ICU4X 1.2 release checklist

[Manishearth]

Extra pre-checklist checklist

• [x] Update Diplomat (https://github.com/rust-diplomat/diplomat/pull/323, https://github.com/unicode-org/icu4x/pull/3304)
• [x] (@Manishearth) Write a changelog (draft: https://hackmd.io/@Manishearth/BkIpZnSGh/edit)
• [x] Utils bumps https://github.com/unicode-org/icu4x/pull/3308

Main checklist

• [x] (@manishearth) Run cargo +nightly fmt -- --config=format_code_in_doc_comments=true --config=doc_comment_code_block_width=80 to prettify our docs (https://github.com/unicode-org/icu4x/pull/3309)
• [x] (@sffc) Run cargo update for each Cargo.lock file to update our CI to freshest dependencies
• [x] (@sffc) Verify that the milestone and checklist are complete
• [x] Verify with component owners that they're ready for release
• [x] (@Manishearth) Verify that ffi/diplomat depends on a released (not Git) version of Diplomat. Get it published (ask manishearth or sffc) otherwise.
• [x] Verify that we have acceptable FFI coverage (should be a part of the checklist issue)
  • [x] Verify that ffi/diplomat/tests/missing_apis.txt is empty. If not, check with the team that we are okay punting these to the next release.
  • NOTE: We can ship 1.2 with a known issue that not all APIs are covered.
• [x] Take a bird-eye view at:
  • [x] (@sffc) READMEs
  • [x] (@echeran) Documentation
  • [x] (@Manishearth) Coverage
  • [x] (done in meeting) Performance / Memory / Size benchmarks
  • [x] (@sffc) Cargo.toml files
  • [x] Cargo.toml files need to specify versions for each entry in dependencies.
  • [x] Cargo.toml should not specify versions for each entry in dev-dependencies.
  • [x] Ensure that any new packages have suitable version numbers.
  • [x] Ensure that the Cargo.toml file includes a set of fields consistent with other ICU4X packages.
• [x] (@sffc) cargo workspaces version --no-git-push to bump the version numbers
  • [x] This will only update crates that have changed, and will ask you which version number to bump for each crate
  • [x] You can use commands like git log icu@1.0.0..@ -- components/plurals/src/ and cargo public-api -p icu_list diff 1.0.0 to figure out whether to do a major, minor, or patch release
  • [x] Get this reviewed and checked in before continuing
• [x] (@Manishearth) Use cargo workspaces publish --from-git to automatically publish the crates in the correct order
  • [x] Add icu4x-release group as owners to each new component you're publishing
  • [x] cargo owner -a github:unicode-org:icu4x-release
• [x] (@sffcManishearth Verify that the tutorial crates work on the newly released crates.io sources by running cargo make test-cargo-tutorial-cratesio
  • [x] If there are any errors, please fix them before advertising the release
• [x] (@Manishearth) Tag the Release

To do in the coming days

• [x] Announce the release to public
  • [x] Send an email to icu4x-announce

[Manishearth]

We should also start drafting a changelog

Edit: I have a draft in https://hackmd.io/@Manishearth/BkIpZnSGh/edit, please help out

[Manishearth]

Changelog is done. @sffc please note that in the experimental section I have marked a bunch of crates as "no other changes", the only changes are edition bumps and clippy fixes. If you want you can publish new versions for them, in which case please note down the version diff since 1.1, otherwise please have it say "No change (still at x.y.z)", the same way we do for the utils.

The idea is that since we only write changelogs for ICU4X releases, we should cover all intermediate versions.

[Manishearth]

Coverage seems fine. icu_properties has low coverage because it's macro-heavy and the tool is very inconsistent about counting that well.

[Manishearth]

note to self: workspaces tool should have --no-git-tag

[Manishearth]

Release is published and tagged. Heading for lunch

[sffc]

For line break, ICU4X 1.2 contains these data entries:

segmenter/lstm/wl_auto@1/my.postcard
segmenter/lstm/wl_auto@1/km.postcard
segmenter/lstm/wl_auto@1/lo.postcard
segmenter/lstm/wl_auto@1/th.postcard
segmenter/grapheme@1/und.postcard
segmenter/line@1/und.postcard

gzipped together, these are 299959 B.

The equivalent in ICU4C 72 should be:

brkitr/burmesedict.dict
brkitr/char.brk
brkitr/khmerdict.dict
brkitr/laodict.dict
brkitr/line.brk
brkitr/line_cj.brk
brkitr/line_loose.brk
brkitr/line_loose_cj.brk
brkitr/line_loose_phrase_cj.brk
brkitr/line_normal.brk
brkitr/line_normal_cj.brk
brkitr/line_normal_phrase_cj.brk
brkitr/line_phrase_cj.brk
brkitr/thaidict.dict

gzipped together, those files are 762631 B.

[sffc]

Comparing for the other three segmenter types:

Grapheme:

ICU4C:
charbrk/char.brk

ICU4X:
charbrk/grapheme@1/und.postcard

Sentence:

ICU4C:
sentbrk/sent.brk

ICU4X:
sentbrk/sentence@1/und.postcard

Word:

ICU4C:
wordbrk/burmesedict.dict
wordbrk/char.brk
wordbrk/laodict.dict
wordbrk/word.brk
wordbrk/thaidict.dict
wordbrk/khmerdict.dict
wordbrk/cjdict.dict

ICU4X:
wordbrk/word@1/und.postcard
wordbrk/w_auto@1/ja.postcard
wordbrk/grapheme@1/und.postcard
wordbrk/wl_auto@1/my.postcard
wordbrk/wl_auto@1/km.postcard
wordbrk/wl_auto@1/lo.postcard
wordbrk/wl_auto@1/th.postcard

All of the above:

ICU4C:
allbrk/sent.brk
allbrk/line_normal_cj.brk
allbrk/burmesedict.dict
allbrk/line_loose_phrase_cj.brk
allbrk/char.brk
allbrk/line_cj.brk
allbrk/line_phrase_cj.brk
allbrk/laodict.dict
allbrk/word.brk
allbrk/thaidict.dict
allbrk/line_loose_cj.brk
allbrk/khmerdict.dict
allbrk/cjdict.dict
allbrk/line.brk
allbrk/line_loose.brk
allbrk/line_normal.brk
allbrk/line_normal_phrase_cj.brk

ICU4X:
allbrk/segmenter/word@1/und.postcard
allbrk/segmenter/sentence@1/und.postcard
allbrk/segmenter/dictionary/w_auto@1
allbrk/segmenter/dictionary/w_auto@1/ja.postcard
allbrk/segmenter/lstm/wl_auto@1/my.postcard
allbrk/segmenter/lstm/wl_auto@1/km.postcard
allbrk/segmenter/lstm/wl_auto@1/lo.postcard
allbrk/segmenter/lstm/wl_auto@1/th.postcard
allbrk/segmenter/grapheme@1/und.postcard
allbrk/segmenter/line@1/und.postcard

Gzip Sizes:

Type	ICU4C	ICU4X	% Reduction
Char	2889	2438	15.6%
Sentence	4352	4295	1.3%
Word	2286466	1852208	19.0%
Line (from above)	762631	299959	60.7%
All of the above	2330198	1862568	20.1%