Details
-
Bug
-
Resolution: Unresolved
-
Major - P3
-
None
-
None
-
None
-
None
-
Query Optimization
-
ALL
Description
Inserting infinity into a histogram causes the valueSpread to neighboring values to be NaN. This results in the area of that DataDistribution bucket to be NaN which causes issues when determining the topK buckets during the max-diff algorithm.
One potential solution is to introduce separate counters for infinity like we did for NaN in https://jira.mongodb.org/browse/SERVER-72807.