Option to retrieve words as opposed to numbers about proposal-intl-relative-time

[icambron]

-1. It isn't supported by the Intl.NumberFormat class, so I don't think it would make sense to support it just for relative time. I'm sure it was discussed when making Intl.NumberFormat and I don't know why they declined to add it, but if I had to guess:

1. Those style guides are about using numbers in prose. It's not clear to me that they apply to the most common use cases for Javascript number formatting.
2. The rules are complicated, e.g. both the major en-US style guides have rules like "if it's at the beginning of a sentence, write it out"
3. The rules are varied. The sytle guides do not, in fact, agree about stuff like this, even for a specific language. For example, the Chicago Manual wants words for up to 100, whereas the AP guide says 1-10.

[zbraniecki]

There's an API for spelling out numbers - http://userguide.icu-project.org/formatparse/numbers/rbnf-examples

Not sure if we want to aim to extend numberformat and this with it in the future but may be worth trying to figure out how such API might look like

[littledan]

@kdex Thanks for coming here and explaining this issue. Given that this proposal is already at Stage 3 in TC39, would you be up for revisiting it as a possible "v2" feature to come after Intl.RelativeTimeFormat ships?

@icambron Thanks a lot for the detailed explanation. As for (1), I would think that there's lots of good applications for using Intl to construct prose. I fully understand (2) and (3) though.

@littledan In general, yes, I'm up for revisiting this, but I'm not sure what the extent of such an feature would entail, as written out numbers aren't necessarily unique. Consider 101:

• one hundred and one
• a hundred and one
• one hundred one
• a hundred one

I'm very drawn to say that, if anything, we might want to extend Intl.NumberFormat to allow writing out numbers, and only then would revisiting this in Intl.RelativeTimeFormat make sense.

[littledan]

@kdex In general, this standard has been not innovating on internationalization data itself but instead following the patterns of other libraries and databases such as CLDR and ICU, and exposing something similar to JavaScript. I'd recommend working in one of those places to solve the underlying problem, and then figure out how to map the solution to JavaScript.

@littledan Thanks a lot for the pointers; I'll be closing this issue for now.

[littledan]

@kdex Feel free to check back here or at #tc39 on irc.freenode.net if you need to reach us for any more discussion.