they all reject. I thought that would be a Bikeshed factory. But then again, cancelable promises is going OK, so maybe I should try that :).

[ljharb]

Seems easier than coming up with a way to do Promise.settle and have a way to distinguish rejection from resolution :-p i say go for it! :-D

[benjamingr]

race and any sound equally good and race is terminology we're using for something similar anyway - so +1 on that.

[bergus]

@domenic Don't forget to rename it in Cancel Tokens.md as well