[SOLR-14628] hl.fragsizeIsMinimum default ought to be 'true' for performance

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 8.5
• Fix Version/s: 8.6
• Component/s: highlighter
• Assignee: David Smiley
• Reporter: David Smiley
• Created: 05/Jul/20 05:52
• Updated: 15/Jul/20 15:21
• Resolved: 06/Jul/20 13:58

Description

hl.fragsizeIsMinimum was added in 8.5 with a default setting of false – thus producing average length fragments of fragsize instead of minimum beyond fragsize. In previous releases, even though the setting didn't exist, it was in-effect true (minimum). Unfortunately, this flip was found to be a significant performance regression for highlighting lots of text (especially when using the SENTENCE break iterator), and furthermore it seems to be of relatively minor value in fragment sizing.

I propose we flip this back to true.

https://lists.apache.org/thread.html/r2e44d290eac9edfa61ed9eced0d9f01a7bd1cbdf910344b2aa2368b8%40%3Csolr-user.lucene.apache.org%3E

Issue Links

• is caused by: LUCENE-9093 Unified highlighter with word separator never gives context to the left
• links to: GitHub Pull Request #1651