SEPURI-SAI-KRISHNA commented on PR #12976: URL: https://github.com/apache/seatunnel/pull/12976#issuecomment-6410956453
@DanielLeens both carryover findings are real. I ran them on JDK 17 rather than take them on source reading, and both are fixed at `70dcb475d`. **Issue 2, confirmed.** `WeekFields.of(locale)` gives SUNDAY/1 for `en-US` and `Locale.ENGLISH`, MONDAY/4 for `en-GB` and `de-DE`, MONDAY/1 for `tr-TR`. Measured for `2021-01-01`, a Friday: | pattern | en-US | en-GB | de-DE | tr-TR | now | | --- | --- | --- | --- | --- | --- | | `YYYY` | 2021 | 2020 | 2020 | 2021 | 2021 | | `ww` | 01 | 53 | 53 | 01 | 01 | | `W` | 1 | 0 | 0 | 1 | 1 | | `e` | 6 | 5 | 5 | 5 | 6 | So the "digits and separators are unaffected" line was wrong, and `en-GB` proves your point about English regional defaults. Both upgrade notes now split Impact into text, localised week fields and unaffected, carry that table, and tell anyone who keyed rows on `Y w W e c` to compute the week upstream. Both function pages got the same note. New regression `testFormatDateTimeWeekFieldsUseEnglishWeekRulesUnderAnyDefaultLocale` runs the real `SQLTransform` under ENGLISH, `en-GB`, `de-DE` and `tr-TR` with a day-of-year control. It is a unit test, not an addition to `func_datetime.conf`, because the E2E container runs one fixed default locale and so cannot express "same output under four different worker defaults". I also checked the `ZetaDateTimeFormat` claim instead of repeating it: all 13 constants are `y M d H m s S` plus literals. **Issue 3, confirmed.** With a fixture whose categories differ from the base: ```text fixture base=en_US FORMAT=de_DE DISPLAY=fr_FR old cleanup, base only -> base=en_US FORMAT=en_US DISPLAY=en_US new cleanup, all three -> base=en_US FORMAT=de_DE DISPLAY=fr_FR ``` Added `snapshotDefaultLocales` and `restoreDefaultLocales` and moved both existing locale tests onto them, taking the full snapshot rather than the FORMAT-only variant. The order is commented: restore the base first, because setting it clobbers both categories. **Validation here.** `spotless:check` clean, `seatunnel-transforms-v2` 1492 tests, 0 failures. Mutation check: reverting both `ofPattern(format, Locale.ENGLISH)` sites fails 3 tests, the new one on `en-GB` with `expected: <2021> but was: <2020>`. One caveat I would rather state than leave implied: that mutation run stops at the first failed assertion, so the newly added era, quarter and am/pm assertions are evidenced by direct measurement of each pattern under `tr-TR`, `ja-JP` and `de-DE`, not by the mutation run itself. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
