JingsongLi opened a new issue, #1310: URL: https://github.com/apache/paimon-rust/issues/1310
## Problem Following the Spark function integration in #1308, several SQL compatibility gaps remain in the DataFusion integration: - `TRUE IS DISTINCT FROM FALSE OR TRUE` can silently return the wrong result because sqlparser 0.62 binds the right-hand side too loosely. This also affects SQL reparsed for DML, functions and views. - Binary inputs, especially Arrow FixedSizeBinary and dictionary-encoded arrays, are not consistently accepted by byte-length and substring functions. - Flink-style `split_index(text, delimiter, index)` has no equivalent registration. Spark's `split` and one-based `split_part` should keep their existing semantics. - Spark's discrete `approx_percentile` / `percentile_approx` aggregate is missing. Aliasing it to DataFusion's interpolating `approx_percentile_cont` would give different answers and accuracy semantics. ## Expected behavior Fix the parser precedence and binary adapters; expose `split_index` as a separate extension; implement the Spark percentile contract, including foldable percentage/accuracy parameters, scalar and array results, partial aggregation, DISTINCT and supported windows. Add regression tests and document the supported semantics. Preserve application overrides and existing Spark function registrations. -- 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]
