ICU4X 0.1

[zbraniecki]

ICU4X is aiming to ship its first release of the meta-package at the end of Q3 or beginning of Q4.

The aim of the release is to collect the available subset of components backed by the DataProvider and release them to the community for evaluation and testing.

In this issue we'll discuss the expected scope of the release and considerations we want to take while preparing for it, and all issues and PRs aiming for 0.1 will be listed under the milestone - https://github.com/unicode-org/icu4x/milestone/5 .

[zbraniecki]

The aim of the initial release is to collect the first set of components backed by the DataProvider and release them to the public for evaluation and testing.

Release Driver: @zbraniecki
Milestone

Based on the current velocity my hope is that by the EOQ we'll be able to ship:

Components:

• icu meta crate (Add ICU Meta Package for 0.1 #306)
• icu-locale
• icu-pluralrules
• icu-unicodeset
• icu-datetime

The releases should work as stand-alone, and be available as icu and/or icu4x meta-crate for easy bundling.

Checklist:

• Documentation
  • Per-component
  • Meta-component
  • Introduction documentation and tutorial
  • Directory structure
  • Naming consistency
  • Generated test data
• Test coverage
  • Above 75%
  • Performance CI (Continous Benchmarking #117)
• Data Provider
  • Basic CLDR JSON runtime deserialization
  • ICU4X JSON serialization/deserialization
  • ICU4X bincode serialization/deserialization (Add Bincode format to FsDataProvider #237)
• Locale
  • Raw API - subtags, extensions, etc.
  • proc-macros for subtags, langid and locale (Add macros for locale/langid #75)
• PluralRules
  • Selection API
  • DataProvider use (Switch PluralRules to use DataProvider #189)
  • FixedDecimal use (Switch PluralRules to use FixedDecimal #190)
• DateTime
  • Add Dates to DataProvider (Add subset of dates data to DataProvider #256)
  • Add Dates to CldrJsonDataProvider (Adds support for CldrJSONDataProvider for Dates (based on #256) #258)
  • Export Dates in FsDataProviderExport (Adds Dates support for FsDataProviderExport (based on #258) #261)
  • Formatting API based on DateStyle/TimeStyle (Add DateTimeFormat #202)
  • DateTimeType trait requirements (DateTime entry point data #240)
• UnicodeSet
  • L1
  • L2
• Examples per component and meta-crate
• vs. ICU4C
  • Performance at least on par
  • Memory use at least on par
• ECMA402
  • Example component (PluralRules) integrated into ECMA402 traits (Implements ecma402_traits with icu4x. #271)
• Release
  • Review entry points (README, CONTRIBUTING etc.)
  • Blog post
  • Unicode Conference Presentation (Unicode Conference 44 Presentation #241)

[zbraniecki]

I'd like to suggest accumulating the checklist in this issue and starting a Wiki in this repo with the checklist and release plan there.

[sffc]

Thanks for writing this up!

Initial feedback on the checkboxes:

Data Provider, ICU4X bincode production and runtime: Maybe in another thread, I'd like to make this a little more specific. #200 introduces FsDataProvider, reading JSON from the filesystem, which is modular and easy to understand, but likely costs nontrivial code size and performance. I see a future StaticDataProvider as one that reads from bincode buffers loaded into static memory, likely with higher performance and smaller code size. So, in other words, I see this checkbox as less about bincode and more about a data provider reading from static data rather than the filesystem.

Data Provider, Async read: This is important, but I want to punt it to v0.2. We still have a lot of other even more fundamental problems to solve in the data provider space. I want to introduce an async DataProvider once we have robust code size and performance infrastructure.

UnicodeSet: I would love to see an "L3a" Unicode properties API powered by the data provider, time permitting. If Evan doesn't have enough time during the remainder of his internship, I'm willing to take a stab at this.

[zbraniecki]

Thank you!

So, in other words, I see this checkbox as less about bincode and more about a data provider reading from static data rather than the filesystem.

Maybe we should then separate those two? We can still serialize/deserialize ICU4X data to/from bincode, but separately we want to be able to read from static data.

This is important, but I want to punt it to v0.2.

Fair!

I would love to see an "L3a" Unicode properties API powered by the data provider, time permitting.

Added!

[sffc]

Maybe we should then separate those two? We can still serialize/deserialize ICU4X data to/from bincode, but separately we want to be able to read from static data.

Can you elaborate on what you envisioned when you wrote this bullet point?

[zbraniecki]

Can you elaborate on what you envisioned when you wrote this bullet point?

I think it would be useful to, in parallel to JSON, be able to produce bincode serialized files, store them, and load using DataProvider.

That is, according to what I understand from what you said, separate from loading static data into memory and then deserializing it using bincode.

If you don't think it is useful, I can remove it.

[sffc]

Ok, thanks for clarifying. Doing that piece should be a small PR based on #200. I'm not sure whether it's "hard enough" to rise to the level of a bullet point on the v0.1 checklist, but I don't feel strongly.

[filmil]

Looking at this now.

[filmil]

Seems reasonable to me.

One small request, would you be willing to let Locale implement ecma402_traits::Locale for the 0.1 release?

[zbraniecki]

One small request, would you be willing to let Locale implement ecma402_traits::Locale for the 0.1 release?

That's on the list as Components integrated into ECMA402 traits - my intention was to get Locale, PluralRules and a subset of DateTime to implement the traits.

[zbraniecki]

I updated the checklist, added 0.2 issue and wrote initial Release Process doc.

[zbraniecki]

ICU4X 0.1 released!

• https://crates.io/crates/icu
• https://crates.io/crates/icu_locid (and other components)

Thank you everyone! Onto the next milestone - https://github.com/unicode-org/icu4x/milestone/7 :)