[LUCENE-10607] NRTSuggesterBuilder扩展input时溢出

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 9.2
• Fix Version/s: 9.3
• Component/s: core/FSTs
• Lucene Fields: New
• Assignee: Unassigned
• Reporter: chaseny
• Created: 09/Jun/22 10:40
• Updated: 28/Aug/22 16:41
• Resolved: 22/Jun/22 20:24

Description

suggest模块在创建索引时，调用NRTSuggestBuilder的finishTerm来写入suggest索引。

会调用maxNumArcsForDedupByte函数来扩展analyzed,向后扩展3 5 7 .... 255。

当entries长度过长（9000000）时，调用maxNumArcsForDedupByte扩展时

 

private static int maxNumArcsForDedupByte(int currentNumDedupBytes) {
int maxArcs = 1 + (2 * currentNumDedupBytes);
if (currentNumDedupBytes > 5)

{ maxArcs *= currentNumDedupBytes;  //当currentNumDedupBytes大于等于32768时，int相乘会大于int最大值 }

return Math.min(maxArcs, 255);
}

 

另外在扩展时，是否可以选择固定4字节来有序扩展。代替 3 5 7 ... 255的扩展方式

 

Issue Links

• links to: GitHub Pull Request #970