[ANN] MemoizationKit.jl: memoization with bounded memory, disk persistence, and a live dashboard

Package Announcements
,
1

MemoizationKit.jl v0.1.0 memoizes expensive functions and gives you tools to manage the caches: limit memory, keep results across runs, and see what they’re doing.

using MemoizationKit

@cached function fib(n::Int)::BigInt
    n < 2 ? BigInt(n) : fib(n - 1) + fib(n - 2)
end

fib(100)                   # compute and cache
cache_info(fib)            # size and hit/miss statistics
set_cache_size!(fib, 1000) # bound the cache

Features

  • Bounded storage: Clock and LRU eviction strategy, with limits in entries or bytes across all methods of a function.
  • Disk persistence: load SQLite.jl to keep results across runs, or ship precomputed results as artifacts.
    Lookups go RAM, then artifact, then local database, then computation.
  • Live dashboard: load Tachikoma.jl and call cache_dashboard() to browse hit rates, sizes, and activity for RAM and disk caches.
    You can also clear or resize caches from the terminal.
  • Fast RAM hits: return-type inference is preserved, and RAM hits don’t allocate for concrete keys with the built-in containers.
  • Strategies by function and argument type: shared, task-local, or no caching via CacheStyle.
  • Custom keys: MemoizationKit.cachekey lets equivalent inputs share a result, and Hashed customizes hashing and equality.
  • Configuration and profiling: Preferences set defaults per package or function, and TimerOutputs.jl integration profiles lookups and computations.

If a single in-memory cache is all you need, Memoize.jl or Memoization.jl with an LRU container may already be enough.
MemoizationKit targets workloads where the cache needs ongoing management.
The docs include a comparison.

Feedback, issues, and PRs are very welcome!

5 Likes