tisonkun commented on code in PR #196: URL: https://github.com/apache/datasketches-rust/pull/196#discussion_r3754843098
########## datasketches/src/frequencies/mod.rs: ########## Review Comment: Applied. The overview now states only the core requirement, while the Serialization section states only the requirement. ########## datasketches/src/hll/union.rs: ########## @@ -45,6 +45,12 @@ use crate::hll::mode::Mode; /// the union of all input sketches. It automatically handles sketches with /// different configurations and modes. /// +/// Coupon-mode inputs are replayed into the current gadget without reducing its `lg_k`. An +/// array-mode input with a smaller `lg_k` reduces the gadget to that value; larger array inputs are +/// downsampled to the gadget's current configuration. Once reduced, the effective `lg_k` remains +/// lower until [`reset`](Self::reset), so estimates and bounds reflect the reduced register count. +/// The requested [`HllType`] changes only the result representation, not its statistical accuracy. +/// Review Comment: They are internal representation concepts. I removed those terms from both the type documentation and the module overview, and kept the contract in terms of the observable effective and its effect on accuracy. ########## datasketches/src/hll/mod.rs: ########## @@ -42,24 +42,29 @@ //! //! # HLL Types //! -//! Three target HLL types are supported, trading precision for memory: +//! The three target HLL types are isomorphic representations of the same registers. Given the +//! same `lg_k` and input, they produce identical estimates and error distributions; they trade +//! memory layout and update performance, not statistical precision: //! -//! * [`HllType::Hll4`]: 4 bits per bucket (most compact) -//! * [`HllType::Hll6`]: 6 bits per bucket (balanced) -//! * [`HllType::Hll8`]: 8 bits per bucket (highest precision) +//! * [`HllType::Hll4`]: 4 bits per bucket plus an auxiliary table for rare exceptions (most +//! compact) +//! * [`HllType::Hll6`]: 6 bits per bucket (fixed-size middle ground) Review Comment: Applied. The Hll4 bullet now stays at the user-visible 4-bit/compact trade-off and no longer mentions the auxiliary table. -- 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]
