wgtmac opened a new issue, #55448:
URL: https://github.com/apache/arrow/issues/55448
### Describe the enhancement requested
Track the remaining ALP improvements after the initial implementation in
#48345.
- [ ] Accept unsorted and duplicate exception positions while preserving the
optimized patch loop for strictly increasing positions.
- [ ] Serialize vectors one at a time instead of retaining all encoded
vectors
and a separate offset array for the page.
- [ ] Make ALP vector size configurable through writer properties and encoder
factories.
- [ ] Cache and reuse a column-chunk preset across pages.
- [ ] Encode complete vectors as values arrive in `Put()` / `PutSpaced()`,
retaining only the unfinished vector instead of buffering the entire
page
as raw values until `FlushValues()`.
- [ ] Support falling back from ALP when ALP would produce a larger page.
- [ ] Improve `EstimatedDataEncodedSize()` using observed encoded bytes per
value,
including metadata and exception overhead, instead of returning the raw
buffered byte count.
### Component(s)
C++, Parquet
--
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]
