Feature flags often combine explicit overrides, audience targeting, percentage rollouts, and a default. When rules overlap, their order is part of the configuration contract.
Design a FeatureFlagRuleEvaluator:
- The constructor receives aligned
ruleTypes, ruleValues, and decisions arrays. int matchedRule(String userId, String region, String plan) returns the first matching rule index, or -1.String evaluate(String userId, String region, String plan) returns that rule's "on" or "off" decision. No match defaults to "off".
Rules match as follows:
"USER": ruleValue == userId"REGION": ruleValue == region"PLAN": ruleValue == plan"PERCENT": the user's stable bucket is below the integer percentage in ruleValue"DEFAULT": always matches
For percentage rules, use the unsigned 32-bit polynomial hash from the sticky-bucketing exercise and bucket = hash % 100. Scan rules strictly in input order.
Example 1:
Input:
Output:
Explanation: Alice's explicit off rule wins before the India rule. Bob has no user override, so the India rule enables the flag.
Example 2:
Input:
Output:
Explanation: Eve hashes to bucket 20, while Frank hashes to bucket 50.
Constraints
0 <= ruleTypes.length == ruleValues.length == decisions.length <= 200- Rule types are
"USER", "REGION", "PLAN", "PERCENT", or "DEFAULT". - Percentage values are decimal integers from
0 through 100. - Decisions are
"on" or "off". - User, region, and plan strings contain printable ASCII characters.
- At most
500 total method calls are made.
Evaluate Ordered Feature-Flag Rulesmedium