sequelize-simple-cache

sequelize-simple-cache

This is a simple, transparent, client-side, in-memory cache for Sequelize v4 and v5. Cache invalidation is based on time-to-live (ttl). Selectively add your Sequelize models to the cache. Works with all storage engines supported by Sequelize.

This cache might work for you if you have database tables that (1) are frequently read but very rarely written and (2) contain only few rows of data.

In a project, we had a couple of database tables with a sort of configuration. Something like 4 or 5 tables with some 10 rows of data. Nearly every request needed this data, i.e., it was read all the time. But updated only very rarely, e.g, once a day. So, pre-fetching or simple in-memory caching would work for us.

If that's not matching your scenario, better look for something more sophisticated such as Redis or Memcached.

Install

npm install sequelize-simple-cache

Usage

Setup the cache along with loading your Sequelize models like this:

const Sequelize = require('sequelize');
const SequelizeSimpleCache = require('sequelize-simple-cache');

const sequelize = new Sequelize('database', 'username', 'password', { ... });

// initialize cache
const cache = new SequelizeSimpleCache({
  User: { ttl: 5 * 60 }, // 5 minutes
  Page: { }, // default ttl is 1 hour
});

// add your models to the cache like this
const User = cache.init(sequelize.import('./models/user'));
const Page = cache.init(sequelize.import('./models/page'));

// no caching for this one (because it's not configured to be cached)
// will only add dummy decorators to the model for a homogeneous interface to all models
const Order = cache.init(sequelize.import('./models/order'));

// the Sequelize model API is fully transparent, no need to change anything.
// first time resolved from database, subsequent times from local cache.
const fred = User.findOne({ where: { username: 'fred' }});

More Details

Supported methods

The following methods on Sequelize model instances are supported for caching: findOne, findAndCountAll, findByPk, findAll, count, min, max, sum. In addition, for Sequelize v4: find, findAndCount, findById, findByPrimary, all.

Non-cacheable queries / bypass caching

You need to avoid non-cacheable queries, e.g., queries containing dynamic timestamps.

const { Op, fn } = require('sequelize');
// this is not good
Model.findAll({ where: { startDate: { [Op.lte]: new Date() }, } });
// you should do it this way
Model.findAll({ where: { startDate: { [Op.lte]: fn('NOW') }, } });
// if you don't want a query to be cached, you may explicitly bypass the cache like this
Model.noCache().findAll(...);

Time-to-live (ttl)

Each model has its individual time-to-live (ttl), i.e., all database requests on a model are cached for a particular number of seconds. Default is one hour. For eternal caching, i.e., no automatic cache invalidation, simply set the model's ttl to false (or any number less or equals 0).

const cache = new SequelizeSimpleCache({
  User: { ttl: 5 * 60 }, // 5 minutes
  Page: { }, // default ttl is 1 hour
  Foo: { ttl: false } // cache forever
});

Clear cache

There are these ways to clear the cache.

const cache = new SequelizeSimpleCache({...});
// clear all
cache.clear();
// clear all entries of specific models
cache.clear('User', 'Page');
// or do the same on any model
Model.clearCache(); // only model
Model.clearCacheAll(); // entire cache

By default, the model's cache is automatically cleared if these methods are called: update, create, upsert, destroy, findOrBuild. In addition, for Sequelize v4: insertOrUpdate, findOrInitialize, updateAttributes.

You can change this default behavior like this:

const cache = new SequelizeSimpleCache({
  User: { }, // default clearOnUpdate is true
  Page: { clearOnUpdate: false },
});

If you run multiple instances (clients or containers or PODs or alike), be aware that cache invalidation is more complex that the above simple approach.

Bypass caching

Caching can explicitly be bypassed like this:

Model.noCache().findOne(...);

Limit

This cache is meant as a simple in-memory cache for a very limited amount of data. So, you should be able to control the size of the cache.

const cache = new SequelizeSimpleCache({
  User: { }, // default limit is 50
  Page: { limit: 30 },
});

Logging

There is "debug" and "ops" logging -- both are off by default. Logging goes to console.debug() unless you set delegate to log somewhere else. event is one of: init, hit, miss, load, purge or ops.

const cache = new SequelizeSimpleCache({
  // ...
}, {
  debug: true,
  ops: 60, // seconds
  delegate: (event, details) => { ... },
});

Unit testing

If you are mocking your Sequelize models in unit tests with Sinon et al., caching might be somewhat counterproductive. So, either clear the cache as needed in your unit tests. For example (using mocha):

describe('My Test Suite', () => {
  beforeEach(() => {
    Model.clearCacheAll(); // on any model with the same effect
  });
  // ...

Or disable the cache right from the beginning. A quick idea... have a specific config value in your project's /config/default.js and /config/test.js to enable or disable the cache respectively. And start your unit tests with setting NODE_ENV=test before. This is actually the way I am doing it; plus a few extra unit tests for caching.

const config = require('config');
const useCache = config.get('database.cache');
// initializing the cache
const cache = useCache ? new SequelizeSimpleCache({...}) : undefined;
// loading the models
const model = sequelize.import('./models/model');
const Model = useCache ? cache.init(model) : model;