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 5 KB gzipped.
Installation:
npm install --save normalized-cacheimport { 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 { data } = cache.read({
type: "Post",
id: "1",
});
const { data } = cache.read({
type: "Author",
id: "2",
});class Cache {
get(entityID: string, optimistic?: boolean): Entity | undefined;
set(entity: Entity, optimistic?: boolean): Entity;
identify(options: IdentifyOptions): string | undefined;
read(options: ReadOptions): ReadResult;
write(options: WriteOptions): WriteResult;
delete(options: DeleteOptions): DeleteResult;
invalidate(options: InvalidateOptions): InvalidateResult;
watch(options: WatchOptions): UnsubscribeFn;
silent(fn: () => void): void;
transaction(fn: () => void): void;
reset(): void;
gc(): void;
retain(entityID: string): DisposeFn;
addOptimisticUpdate(updateFn: OptimisticUpdateFn): number;
removeOptimisticUpdate(id: number): void;
}
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 types allow you to define entities, relationships and fields.
Learn more about the type system here.
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 from the cache can be done with the read method.
When no selector is given, all fields and related entities will be returned:
const { data } = cache.read({
type: "Post",
id: "1",
});Selectors can be used to select specific fields to a certain depth:
import { cql } from "normalized-cache";
const { data } = cache.read({
type: "Post",
id: "1",
select: cql`{ title comments { text } }`,
});Use the star operator to select all fields on a certain level:
const { data } = cache.read({
type: "Post",
id: "1",
select: cql`{ * comments { text } }`,
});Quotes can be used to specify non-aplhanumeric fields:
const { data } = cache.read({
type: "Post",
id: "1",
select: cql`{ "field with spaces" { text } }`,
});Fields can also be aliased:
const { data } = cache.read({
type: "Post",
id: "1",
select: cql`{ myTitle: title } }`,
});Learn more about CQL here.
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.lineItems.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 });
},
},
},
});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 { invalidFields } = cache.read({ type: "LoggedIn" });
if (invalidFields) {
console.log("Invalid data");
}Fields that are missing will be reported in the missingFields array:
const LoggedIn = schema.boolean({ name: "LoggedIn" });
const cache = new Cache({ types: [LoggedIn] });
const { missingFields } = cache.read({ type: "LoggedIn" });
if (missingFields) {
console.log("Missing data");
}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 { stale } = cache.read({ type: "LoggedIn" });
if (stale) {
console.log("Stale data");
}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
}
},
});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 }`,
});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,
},
});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 }`,
});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 { data } = cache.read({
type: "Post",
id: postID,
select: cql`{ comments }`,
});
cache.write({
type: "Post",
id: postID,
data: { comments: [...data.comments, comment] },
});
}
const updateID = cache.addOptimisticUpdate(() => {
const optimisticComment = { id: uuid(), text };
addCommentToPost(optimisticComment);
});
const comment = await api.addComment(postID, text);
cache.transaction(() => {
cache.removeOptimisticUpdate(updateID);
addCommentToPost(comment);
});
}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();
}
},
},
},
});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" } });
});Wrap changes with silent to prevent watchers from being notified:
cache.silent(() => {
cache.write({ type: "Post", data: { id: "1", title: "1" } });
});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.