JingsongLi opened a new pull request, #9480: URL: https://github.com/apache/paimon/pull/9480
## What is changed - Revert #9473 and restore the aligned file-group model in Paimon’s data-evolution layer. - Add LeRobot v3 import support, with normalization performed per episode. - Roll normal files only at episode boundaries so normal and dedicated-format files remain aligned. - Support multiple video fields when their episode boundaries are nested. - Reuse source video shards when possible and remux only when necessary. ## Import strategies 1. **Exact boundaries:** reuse all source shards directly at their common boundaries. 2. **Nested boundaries with an acceptable common span:** import at the coarsest shared boundary. 3. **Oversized or incompatible boundaries:** reshard/remux in the LeRobot import layer. This keeps boundary normalization in the LeRobot-specific Python import path instead of adding general spanning-sidecar behavior to Paimon scan, split, and read internals. ## Commit structure 1. Complete revert of #9473. 2. LeRobot v3 import baseline. 3. Episode-aligned normalization and writer support. 4. Adaptive shard reuse / remux strategy. ## Verification - Python: 112 passed, 6 skipped - Java: 19 passed - Spotless: passed for paimon-api and paimon-core - Final tree is byte-for-byte identical to the previously validated implementation branch. -- 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]
