Handles Redis-compatible TDIGEST.* commands.

A t-digest is a probabilistic data structure for accurate on-line accumulation of rank-based statistics such as quantiles, trimmed means, and cumulative distribution values. It provides high accuracy at the tails (P99, P99.9) while using bounded memory.

Storage format

T-digests are stored as tagged tuples via the injected store map:

{:tdigest, centroids_list, metadata}

where centroids_list is a list of {mean, weight} tuples sorted by mean, and metadata is a map containing compression, count, min, max, buffer, and total_compressions.

Supported commands

• TDIGEST.CREATE key [COMPRESSION compression] -- create a new t-digest
• TDIGEST.ADD key value [value ...] -- add observations
• TDIGEST.RESET key -- clear data, preserve compression
• TDIGEST.QUANTILE key quantile [quantile ...] -- estimate values at quantiles
• TDIGEST.CDF key value [value ...] -- estimate CDF at values
• TDIGEST.RANK key value [value ...] -- estimate rank of values
• TDIGEST.REVRANK key value [value ...] -- estimate reverse rank
• TDIGEST.BYRANK key rank [rank ...] -- estimate value at rank
• TDIGEST.BYREVRANK key rank [rank ...] -- estimate value at reverse rank
• TDIGEST.TRIMMED_MEAN key low_quantile high_quantile -- trimmed mean
• TDIGEST.MIN key -- minimum observed value
• TDIGEST.MAX key -- maximum observed value
• TDIGEST.INFO key -- digest metadata
• TDIGEST.MERGE destkey numkeys src [src ...] [COMPRESSION c] [OVERRIDE]