Normalized Cache

This normalized cache provides the following functionality:

Data (de)normalization
Data subscriptions
Data validation
Data invalidation
Data expiration
Computed fields
Optimistic updates
Garbage collection

The library is around 6 KB gzipped when using all features.

Setup

Installation:

npm install --save normalized-cache

Usage

import { Cache, schema } from "normalized-cache";

const Author = schema.object({
  name: "Author",
});

const Post = schema.object({
  name: "Post",
  fields: {
    author: Author,
  },
});

const cache = new Cache({
  types: [Post],
});

cache.write({
  type: "Post",
  data: {
    id: "1",
    title: "Title",
    author: {
      id: "2",
      name: "Name",
    },
  },
});

const result = cache.read({
  type: "Post",
  id: "1",
});

console.log(result.data);

// {
//   id: "1",
//   title: "Title",
//   author: {
//     id: "2",
//     name: "Name",
//   },
// }

const result2 = cache.read({
  type: "Author",
  id: "2",
});

console.log(result2.data);

// {
//   id: "2",
//   name: "Name",
// }

const result3 = cache.read({
  type: "Post",
  id: "1",
  select: cql`{ title author { name } }`,
});

console.log(result3.data);

// {
//   title: "Title",
//   author: {
//     name: "Name",
//   },
// }

API

class Cache {
  constructor(config?: CacheConfig): Cache;
  get(entityID: string, optimistic?: boolean): Entity | undefined;
  set(entityID: string, entity: Entity | undefined, optimistic?: boolean): Entity;
  identify(options: IdentifyOptions): string | undefined;
  read(options: ReadOptions): ReadResult | undefined;
  write(options: WriteOptions): WriteResult;
  delete(options: DeleteOptions): DeleteResult | undefined;
  invalidate(options: InvalidateOptions): InvalidateResult | undefined;
  watch(options: WatchOptions): Unsubscribable;
  silent(fn: () => void): void;
  transaction(fn: () => void): void;
  extract(optimistic?: boolean): SerializedData;
  restore(data: SerializedData): void;
  reset(): void;
  gc(): void;
  retain(entityID: string): Disposable;
  addOptimisticUpdate(updateFn: OptimisticUpdateFn): OptimisticUpdateDisposable;
  removeOptimisticUpdate(id: number): void;
}

interface CacheConfig {
  types?: ValueType[];
  onlyReadKnownFields?: boolean;
  onlyWriteKnownFields?: boolean;
}

const schema = {
  array(config?: ArrayTypeConfig | ValueType): ArrayType
  boolean(config?: BooleanTypeConfig): BooleanType
  nonNullable(config: NonNullableTypeConfig | ValueType): NonNullableType
  number(config?: NumberTypeConfig): NumberType
  object(config?: ObjectTypeConfig): ObjectType
  string(config?: StringTypeConfig | string): StringType
  union(config: UnionTypeConfig | ValueType[]): UnionType
}

Schema

Schema types allow you to define entities, relationships and fields.

Learn more about the type system here.

Writing

When writing to the cache, a type must be provided.

cache.write({
  type: "Post",
  data: { id: "1", title: "Title" },
});

A ID can be specified if this cannot be inferred from the data itself:

cache.write({
  type: "Post",
  id: "1",
  data: { title: "Title" },
});

If the ID is an object or array it will be automatically serialized to a stable string:

cache.write({
  type: "Posts",
  id: { page: 1, limit: 10 },
  data: [],
});

Reading

Reading from the cache can be done with the read method.

Without selector

When no selector is given, all data related to the entity will be returned:

cache.write({
  type: "Author",
  data: {
    id: "2",
    name: "Name",
  },
});

cache.write({
  type: "Post",
  data: {
    id: "1",
    title: "Title",
    author: {
      id: "2",
    },
  },
});

const result = cache.read({
  type: "Post",
  id: "1",
});

console.log(result.data);

// {
//   id: "1",
//   title: "Title",
//   author: {
//     id: "2",
//     name: "Name",
//   },
// }

The resulting data can contain circular references when entities refer to each other.

With selector

Selectors can be used to select specific fields:

import { cql } from "normalized-cache";

cache.write({
  type: "Author",
  data: {
    id: "2",
    name: "Name",
  },
});

cache.write({
  type: "Post",
  data: {
    id: "1",
    title: "Title",
    author: {
      id: "2",
    },
  },
});

const result = cache.read({
  type: "Post",
  id: "1",
  select: cql`{ title author { name } }`,
});

console.log(result.data);

// {
//   title: "Title",
//   author: {
//     name: "Name",
//   },
// }

Learn more about selectors here.

With write selector

The write method also returns a selector that matches the exact shape of the input:

cache.write({
  type: "Author",
  data: {
    id: "2",
    name: "Name",
  },
});

const { selector } = cache.write({
  type: "Post",
  data: {
    id: "1",
    title: "Title",
    author: {
      id: "2",
    },
  },
});

const result = cache.read({
  type: "Post",
  id: "1",
  select: selector,
});

console.log(result.data);

// {
//   id: "1",
//   title: "Title",
//   author: {
//     id: "2",
//   },
// }

Computed fields

Computed fields can be created by defining a field with a read function.

Defining a computed field for calculations:

const Cart = schema.object({
  name: "Cart",
  fields: {
    totalPrice: {
      read: (cart) => {
        return cart.items.reduce((total, item) => total + item.price, 0);
      },
    },
  },
});

Defining a relational field based on another field:

const Author = schema.object({
  name: "Author",
});

const Post = schema.object({
  name: "Post",
  fields: {
    author: {
      read: (post, { toReference }) => {
        return toReference({ type: "Author", id: post.authorId });
      },
    },
  },
});

Invalid fields

Fields that do not match with the schema will be reported in the invalidFields array:

const LoggedIn = schema.boolean({
  name: "LoggedIn",
});

const cache = new Cache({
  types: [LoggedIn],
});

cache.write({
  type: "LoggedIn",
  data: "string",
});

const result = cache.read({
  type: "LoggedIn",
});

if (result.invalidFields) {
  console.log("Invalid data");
}

Missing fields

Fields that are missing will be reported in the missingFields array:

const LoggedIn = schema.boolean({
  name: "LoggedIn",
});

const cache = new Cache({
  types: [LoggedIn],
});

const result = cache.read({
  type: "LoggedIn",
});

if (result.missingFields) {
  console.log("Missing data");
}

Stale flag

The stale flag indicates if some entity or field has been invalidated or if any expiresAt has past:

const LoggedIn = schema.boolean({
  name: "LoggedIn",
});

const cache = new Cache({
  types: [LoggedIn],
});

cache.write({
  type: "LoggedIn",
  data: true,
  expiresAt: 0,
});

const result = cache.read({
  type: "LoggedIn",
});

if (result.stale) {
  console.log("Stale data");
}

Watching

Data in the cache can be watched with the watch method.

Watching for any change in a specific post and all related data:

const { unsubscribe } = cache.watch({
  type: "Post",
  id: "1",
  callback: (result, prevResult) => {
    // log
  },
});

unsubscribe();

Watching specific fields:

cache.watch({
  type: "Post",
  id: "1",
  select: cql`{ title }`,
  callback: (result, prevResult) => {
    if (!prevResult.stale && result.stale) {
      // The title became stale
    }
  },
});

Invalidation

Entities and fields can be invalidated with the invalidate method.

When an entity or field is invalidated, all related watchers will be notified.

Invalidate an entity:

cache.invalidate({
  type: "Post",
  id: "1",
});

Invalidate entity fields:

cache.invalidate({
  type: "Post",
  id: "1",
  select: cql`{ comments }`,
});

Expiration

when expiresAt is specified, all affected fields will be considered stale after the given time:

cache.write({
  type: "Post",
  data: { id: "1" },
  expiresAt: Date.now() + 60 * 1000,
});

Set expiration for certain types:

cache.write({
  type: "Post",
  data: { id: "1" },
  expiresAt: {
    Comment: Date.now() + 60 * 1000,
  },
});

Deletion

Entities and fields can be deleted with the delete method.

Deleting an entity:

cache.delete({
  type: "Post",
  id: "1",
});

Deleting specific fields:

cache.delete({
  type: "Post",
  id: "1",
  select: cql`{ title }`,
});

Optimistic updates

An optimistic update function can be used to update the cache optimistically.

These functions will be executed everytime the cache is updated, until they are removed.

This means that if new data is written to the cache, the optimistic update will be re-applied / rebased on top of the new data.

async function addComment(postID, text) {
  function addCommentToPost(comment) {
    const result = cache.read({
      type: "Post",
      id: postID,
      select: cql`{ comments }`,
    });

    cache.write({
      type: "Post",
      id: postID,
      data: { comments: [...result.data.comments, comment] },
    });
  }

  const update = cache.addOptimisticUpdate(() => {
    addCommentToPost({ id: uuid(), text });
  });

  const comment = await api.addComment(postID, text);

  cache.transaction(() => {
    update.dispose();
    addCommentToPost(comment);
  });
}

Merging

By default entities are shallowly merged and non-entity values are replaced.

This behavior can be customized by defining custom write functions on entities and fields.

Replacing entities instead of merging:

const Author = schema.object({
  name: "Author",
  write: (incoming) => {
    return incoming;
  },
});

Merging objects instead of replacing:

const Post = schema.object({
  name: "Post",
  fields: {
    content: {
      type: schema.object(),
      write: (incoming, existing) => {
        return { ...existing, ...incoming };
      },
    },
  },
});

Transforming values when writing:

const Post = schema.object({
  name: "Post",
  fields: {
    title: {
      write: (incoming) => {
        if (typeof incoming === "string") {
          return incoming.toUpperCase();
        }
      },
    },
  },
});

Transactions

Multiple changes can be wrapped in a transaction to make sure watchers are only notified once after the last change:

cache.transaction(() => {
  cache.write({
    type: "Post",
    data: { id: "1", title: "1" },
  });
  cache.write({
    type: "Post",
    data: { id: "2", title: "2" },
  });
});

Silent changes

Wrap changes with silent to prevent watchers from being notified:

cache.silent(() => {
  cache.write({
    type: "Post",
    data: { id: "1", title: "1" },
  });
});

Garbage collection

The gc method can be used to remove all unwatched and unreachable entities from the cache.

Use the retain method to prevent an entity from being removed.

boschni/normalized-cache