SEPURI-SAI-KRISHNA commented on PR #11724: URL: https://github.com/apache/seatunnel/pull/11724#issuecomment-5367323569
@DanielLeens closing the loop on the follow-up I owed from your Issue Summary. **Issue 3 — `toBigDecimal` unification** is now up as its own PR: #11918. It does what was promised and nothing more. The two copies were byte-identical in behaviour, so this is a pure de-duplication: `NumericFunction.toBigDecimal` goes `private` → `public`, `ZetaSQLFunction`'s private copy is deleted, and the DECIMAL branch of `executeBinaryExpr` calls the surviving one. Net `+16/-32` across the two files, no behaviour change and no new tests, since the existing `SQLDecimalArithmeticTest` and `NumericFunctionTest` already cover both call paths and now exercise a single implementation. The merged Javadoc keeps the reasoning from both originals — why integral types go through `longValue()` rather than `doubleValue()`, and why `Float`/`Double` use `BigDecimal.valueOf` — and adds a line naming both callers, so the next person to touch it knows it is shared rather than local to `NumericFunction`. **Issues 1, 2, and 4** remain tracked and unchanged: the addition/subtraction scale question, the absence of anything bounding result precision (the near-ceiling test plus `MathContext`/explicit overflow check), and the divide-by-zero asymmetry where `DOUBLE` silently yields `Infinity`/`NaN` while `INT`/`BIGINT` throw a bare `ArithmeticException`. None is made worse by #11918. Separately, while working in `NumericFunction` for this I turned up a few defects that are outside both PRs and will get their own issues rather than being folded in here — integral overflow in the `ROUND`/`CEIL`/`FLOOR`/`TRUNC` family at negative scale, `ABS` at `Integer.MIN_VALUE`/`Long.MIN_VALUE`, and TINYINT/SMALLINT being declared supported by `ZetaSQLType` while the function bodies either throw or silently return the input unrounded. I will link them here once filed, since they sit next to the code you reviewed. Thanks again for the seven rounds on this one. -- 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]
