srielau commented on PR #58549: URL: https://github.com/apache/spark/pull/58549#issuecomment-5912608293
Addressed the 2026-09-29 P2s in c332e878b89: 1. JVM `UserDefinedPythonFunction.builder` now checks return type and buffer schema independently, so a valid STRING result with a CHAR/VARCHAR buffer reports `Python UDAF buffer schemas` and the buffer catalog string. 2. `CHAR_VARCHAR_NOT_SUPPORTED_IN_PYTHON` now says to use STRING in the declaration (not SQL CAST). Same class, Classic and Connect. 3. Direct regressions: JVM builder (return, buffer, UDT storage, UDTF), Classic/Connect Aggregator `bufferSchema`/`outputType`, Classic parsed UDTF DDL, Connect server-parsed UDTF DDL. Scope for the review-level question: those routes are in this PR. JVM recursion now unwraps `UserDefinedType` storage (`hasCharVarcharIncludingUDT`), matching Python `_has_type`. Analyze-derived UDTF schemas are rejected after the analyze worker returns the schema, with the same error class as a fixed return type. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
