Limited-Cache

A minimal JS cache. Like using an object to store keys and values, except it won't grow forever

Motivation

A plain Javascript object is often good enough for simple key-value caching.

The problem is that a plain object cache can grow forever -- especially if you persist it in local storage. This library adds a size limit, plus maxCacheTime and smarter removal of old items.

Usage

The plain API provides a standard cache interface:

1const recentResults = LimitedCache({
2  maxCacheSize: 100,
3});
4recentResults.set('abc', thingToSave);
5recentResults.get('abc');
6recentResults.has('abc');
7recentResults.getAll();
8recentResults.reset();

Use LimitedCacheObject for a nicer developer experience, using Proxy:

1const recentResults = LimitedCacheObject();
2recentResults['abc'] = thingToSave;

Low-level functions using plain objects, if you need to stay serializable or want to store offline:

1const reduxState = {
2  childIdsByParentId: {},
3  cacheMeta: limitedCacheUtil.init(),
4};
5
6// cacheMeta is a plain, serializable object that contains the cache's internal state
7cacheMeta = limitedCacheUtil.set(cacheMeta, 'abc', thingToSave);
8
9return {
10  ...reduxState,
11  childIdsByParentId: limitedCacheUtil.getAll(cacheMeta),
12  cacheMeta,
13};

Typescript generics, to define a type for items in the cache:

1const stringCache = LimitedCache<string>();
2const myClassCache = LimitedCacheObject<SomeClass>();
3const offlineCacheMeta = lowLevelInit<SomeObjectShape>();

Note: The React hooks were removed in v1.0 The code for useLimitedCache and useLimitedCacheObject is here if you want to implement them yourself. For most cases, a useMemo(() => LimitedCache(), [])) should be enough.

Install and Import

npm install limited-cache or yarn add limited-cache

1import { LimitedCache, LimitedCacheObject, limitedCacheUtil } from 'limited-cache';

Options

maxCacheSize (number, default: 100)

Number of key/value pairs to keep in the cache. A falsy value will make it limitless.

maxCacheTime (milliseconds, default: 1 day, max: 1 year)

Time after which an item is removed. A falsy value will make it the 1-year maximum.

warnIfItemPurgedBeforeTime (milliseconds, default: 5000, development only)

If an item rotates out of the cache before this time passes, emits a warning to suggest you increase the cache size. Use a falsy value to disable.

Low-level functions

Under the hood, everything is tracked inside a single, serializable object (cacheMeta) which can be persisted to storage or kept in Redux or any other state.

You can retrieve this object from a LimitedCache or LimitedCacheObject, or create it directly via lowLevelInit:

1myLimitedCache.getCacheMeta();
2getCacheMetaFromObject(myLimitedCacheObject);

Do not manipulate cacheMeta directly: a set of low-level functions is available for that. Every action available on the higher-level LimitedCache and LimitedCacheObject is available as a low-level function.

• lowLevelInit(options)
• lowLevelGetOne(cacheMeta, cacheKey)
• lowLevelGetAll(cacheMeta) - returns the entire cache, excluding expired items
• lowLevelHas(cacheMeta, cacheKey)
• lowLevelSet(cacheMeta, cacheKey, value)
• lowLevelRemove(cacheMeta, cacheKey)
• lowLevelReset(cacheMeta)
• lowLevelSetOptions(cacheMeta, options) - you can update options anytime

These functions are also grouped together as limitedCacheUtil -- but minimization and tree-shaking will be slightly better if you import each individually.

FAQ

Immutable?

The cache itself is, but the low-level cacheMeta is persistent/mutated.

API for bulk operations?

Only reset and getAll. The other actions aren't as optimizable, so they're omitted to keep this small.

When are old items removed?

When new items are added, or if you try to get an item that has expired.

Is this a least-recently-used cache?

Not by default: For performance it only tracks by set time.

You can turn it into a least-recently-used cache by calling set each time you get an item, though: items will then expire based on when they were last accessed. This case has been optimized so performance won't suffer.