Use ULP float comparison instead of epsilon-based comparison
- Dominant language
- Java
- Stars
- 3.6k
- Forks
- 1.4k
- Avg merge
- 2d 11h
- Merged PRs (30d)
- 88
Description
### Description
#13723 introduced new methods for testing floating point equality based on number of units-of-least-precision rather than a fixed error. We should check which tests would benefit form using it. One [candidate](https://github.com/apache/lucene/pull/13723#pullrequestreview-2304393385) are the Panama vs scalar vector tests. The tricky bit will be to figure out the correct number of ULPs.
Contributor guide
Research direction
Start by reviewing the ULP-based test methods introduced in PR #13723, then inspect the Panama versus scalar vector tests mentioned in the issue. Identify epsilon-based floating-point comparisons that benefit from ULP comparisons and determine suitable ULP counts. Done means the appropriate tests use validated ULP thresholds and continue to pass.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- java
- Domain
- testing
- Issue type
- Refactor
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 35/100