Manish Goregaokar

I think this is in scope as a utils crate we publish separately, perhaps with icu_locid integration. I don't consider this priority.

Discussed: - @robertbastian: should people even be making giant blobs? slice them! - @manishearth: monolithic data bundle is often convenient from a versioning and distribution standpont - @zbraniecki: it's what...

Going to try and knock this out. Hopefully the following patch is sufficient to repro: ```diff diff --git a/tools/testdata-scripts/src/bin/make-testdata-legacy.rs b/tools/testdata-scripts/src/bin/make-testdata-legacy.rs index 46685cacc..93a908feb 100644 --- a/tools/testdata-scripts/src/bin/make-testdata-legacy.rs +++ b/tools/testdata-scripts/src/bin/make-testdata-legacy.rs @@ -45,7 +45,7...

That didn't work. Hm.

``` $ RUST_BACKTRACE=1 cargo run --bin icu4x-datagen -- --all-keys --locales recommended --format blob2 --out test.postcard ``` works

@sffc do you have a suggestion for a good way to structure a test for this? We could have it generate some fake key for thousands of fake locales, or...

Yeah that's basically what I was wondering for "generate thousands of locales". `abc-DE` seems easy.

https://github.com/unicode-org/icu4x/pull/4856

There's a PR open for Chinese (https://github.com/unicode-org/icu4x/pulls), but islamic/hebrew also need similar work done.

#4051 solves this for the milestone, kicking the can to 1.4