Enum SelectionStrategy

Specifies the algorithm used by ProbabilityList<T> to sample entries according to their weighted probabilities.

Namespace: Scylla.Core.Util.Random

Assembly: ScyllaCore.dll

Syntax

public enum SelectionStrategy

Remarks

All three strategies produce the same statistical distribution for the same set of probabilities; the choice is purely a performance trade-off:

Strategy	Build cost | Sample cost | Best use case
AliasTable	O(n) | O(1) | Large lists, infrequent modification
LinearScan	O(1) | O(n) | Small lists or frequent probability changes
BinarySearchCDF	O(n) | O(log n) | Medium lists, moderate pick frequency

When any entry has an IProbabilityInfluence provider attached, the alias table and CDF are bypassed regardless of the configured strategy, and a runtime linear scan over dynamically computed effective probabilities is used instead.

Fields

Name	Description
AliasTable	Walker's Alias Method. The internal table is built in O(n) time when entries change, and each sample is O(1) requiring exactly two uniform random values. This is the default strategy and is optimal for large lists where picks far outnumber modifications.
BinarySearchCDF	Binary search over a precomputed cumulative distribution function (CDF). The CDF is built in O(n) time and samples are O(log n), placing this strategy between LinearScan and AliasTable for medium-sized lists.
LinearScan	Linear scan through accumulated probabilities. Requires no precomputed structure (O(1) build cost) but samples in O(n) time. Best suited for small lists (typically fewer than eight entries) or for lists whose entries change frequently between picks, where the alias table rebuild cost would dominate.