Fix Uniform::new_inclusive overflow on large finite float ranges - #1821
Open
dhardy wants to merge 7 commits into
Open
Fix Uniform::new_inclusive overflow on large finite float ranges#1821dhardy wants to merge 7 commits into
dhardy wants to merge 7 commits into
Conversation
Dividing high - low by 1 - EPSILON can round to infinity even when the range itself is finite (e.g. 0.0..=f64::MAX), yielding a spurious NonFinite error while Uniform::new and sample_single_inclusive both accept the same range. Clamp infinite lanes to the largest finite value and let new_bounded reduce scale as usual, so that samples still cannot exceed high. Noted by dhardy in #1603.
1 task
There was a problem hiding this comment.
Pull request overview
This PR fixes Uniform::new_inclusive for floating-point ranges where (high - low) / (1 - ε) can round to infinity even though (high - low) is finite (e.g. 0.0..=f64::MAX), aligning behavior with sample_single_inclusive and avoiding spurious Error::NonFinite.
Changes:
- Adjust
UniformFloat::new_inclusiveto validate(high - low)for finiteness before scaling, allowing large finite inclusive ranges that previously overflowed during scaling. - Update the
new_boundedscale-reduction loop to use<=masking plusall()(instead of>plusany()), and add targeted tests fordecrease_masked(including infinity cases). - Add a
CHANGELOG.mdentry documenting the fix.
Reviewed changes
Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.
| File | Description |
|---|---|
src/distr/utils.rs |
Renames/adjusts float mask utilities and adds tests for decrease_masked, including infinity handling. |
src/distr/uniform_float.rs |
Fixes new_inclusive overflow-to-infinity rejection for large finite float ranges and extends test coverage for those bounds. |
CHANGELOG.md |
Documents the Uniform::new_inclusive large-range fix in Unreleased notes. |
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
Comment on lines
+60
to
+61
| /// Requirements: `low` and `high` must be finite. `scale` may be | ||
| /// infinite but must not be NaN. |
Comment on lines
+224
to
+226
| // Decrease all lanes where the mask is `true` to the next lower value | ||
| // representable by the floating-point type. At least one of the lanes | ||
| // must be set. | ||
| // must be set. Inputs may be non-finite but must not be NaN. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
CHANGELOG.mdentrySummary
Replaces #1809, allowing
Uniform::new_inclusiveto accept ranges like0.0..=f64::MAXwhich overflow-to-infinity, without extra code.Details
This allows
UniformFloat::new_inclusive(0.0, f64::MAX)to return a valid sampler, though in practice the largest value which may be sampled remains1.next_down() * MAX = MAX.next_down().This is slightly more consistent with
sample_single_inclusivewhich also accepts this range and also does not actually ever emit the largest value.