Skip to content

chore(tdigest): better handle Centroid adds #88

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
tisonkun merged 1 commit into from
Feb 11, 2026
Merged

chore(tdigest): better handle Centroid adds #88

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
19 changes: 14 additions & 5 deletions datasketches/src/tdigest/sketch.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1251,15 +1251,24 @@ impl Centroid {
fn add(&mut self, other: Centroid) {
let (self_weight, other_weight) = (self.weight(), other.weight());
let total_weight = self_weight + other_weight;
self.weight = self.weight.saturating_add(other.weight.get());
self.weight = self
.weight
.checked_add(other.weight.get())
.expect("weight overflow");
Comment on lines +1254 to +1257

@tisonkun tisonkun Feb 10, 2026

Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Instead of silently saturating_add, panic if overflow.

This should never happen with real inputs, and let it crash should be better than silently holding inaccurate values.

Comment thread
tisonkun marked this conversation as resolved.

let (self_mean, other_mean) = (self.mean, other.mean);
let ratio_self = self_weight / total_weight;
let ratio_other = other_weight / total_weight;
self.mean = self_mean.mul_add(ratio_self, other_mean * ratio_other);
let delta = other_mean - self_mean;
self.mean = if delta.is_finite() {
delta.mul_add(ratio_other, self_mean)
} else {
let ratio_self = self_weight / total_weight;
self_mean.mul_add(ratio_self, other_mean * ratio_other)
};
Comment on lines +1261 to +1267

@tisonkun tisonkun Feb 10, 2026

Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If other_mean - self_mean doesn't overflow/underflow, save one mul + one div to reduce precision loss.

This can overflow/underflow, so we fallback to conservative weighted average computed in that case. In both cases, use mul_add to best effort reduce precision loss.

Comment thread
tisonkun marked this conversation as resolved.

debug_assert!(
!self.mean.is_nan(),
"NaN values should never be present in centroids; self: {}, other: {}",
self.mean.is_finite(),
"Centroid's mean must be finite; self: {}, other: {}",
self_mean,
other_mean
);
Expand Down