Nemo's Parental Misguidance

    TypeScript icon, indicating that this package has built-in type declarations

    1.1.2 • Public • Published

    Actions Status Actions Status Actions Status Test Coverage Maintainability Packages npm version

    A simple semaphore/cache lib using redis, extended from remembered package

    How to Install

    npm i @remembered/redis

    How to use

    Create an Remembered instance passing a redis-io instance:

    const remembered = new RememberedRedis({
      ttl: 200, // In milliseconds
      redisTtl: 10000 // In seconds
    }, IORedisInstance);

    Now, call the method you want to use the semaphore using get

    const result = await remembered.get(request.myKey, () => myRequest(request));

    You can also use the wrap method to create a new version of your function that uses the semaphore under the hood

    const semaphoredMyRequest = remembered.wrap(myRequest, (request) => request.myKey);
    const result = await semaphoredMyRequest(request);

    Notice that the remembering key in the wrap option is defined by the second callback. This function receives the same parameters as the first one.


    The ttl refers to the local ttl, the one that the remembered package uses to reuse promises. redisTtl, in the other hand, refers to the ttl the cache will have in Redis. This distinction is important as, locally, you have a limited memory. You can make the local ttl be just some seconds or even 0 (for the promise to be reused just while it is not resolved), while RedisTtl can be larger, without affecting memory consumption in your service.

    How to use alternative persistence

    Sometimes the size of your data is just too big for redis to be a cheap solution, and you need to take some common strategy, like, to save the data in S3 and to use a reference to the saved file in Redis, for ttl control. This package offers a way to apply this strategy seamlessly through the configuration alternativePersistence. Here's an example:

    First, you need to implement the interface AlternativePersistence. The implementation below is a valid one to use with S3

    export class S3Cache implements AlternativePersistence {
        private s3: S3, //AWS S3 object
        private bucketName: string,
        public maxSavingDelay: number,
      ) {}
      async save(key: string, content: string | Buffer): Promise<void> {
        const objectKey = this.getObjectKey(key);
        const params = {
          Body: content,
          Bucket: this.bucketName,
          Key: objectKey,
          ContentType: 'application/octet-stream',
        await this.s3.putObject(params).promise();
      async get(key: string): Promise<string | Buffer | undefined> {
        try {
          const params = {
            Bucket: this.bucketName,
            Key: this.getObjectKey(key),
          const { Body } = await this.s3.getObject(params).promise();
          return Body as string | Buffer | undefined;
        } catch {
          return undefined;
      private getObjectKey(key: string): string {
        return `my-cache/${key}.bin`;

    Then, you can use it with remembered-redis:

    const remembered = new RememberedRedis(
        ttl: 0,
        redisTtl: 3600,
        redisPrefix: 's3-cache',
        alternativePersistence: new S3Cache(s3, 'my-bucket', 3),

    That's it! Now S3 will get the heavy data while Redis control semaphore, and ttl! But there's more to it: maxSavingDelay is part of the AlternativePersistence interface, and it determines how much time remembered-redis will wait to save the new data and to release the semaphore. But why wait? Because more data can be gathered from another calls and all that data can be saved in the same S3 file. The data is also saved compressed and it will benefit from a larger content, and the number of objects sent to S3 will be much lesser than the number of results you have. For example, if you have 100 calls in 3 seconds, you'll only send 1 file to S3, avoiding 99 calls! This is important because S3 can charge you very high if your number of api calls is big, and with this strategy you can decrease your cost by almost 50%!

    You can also use even a redis AlternativePersistence implementation:

    export class RedisCache implements AlternativePersistence {
        private redis: Redis,
        private bucketName: string,
        public maxSavingDelay: number,
      ) {}
      async save(key: string, content: string | Buffer, ttl: number): Promise<void> {
        await this.redis.setex(key, ttl, content);
      async get(key: string): Promise<string | Buffer | undefined> {
        return this.redis.getBuffer(key);

    This may not seem reasonable, but, as you join many results into one can when maxSavingDelay > 0, you can favor the compression with this, and save a lot of memory in your instance.


    Licensed under MIT.




    npm i @remembered/redis

    DownloadsWeekly Downloads






    Unpacked Size

    40.3 kB

    Total Files


    Last publish


    • farenheith
    • danielgaleni