Kind of surpised by this one... ICU 73.1 already contains timezone 2023c.

Kind of surpised by this one... ICU 73.1 already contains timezone 2023c.

The PR/commit message is misleading, but it does appear to remove an unneeded file, I think?

I doesn't remove the file. It reduces its size (probably by removing some data in it).

/cc @srl295

Kind of surpised by this one... ICU 73.1 already contains timezone 2023c.

It should perhaps check process.versions.tz before running.

I doesn't remove the file. It reduces its size (probably by removing some data in it).

Maybe just a re pack. I'll check it sometime today.

in tools/update-timezone.mjs

renameSync(`icu-data/tzdata/icunew/${latestVersion}/44/le/${file}`, `deps/icu-small/source/data/in/${file}`);

this should be a copy instead of a move. Otherwise it's destructive to the icu-data repo :)

uncompressed the files are identical in size (but different in content) so the 500 byte 'reduction' was a recompression.

$ ls -ld icudt73l*
-rw-r--r-- 1 srl295 wheel 31998544 Apr 24 09:44 icudt73l.dat
-rw-r--r-- 1 srl295 wheel 31998544 Apr 24 09:45 icudt73lPLUS.dat

./tools/icu/shrink-icu-src.py has

# compression stuff. Keep the suffix and the compression function in sync.
compression_suffix = '.bz2'
def compress_data(infp, outfp):
    with open(infp, 'rb') as inf:
        with bz2.BZ2File(outfp, 'wb') as outf:
            shutil.copyfileobj(inf, outf)

whereas tools/update-timezone.mjs has

execSync('bzip2 -z deps/icu-small/source/data/in/icudt*.dat');

also see #47302

ok my analysis tools are in #47248 (comment) (document somewhere?)

$ diff ICU0/timezoneTypes.txt ICU1/timezoneTypes.txt
10d9
<             camtr { "cator" }
35d33
<             America:Montreal { "America/Toronto" }
308a307
>             America:Montreal { "camtr" }

this shouldn't have made a change, but it did. Investigating.

Update This is a bad change (regression).

@yumaoka any ideas here? Node.js has ICU4C 73.1, and this script "updated" it using https://github.com/unicode-org/icu-data/tree/main/tzdata/icunew/2023c/44 (It's a bug, because 73.1 is already at 2023c)

But it looks like the icu-data repository doesn't reflect unicode-org/icu@2c584ab#diff-6b5713193269b4b1d788bdb509f542cf6c6148ea8821174310b73ecf86b0b1cd

• your commit to icu-data: March 29, 2023 unicode-org/icu-data@55248ee
• the icu4c maint-73 commit was March 30, 2023 unicode-org/icu@2c584ab

Is icu-data out of date?

Is icu-data out of date?

Probably a better way to say it is: icu-data's 2023c update is valid for ICU4C 72.1 as it does contain tzdata 2023c. however, it doesn't include the CLDR 43 changes that went into ICU4C 73.1. Since ICU4C 73.1 already included 2023c, one wouldn't expect users to update from 2023c "to" 2023c. So it is kind of an edge case.

If there was a 2023d (always likely), it will undoubtably contain the CLDR 43 changes.

Anyway, I propose to close this PR and fix issues in #47702