Repository navigation
Intl: Consider deprecating Intl.v8BreakIterator #8865
Description
@littledan has put forward a TC-39 proposal to introduce Intl.Segmenter as a standardized replacement for the V8 specific Intl.v8BreakIterator.
Currently in Node.js, Intl.v8BreakIterator only works if the full-icu data is being used. Attempting to create an instance of Intl.v8BreakIterator with the default small-icu with throw. That said, given the desire to replace it with a standardized alternative we should consider emitting a runtime deprecation warning when it is used.
Activity
srl295 commentedon Oct 1, 2016
Makes sense… is there a need to deprecate something that's really v8 internal? So I'm +0.0 on this subject to rounding err!
littledan commentedon Oct 1, 2016
I hope no one new starts using the deprecated API, but it's currently exposed to some Node users. I'd support a runtime deprecation warning but would be even happier if it were just deleted. I hope to remove it entirely from a future version of V8.
srl295 commentedon Oct 1, 2016
Deletion makes sense too. I'm +1 on principle just not sure about policy details.
Enviado desde nuestro iPhone.
jasnell commentedon Oct 2, 2016
We have to deprecate first in one major before we can remove in the next.
On Saturday, October 1, 2016, Steven R. Loomis notifications@github.com
wrote:
srl295 commentedon Oct 2, 2016
caridy commentedon Oct 3, 2016
@srl295 I thought we had discussed this in the past, and agreed on throwing an error, at least that was my understanding. Anyhow, let's try to take action asap, to prevent a bigger problem later on.
Fishrock123 commentedon Oct 3, 2016
If there is a chance this will be removed before the next major we should deprecate in v7.
jasnell commentedon Oct 3, 2016
I'm thinking we probably should. I'm not able to add the ctc-agenda label
at the moment. We would need ctc ok to land it in v7.
On Monday, October 3, 2016, Jeremiah Senkpiel notifications@github.com
wrote:
intl: add deprecation warning for v8BreakIterator
targos commentedon Oct 3, 2016
Here is a PR adding the deprecation warning: #8908
rvagg commentedon Oct 5, 2016
One real problem here is that if
v8BreakIteratorgets removed in a near-future version of V8 then it will increase the difficulty in upgrading V8 mid-major, we'll either have to polyfill/backfill in a patched version of V8 when we upgrade to avoid breakage or break for users who may be using it (do we really think folks will be using it?).The problem here is the speculative nature of the problem here.
Intl.Segmenteris before TC39, but that could take many months, no?and
@littledan:
What timeframe might we be talking about? Would this depend on TC39? What if
Intl.Segmenterdoesn't get accepted by TC39? WillIntl.v8BreakIteratorbe removed regardless?rvagg commentedon Oct 5, 2016
Some discussion at CTC meeting, lots of uncertainty still, @littledan is on vacation for a few weeks apparently. This should come back on the agenda.
4 remaining items
littledan commentedon Oct 18, 2016
To give a little more context on top of what @s3ththompson,
Intl.Segmenterwill take at least six more months, and could be even longer (a year or two), but I'm optimistic that it'll eventually make it through standardization.intl: unexpose Intl.v8BreakIterator
intl: unexpose Intl.v8BreakIterator
AaronPresley commentedon Apr 8, 2019
Hi all,
I'm looking to integrate segmentation in some node (v10) code.
From what I understand after a twitter chat with @srl295,
Intl.v8BreakIteratorhas been completely removed; however its replacement ofIntl.Segmentationis still in Stage 3 and not yet approved.How would you recommend I use these APIs in node? Am I just stuck with an older version that supports
Intl.v8BreakIterator? Would appreciate some guidance.Secondly, would also love to understand why this feature was completely removed before a suitable replacement was approved? Or am I misunderstanding something?
Thanks for your time.
devsnek commentedon Apr 8, 2019
if you only need en-US, this should work. https://github.com/proxy/gist.github.com/inexorabletash/8c4d869a584bcaa18514729332300356
AaronPresley commentedon Apr 8, 2019
@devsnek Unfortunately en-US is the least of my worries. Mostly working with CJK languages.
srl295 commentedon Apr 8, 2019
Because it wasn't actually a Node.js feature. It was an unsupported, lightly documented
undocumented, non standard, internal function in v8. It tended to cause a process crash (#3111) rather than a graceful response to data loading or parameter errors.A couple of other interim options which link against another copy of ICU4C might include: https://www.npmjs.com/package/node-icu-tokenizer or https://www.npmjs.com/package/icu-wordsplit — maybe these could be made to work with the new API.
AaronPresley commentedon Apr 8, 2019
Ah, that makes sense then. Thanks for the explanation. That would also explain why it's been so impossible to find documentation for that function :)
littledan commentedon Apr 8, 2019
It used to be documented in a Google Code repository, but that was removed when Google Code was shut down, and I'm not sure where to find it in the Internet Archive :(
srl295 commentedon Apr 8, 2019
@littledan My mistake! I updated my commend above with a link to the docs ☝️
intl: add deprecation warning for v8BreakIterator