SEZ9 commented on PR #12802:
URL: https://github.com/apache/seatunnel/pull/12802#issuecomment-6816724824

   Thanks @fatmanverse, and sorry for the slow follow-up.
   
   On the two items I can see in your status for 27340ef916:
   
   1. **Bounding `toPlainString()`** — the sink-side approach sounds right: 
computing `plainTextLength(value)` from precision and scale without building 
the text, and falling back to `BigDecimal#toString()` above Excel's 
32767-character cell limit, is the guard I was after. Covering `1E+2147483647` 
in `ExcelGeneratorTest#testDecimalWithLargeExponentIsWrittenAsScientificText` 
and pinning both sides of the limit is good. On the read side, I accept that 
`toDeclaredScale` checks the declared precision before rescaling and that the 
CSV/Text/JSON/XML readers parse DECIMAL the same way, so an Excel-specific 
rejection would be inconsistent. I'm fine deferring that; a one-line note in 
the PR description that the read side intentionally matches the other file 
formats would be enough to close it out.
   
   2. **EasyExcel `.xls`** — your comment is cut off in my view right after 
"`NumberRecordHandler` and `FormulaRecordHandler` also set 
`originalNumberValue`". If EasyExcel 4.0.3 populates `originalNumberValue` for 
the `.xls` path as well, then no change is needed and I'm happy to drop this 
finding. Could you confirm that is the conclusion, and ideally point to (or 
add) a test that reads a `.xls` fixture through `excel_engine = EasyExcel` and 
asserts the full-precision value, so the fast path is covered for both formats?
   
   For the remaining items (docs placement in the connector source/sink docs, a 
sink option to keep all-numeric output, declared scale on read, the 
upgrade-note wording, the `(BigDecimal)` cast, and the test comment), your 
per-item status did not reach me. I'll go back to 
https://github.com/apache/seatunnel/pull/12802#issuecomment-6557065213 and the 
current head, but if you could re-post the short status list for those (fixed / 
deferred / won't fix, with the test or doc location where applicable), that 
would let me finish the review in one pass. No new topics from my side.
   
   <!-- streview-comment:2277 -->


-- 
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]

Reply via email to