advanced-caching 0.2.2b0

Production-ready composable caching with TTL, SWR, and background refresh patterns for Python.

advanced-caching

Production-ready caching library for Python with TTL, stale-while-revalidate (SWR), and background refresh.
Type-safe, fast, thread-safe, async-friendly, and framework-agnostic.

Issues & feature requests: new issue

---

Table of Contents

• Installation
• Quick Start
• Key Templates
• Storage Backends
  • InMemCache
  • RedisCache & Serializers
  • HybridCache (L1 + L2)
  • ChainCache (multi-level)
  • Custom Storage
• API Reference
• Testing & Benchmarks
• Use Cases
• Comparison
• Contributing
• License
• BGCache (Background)
  • Production example

---

Installation

uv pip install advanced-caching            # core
uv pip install "advanced-caching[redis]"  # Redis support
# pip works too

---

Quick Start

from advanced_caching import TTLCache, SWRCache, BGCache

# Sync function
@TTLCache.cached("user:{}", ttl=300)
def get_user(user_id: int) -> dict:
    return db.fetch(user_id)

# Async function (works natively)
@TTLCache.cached("user:{}", ttl=300)
async def get_user_async(user_id: int) -> dict:
    return await db.fetch(user_id)

# Stale-While-Revalidate (Sync)
@SWRCache.cached("product:{}", ttl=60, stale_ttl=30)
def get_product(product_id: int) -> dict:
    return api.fetch_product(product_id)

# Stale-While-Revalidate (Async)
@SWRCache.cached("async:product:{}", ttl=60, stale_ttl=30)
async def get_product_async(product_id: int) -> dict:
    return await api.fetch_product(product_id)

# Background refresh (Sync)
@BGCache.register_loader("inventory", interval_seconds=300)
def load_inventory() -> list[dict]:
    return warehouse_api.get_all_items()

# Background refresh (Async)
@BGCache.register_loader("inventory_async", interval_seconds=300)
async def load_inventory_async() -> list[dict]:
    return await warehouse_api.get_all_items()

# Configured Cache (Reusable Backend)
# Create a decorator pre-wired with a specific cache (e.g., Redis)
RedisTTL = TTLCache.configure(cache=RedisCache(redis_client))

@RedisTTL.cached("user:{}", ttl=300)
async def get_user_redis(user_id: int):
    return await db.fetch(user_id)

---

Key Templates

The library supports smart key generation that handles both positional and keyword arguments seamlessly.

• Positional Placeholder: "user:{}"

  • Uses the first argument, whether passed positionally or as a keyword.
  • Example: get_user(123) or get_user(user_id=123) -> "user:123"

• Named Placeholder: "user:{user_id}"

  • Resolves user_id from keyword arguments OR positional arguments (by inspecting the function signature).
  • Example: def get_user(user_id): ... called as get_user(123) -> "user:123"

• Custom Function:

  • For complex logic, pass a callable.
  • Example1 for kw/args with default values use : key=lambda *a, **k: f"user:{k.get('user_id', a[0])}"
  • Example2 fns with no defaults use : key=lambda user_id: f"user:{user_id}"

---

Storage Backends

• InMemCache (default): Fast, process-local
• RedisCache: Distributed in-memory
• HybridCache: L1 (memory) + L2 (Redis)
• ChainCache: Arbitrary multi-level chain (e.g., InMem -> Redis -> S3/GCS)
• S3Cache: Object storage backend (AWS)
• GCSCache: Object storage backend (Google Cloud)
• LocalFileCache: Filesystem-backed cache (per-host)

InMemCache

Thread-safe in-memory cache with TTL.

from advanced_caching import InMemCache

cache = InMemCache()
cache.set("key", "value", ttl=60)
cache.get("key")
cache.delete("key")
cache.exists("key")
cache.set_if_not_exists("key", "value", ttl=60)
cache.cleanup_expired()

---

RedisCache & Serializers

import redis
from advanced_caching import RedisCache, JsonSerializer

client = redis.Redis(host="localhost", port=6379)

cache = RedisCache(client, prefix="app:")
json_cache = RedisCache(client, prefix="app:json:", serializer="json")
custom_json = RedisCache(client, prefix="app:json2:", serializer=JsonSerializer())

---

HybridCache (L1 + L2)

Two-level cache:

• L1: In-memory
• L2: Redis

Simple setup

import redis
from advanced_caching import HybridCache, TTLCache

client = redis.Redis()
hybrid = HybridCache.from_redis(client, prefix="app:", l1_ttl=60)

@TTLCache.cached("user:{}", ttl=300, cache=hybrid)
def get_user(user_id: int):
    return {"id": user_id}

Manual wiring

from advanced_caching import HybridCache, InMemCache, RedisCache

l1 = InMemCache()
l2 = RedisCache(client, prefix="app:")
# l2_ttl defaults to l1_ttl * 2 if not specified
hybrid = HybridCache(l1_cache=l1, l2_cache=l2, l1_ttl=60)

# Explicit l2_ttl for longer L2 persistence
hybrid_long_l2 = HybridCache(l1_cache=l1, l2_cache=l2, l1_ttl=60, l2_ttl=3600)

TTL behavior:

• l1_ttl: How long data stays in fast L1 memory cache
• l2_ttl: How long data persists in L2 (Redis). Defaults to l1_ttl * 2
• When data expires from L1 but exists in L2, it's automatically repopulated to L1

With BGCache using lambda factory

For lazy initialization (e.g., deferred Redis connection):

from advanced_caching import BGCache, HybridCache, InMemCache, RedisCache

def get_redis_cache():
    """Lazy Redis connection factory."""
    import redis
    client = redis.Redis(host="localhost", port=6379)
    return RedisCache(client, prefix="app:")

@BGCache.register_loader(
    "config_map",
    interval_seconds=3600,
    run_immediately=True,
    cache=lambda: HybridCache(
        l1_cache=InMemCache(),
        l2_cache=get_redis_cache(),
        l1_ttl=3600,
        l2_ttl=86400  # L2 persists longer than L1
    )
)
def load_config_map() -> dict[str, dict]:
    return {"db": {"host": "localhost"}, "cache": {"ttl": 300}}

# Access nested data
db_host = load_config_map().get("db", {}).get("host")

---

ChainCache (multi-level)

from advanced_caching import InMemCache, RedisCache, S3Cache, ChainCache

chain = ChainCache([
    (InMemCache(), 60),                    # L1 fast
    (RedisCache(redis_client), 300),       # L2 distributed
    (S3Cache(bucket="my-cache"), 3600),   # L3 durable
])

# Write-through all levels (per-level TTL caps applied)
chain.set("user:123", {"name": "Ana"}, ttl=900)

# Read-through with promotion to faster levels
user = chain.get("user:123")

Notes:

• Provide per-level TTL caps in the tuple; if None, the passed ttl is used.
• set_if_not_exists delegates atomicity to the deepest level and backfills upper levels on success.
• get/get_entry promote hits upward for hotter reads.

---

Object Storage Backends (S3/GCS)

Store large cached objects cheaply in AWS S3 or Google Cloud Storage. Supports compression and metadata-based TTL checks to minimize costs.

📚 Full Documentation & Best Practices

from advanced_caching import S3Cache, GCSCache

user_cache = S3Cache(
    bucket="my-cache-bucket",
    prefix="users/",
    serializer="json",
    compress=True,
    dedupe_writes=True,  # optional: skip uploads when content unchanged (adds HEAD)
)

gcs_cache = GCSCache(
    bucket="my-cache-bucket",
    prefix="users/",
    serializer="json",
    compress=True,
    dedupe_writes=True,  # optional: skip uploads when content unchanged (adds metadata check)
)

RedisCache dedupe_writes

RedisCache(..., dedupe_writes=True) compares the serialized payload to the stored value; if unchanged, it skips rewriting and only refreshes TTL when provided.

LocalFileCache (filesystem)

from advanced_caching import LocalFileCache

cache = LocalFileCache("/var/tmp/ac-cache", dedupe_writes=True)
cache.set("user:123", {"name": "Ana"}, ttl=300)
user = cache.get("user:123")

Notes: one file per key; atomic writes; optional compression and dedupe to skip rewriting identical content.

---

BGCache (Background)

Single-writer/multi-reader pattern with background refresh and optional independent reader caches.

from advanced_caching import BGCache, InMemCache

# Writer: enforced single registration per key; refreshes cache on a schedule
@BGCache.register_writer(
    "daily_config",
    interval_seconds=300,   # refresh every 5 minutes
    ttl=None,               # defaults to interval*2
    run_immediately=True,
    cache=InMemCache(),     # or RedisCache / ChainCache
)
def load_config():
    return expensive_fetch()

# Readers: read-only; keep a local cache warm by pulling from the writer's cache
reader = BGCache.get_reader(
    "daily_config",
    interval_seconds=60,    # periodically pull from source cache into local cache
    ttl=None,               # local cache TTL defaults to interval*2
    run_immediately=True,
    cache=InMemCache(),     # local cache for this process
)

# Usage
cfg = reader()   # returns value from local cache; on miss pulls once from source cache

Notes:

• register_writer enforces one writer per key globally; raises if duplicate.
• interval_seconds <= 0 disables scheduling; wrapper still writes-on-demand on misses.
• run_immediately=True triggers an initial refresh if the cache is empty.
• get_reader creates a read-only accessor backed by its own cache; it pulls from the provided cache (usually the writer’s cache) and optionally keeps it warm on a schedule.
• Use cache= on readers to override the local cache backend (e.g., InMemCache in each process) while sourcing data from the writer’s cache backend.

See docs/bgcache.md for a production-grade example with Redis/ChainCache, error handling, and reader-local caches.

---

API Reference

• TTLCache.cached(key, ttl, cache=None)

• SWRCache.cached(key, ttl, stale_ttl=0, cache=None)

• BGCache.register_loader(key, interval_seconds, ttl=None, run_immediately=True)

• Storages:

  • InMemCache()
  • RedisCache(redis_client, prefix="", serializer="pickle"|"json"|custom)
  • HybridCache(l1_cache, l2_cache, l1_ttl=60, l2_ttl=None) - l2_ttl defaults to l1_ttl * 2

• Utilities:

  • CacheEntry
  • CacheStorage
  • validate_cache_storage()

---

Testing & Benchmarks

uv run pytest -q
uv run python tests/benchmark.py

---

Use Cases

• Web & API caching (FastAPI, Flask, Django)
• Database query caching
• SWR for upstream APIs
• Background refresh for configs & datasets
• Distributed caching with Redis
• Hybrid L1/L2 hot-path optimization

---

Comparison

Feature	advanced-caching	lru_cache	cachetools	Redis	Memcached
TTL	✅	❌	✅	✅	✅
SWR	✅	❌	❌	Manual	Manual
Background refresh	✅	❌	❌	Manual	Manual
Custom backends	✅ (InMem/Redis/S3/GCS/Chain)	❌	❌	N/A	N/A
Distributed	✅ (Redis, ChainCache)	❌	❌	✅	✅
Multi-level chain	✅ (ChainCache)	❌	❌	Manual	Manual
Dedupe writes	✅ (Redis/S3/GCS opt-in)	❌	❌	Manual	Manual
Async support	✅	❌	❌	✅	✅
Type hints	✅	✅	✅	❌	❌

---

Contributing

1. Fork the repo
2. Create a feature branch
3. Add tests
4. Run uv run pytest
5. Open a pull request

---

License

MIT License – see LICENSE.