tisonkun opened a new pull request, #284: URL: https://github.com/apache/datasketches-rust/pull/284
REQ CDF queries lose precision by accumulating PMF values: a 70-item sketch can end at `0.9999999999999999`. KLL can also return a smaller retained item for a rank-1 query when a stream weight above `2^53` rounds down to `f64`. Compute CDF values directly from ranks and handle the upper quantile endpoint explicitly. - Add `estimated_size()` to KLL and REQ, counting inline state and owned buffer capacity. Document the exclusion of allocator overhead and allocations owned by generic items. - Simplify REQ query and compaction code, remove the unused compaction argument and misleading comments, and reserve the full merge scratch length correctly. - List KLL in the README, update the changelog, and cover query boundaries, buffer growth/reset, owned item allocations, and zero-sized items through the existing integration targets. Validation: - `cargo x prepare-testdata` (754 pinned C++/Go/Java snapshots) - `cargo x check` - `cargo x test` (709 tests passed, including serialization compatibility and doctests) - `cargo x lint` - `cargo +1.86.0 test --locked -p tests-integration --test req_test --test kll_test` (67 passed) - `cargo publish --dry-run --locked --all-features -p datasketches` The CDF and KLL endpoint regressions were reproduced before their fixes. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
