HyperLogLog

Use redis.HyperLogLog from the core Respire package on Redis 2.8.9+ to estimate the number of distinct values. A HyperLogLog does not retain an enumerable set of members; use sets when you need membership checks or exact counts.

await using var redis = await RespireClient.ConnectAsync("redis://localhost:6379");
bool changed = await redis.HyperLogLog.AddAsync("{visits}:monday", "ada", "grace", "ada");
long monday = await redis.HyperLogLog.CountAsync("{visits}:monday");
long week = await redis.HyperLogLog.CountAsync("{visits}:monday", "{visits}:tuesday");
bool merged = await redis.HyperLogLog.MergeAsync(
    "{visits}:week", "{visits}:monday", "{visits}:tuesday");

Method	Redis command	Result
AddAsync	PFADD	Whether the sketch changed; not an exact new-member test
CountAsync	PFCOUNT	Estimated distinct count across one or more sketches
MergeAsync	PFMERGE	true on an OK reply

Counting multiple keys estimates their union without creating a destination. MergeAsync stores the union in the destination. Counts remain estimates even when input values overlap; adding the same value repeatedly does not intentionally count it again.

Key prefixes apply to sketch keys, not the values added. All keys in a multi-key count or merge must share a Cluster slot after prefixing; the {visits} tag above ensures this. Span overloads accept a cancellation token when adding values, counting, or merging.

batch.HyperLogLog and transaction.HyperLogLog provide Add, Count, and Merge. Await execution before reading pending results. See batches and transactions.

HyperLogLog needs no module package. For Bloom filters, Cuckoo filters, Count-Min Sketch, Top-K, and t-digest, see probabilistic data structures.