hutiefang76 opened a new issue, #12658: URL: https://github.com/apache/seatunnel/issues/12658
### Search before asking - [x] Searched current issues and pull requests; related prior work is identified below. ### What happened CSV and Text deserializers cache the timestamp formatter selected by the first row. A later supported value with a different fractional-second precision is then parsed using that stale formatter and fails. ### SeaTunnel Version dev at 11e4f5554973a66ba2d63d4df18ee22f10be121f. No claim about untested releases. ### SeaTunnel Config Use a two-column schema (`name` STRING, `ts` TIMESTAMP), comma delimiter, and the same deserializer instance for these successive rows: ```text alpha,2023-01-01 00:00:00 bravo,2023-01-01 00:00:00.123 charlie,2023-01-01 00:00:00.123456 ``` Repeat with TextDeserializationSchema and CsvDeserializationSchema. Reverse the precision order to exercise cache replacement in both directions. ### Running Command Call deserialize for each row and assert the returned LocalDateTime including its nanoseconds. This reproduces the reader behavior without starting an engine. ### Error Exception A DateTimeParseException is raised when a later value uses a different supported precision. Expected: parse every row using a matching supported pattern; unsupported text must still fail. ### Zeta or Flink or Spark Version Not required for the isolated utility/format reproduction. ### Java or Scala Version JDK 17.0.18; regression cases also passed on JDK 1.8.0_492 in the original local repair batch. ### Anything else Merged PR #11264 addressed a similar cache problem in the JSON format. CSV and Text retain the stale-formatter behavior on the current dev source and are the scope here. `timestamp_format` is a sink serialization option, not a strict source parsing pattern. ### Are you willing to submit PR? - [x] Yes I am willing to submit a PR. ### Code of Conduct - [x] I agree to follow this project's Code of Conduct. AI assistance: DeepSeek and Codex assisted reproduction, implementation, tests and this description. -- 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]
