apache / apache/lucene

Remove divide-by-zero formula hacks across sim impls [LUCENE-8023]

Open
#9,071 0 comments 0 reactions 0 assignees View on GitHub
legacy-jira-priority:Major type:bug
Dominant language
Java
Stars
3.6k
Forks
1.4k
Avg merge
2d 11h
Merged PRs (30d)
88

Description

Follow-up to #9068

With the scoring api no longer passing e.g. docFreq=0 in corner cases such as SpanOrQuery, we can revert formula hacks that we added to many similarities. It may not be very significant for ranking but it removes confusion and makes it possible to have good explain() etc.

---
Migrated from [LUCENE-8023](https://issues.apache.org/jira/browse/LUCENE-8023) by Robert Muir (@rmuir)

Contributor guide

Open the contributing guide

Research direction

Start by reading follow-up issue #9068 and locating the similarity implementations containing the divide-by-zero formula workarounds. Compare the affected formulas with the updated scoring API assumptions, then remove the obsolete hacks across the implementations and verify that explanations and corner-case scoring remain correct.

Written by the indexing model from the issue text.

Assessment

Tech stack
java
Domain
search
Issue type
Refactor
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.