Hello all,
The Apache DataSketches team is pleased to announce the release of
Apache DataSketches Go 0.3.0.
Apache DataSketches is a library of stochastic streaming algorithms
("sketches") that answer
approximate queries over large data, such as distinct counts,
quantiles and frequent items,
in a single pass with bounded memory and proven error bounds.
DataSketches Go is the Go implementation and supports cross-language
serialization compatibility for supported sketches.
Highlights:
* REQ sketch serialization, compatible with Java and C++
* Theta: CompactTrimmed
* Count-Min: NewCountMinSketchFromSlice
* Hardened deserialization: corrupt input now returns an error instead
of panicking or over-allocating (HLL, CPC, Theta, Count-Min,
Bloom filter, common SerDe)
* Correctness fixes in t-digest, REQ and HLL
Breaking changes:
* frequencies: GetFrequentItems returns []Row instead of []*Row
* req: Sketch.All() returns iter.Seq[Item] instead of []Item
* tuple: GenerateHashKeyFromStrings now hashes keys as UTF-16LE to
match Java. Array-of-strings sketches built with 0.2.0 must be
rebuilt before combining them with 0.3.0 sketches.
Release notes: https://github.com/apache/datasketches-go/releases/tag/v0.3.0
Go module: go get github.com/apache/[email protected]
API docs: https://pkg.go.dev/github.com/apache/[email protected]
Thank you to the PMC members, the community for reviewing this
release and everyone who contributed to this release.
Best regards,
Hyeonho Kim