IndexCandidateOptions
class IndexCandidateOptions(val minObservations: Long = 16, val minPresence: Double = 0.5, val minTypeStability: Double = 0.9, val minDistinct: Long = 2, val minColumnByteShare: Double = 0.05, val limit: Int = 16)
The thresholds SchemaCatalog.indexCandidates applies.
Every one of them is a heuristic and every one is here rather than buried in the code, because a recommendation engine whose rules cannot be seen or changed is one that gets ignored the first time it is wrong about somebody's workload.
Constructors
Properties
Link copied to clipboard
The share of the corpus's bytes a path must carry before a shredded column is suggested.
Link copied to clipboard
Distinct values a path needs to be worth an inverted index.
Link copied to clipboard
Observations a path needs before it is considered at all.
Link copied to clipboard
How much of the corpus a path must appear in. An index over a rare path rarely pays.
Link copied to clipboard
How dominant the path's type must be.