SEPURI-SAI-KRISHNA commented on PR #13161:
URL: https://github.com/apache/seatunnel/pull/13161#issuecomment-6943952491

   @nzw921rx Thanks for checking this in an isolated copy. Both points are 
fixed in `16d9947645`.
   
   **Binding test.** 
`DatabendSinkWriterTest#testWriteBindsTheAddTimeOfEachCdcRowAsAStringWithMicroseconds`
 builds a real CDC writer with a mocked `Connection` and `PreparedStatement` 
and goes through `write()` and `processCdcRow()`. The writer now takes the 
clock from a field that the test replaces, so three rows are written in one 
fixed millisecond. The test captures parameter 4 and asserts that:
   
   - it is bound with `setString`, three times, and `setTimestamp` is never 
called for it;
   - the values keep their microseconds: `2023-11-14 22:13:20.123`, `.123001` 
and `.123002` (UTC);
   - each value is strictly later than the one before.
   
   With the binding reverted to `setTimestamp`, this test fails, and it also 
fails when the string is built without the microsecond step. The other 13 
writer tests still pass in both cases. Connector tests: 28 run, 0 failures.
   
   **Docs.** The EN and ZH upgrade notes no longer say the latest event always 
wins. They say the merge orders by `add_time`, which is local time of the JVM, 
so in the hour that repeats when daylight saving time ends an event can still 
lose to an older event of the same key in the same merge; that this limit is 
not new; that a time zone without daylight saving time, such as UTC, does not 
have it; and that events from different writers are ordered only by their own 
clocks. The sink pages say the same in two sentences.
   


-- 
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]

Reply via email to