Hello Apache DataSketches PMC and Community,
This is a call for vote to release Apache DataSketches-java candidate
version 9.1.0-RC1
Changes from the previous release (9.0.0):
New features:
- Filters: new Xor filter
- Bloom filter: difference (A NOT B); invert() is deprecated
- HLL: HllUnion.update(HllUnion)
- CPC: CpcUnion.update(MemorySegment)
- Theta sketch: UpdatableThetaSketch.compactTrimmed() compacts and
reduces to k in one pass
- Tuple sketch: custom update seeds, with seed hash checks
Bug fixes and C++ compatibility:
- Stricter validation of serialized sketches: malformed or truncated
input is rejected with an exception (ArrayOfNumbersSerDe and the
string SerDe, Bloom filter, Count-Min, CPC, Frequencies, HLL, Items
Quantiles, REQ, Theta, Tuple, VarOpt)
- HLL: fixes to the deferred KxQ rebuild and to HIP after a merge
- Theta/Tuple: seed hash checks and empty sketch handling aligned
across the families and with C++; fixes for reading historical
images; JaccardSimilarity fixes
- t-digest: interpolation, min/max, extreme value and NaN fixes;
centroids sort by value, then by weight
- KLL items: sorted view fixed after query-before-serialize
- Frequent items: empty now means zero stream weight
- REQ: negative values on deserialization, compactor weight fixes,
getRSE() fixed
- CPC: CpcWrapper no longer throws on empty sketches
Compatibility notes:
- An empty compact tuple sketch now carries a zero seed hash, aligned
with Theta and with C++ and Python. Release 9.0.0 and earlier reject
an empty compact ArrayOfDoubles sketch serialized by this release,
so upgrade the readers before, or together with, the writers.
Non-empty ArrayOfDoubles images, and all other sketch images written
by this release, can be read by 9.0.0.
- Tuple sketches: nominal entries are now limited to 2^26.
Source repository:
- https://github.com/apache/datasketches-java
Git Tag for this release:
- https://github.com/apache/datasketches-java/releases/tag/9.1.0-RC1 on
branch 9.1.X
Git HashId for this release is: cc5eba818c260c0b9011791218819bd32a2d06ca
The Release Candidate Zip Repository:
- https://dist.apache.org/repos/dist/dev/datasketches/java/9.1.0-RC1/
The Maven Staging Repository:
-
https://repository.apache.org/content/groups/staging/org/apache/datasketches/datasketches-java/9.1.0/
The public signing key can be found in the KEYS file:
- https://dist.apache.org/repos/dist/dev/datasketches/KEYS
The artifacts have been signed with --keyid-format SHORT : 8CD4A902
Build & Test Guide:
- https://github.com/apache/datasketches-java/blob/9.1.0-RC1/README.md
Note: this release requires an OpenJDK-compatible build of Java 25.
The vote will be performed as follows:
- This letter will be published on dev@ and remain open for at least 72
hours (excluding weekends and holidays), AND until at least 3 (+1) PMC
votes or
a majority of (+1) PMC votes are acquired. Anyone in the community can vote.
This vote will close no earlier than Monday, Nov 9th 6PM PST (Nov 10th
02:00 UTC)
Please vote accordingly:
[ ] +1 approve
[ ] +0 no opinion
[ ] -1 disapprove with the reason
Thank you,
Lee Rhodes
[email protected]
