kv-orm is an Node.JS ORM for key-value datastores. It is currently in beta.
👤 Greg Brimble
- Github: @GregBrimble
- Personal Website: https://gregbrimble.com/
Contributions, issues and feature requests are welcome! Feel free to check issues page.
Please consider giving this project a ⭐️ if you use it, or if it provides some inspiration!
- In-Memory
- Cloudflare Workers KV
If there is any other datastore that you'd like to see supported, please create an issue, or make a pull request.
-
Support for multiple key-value datastores in a single application.
import { MemoryDatastore } from "@kv-orm/core"; const libraryDatastore = new MemoryDatastore(); const applicationSecrets = new MemoryDatastore();
See above for the full list of Supported Datastores.
-
Easy construction of typed Entities using Typescript.
import { Column, Entity } from "@kv-orm/core"; @Entity({ datastore: libraryDatastore }) class Author { @Column() public firstName: string; @Column() public lastName: string; // ... }
-
On-demand, asynchronous, lazy-loading: kv-orm won't load properties of an Entity until they're needed, and will do so seamlessly at the time of lookup.
import { getRepository } from "@kv-orm/core"; const authorRepository = getRepository(Author); let author = await authorRepository.load("[email protected]"); // 1ms - no properties of the author have been loaded console.log(await author.firstName); // 60ms - author.firstName is fetched
-
No unnecessary reads: if a property is already in memory, kv-orm won't look it up again unless it needs to.
let author = await authorRepository.load("[email protected]"); console.log(await author.lastName); // 60ms - author.lastName is fetched console.log(await author.lastName); // 1ms - author.lastName is retrieved from memory (no lookup performed)
-
Indexable and Unique Columns allowing quick lookups for specific entity instances:
import { UniqueColumn, IndexableColumn } from "@kv-orm/core"; @Entity({ datastore: libraryDatastore }) class Author { // ... @IndexableColumn() public birthYear: number; @UniqueColumn() public phoneNumber: string; // ... } let authors = await authorRepository.search("birthYear", 1564); // An AsyncGenerator yielding authors born in 1564 let author = await authorRepository.find("phoneNumber", "+1234567890"); // A single author with the phone number +1234567890
-
Relationships (*-to-one & *-to-many) with backref support, and on-update & on-delete cascading.
import { ToOne, ToMany } from "@kv-orm/core"; @Entity({ datastore: libraryDatastore }) class Author { // ... @ToOne({ type: Book, backRef: "author", cascade: true }) public books: Book[]; // ... } @Entity({ datastore: libraryDatastore }) class Book { @ToMany({ type: Author, backRef: "books", cascade: true }) public author: Author; // ... }
npm install --save @kv-orm/core
MemoryDatastore
is inbuilt into @kv-orm/core
. It is a simple in-memory key-value datastore, and can be used for prototyping applications.
import { MemoryDatastore } from `@kv-orm/core`;
const libraryDatastore = new MemoryDatastore();
See @kv-orm/cf-workers
for more information.
An Entity is an object which stores data about something e.g. an Author. The Entity decorator takes a datastore to save the Entity instances into.
Optionally, you can also pass in a key
to the decorator, to rename the key name in the datastore.
You can initialize a new instance of the Entity as normal.
import { Entity } from "@kv-orm/core";
@Entity({ datastore: libraryDatastore, key: "Author" })
class Author {
// ...
}
const authorInstance = new Author();
Using the @Column()
decorator on an Entity property is how you mark it as a savable property. You must await
their value. This is because it might need to asynchronously query the datastore, if it doesn't have the value in memory.
Like with Entities, you can optionally pass in a key
to the decorator.
import {
Column,
PrimaryColumn,
UniqueColumn,
IndexableColumn,
} from "@kv-orm/core";
import { Book } from "./Book";
@Entity({ datastore: libraryDatastore })
class Author {
@Column({ key: "givenName" })
public firstName: string;
@Column()
public lastName: string;
@Column()
public nickName: string | undefined;
@PrimaryColumn()
public emailAddress: string;
@IndexableColumn()
public birthYear: number;
@UniqueColumn()
public phoneNumber: string;
public someUnsavedProperty: any;
@ToMany({ type: () => Book, backRef: "author", cascade: true }) // More on this later
public books: Book[] = [];
public constructor({
firstName,
lastName,
emailAddress,
birthYear,
phoneNumber,
}: {
firstName: string;
lastName: string;
emailAddress: string;
birthYear: number;
phoneNumber: string;
}) {
this.firstName = firstName;
this.lastName = lastName;
this.emailAddress = emailAddress;
this.birthYear = birthYear;
this.phoneNumber = phoneNumber;
}
}
const williamShakespeare = new Author({
firstName: "William",
lastName: "Shakespeare",
emailAddress: "[email protected]",
birthYear: 1564,
phoneNumber: "+1234567890",
});
williamShakespeare.nickName = "Bill";
williamShakespeare.someUnsavedProperty = "Won't get saved!";
// When in an async function, you can fetch the value with `await`
(async () => {
console.log(await author.firstName);
})();
Any non-singleton class needs a PrimaryColumn used to differentiate Entity instances. For this reason, PrimaryColumn values are required and must be unique.
@Entity({ datastore: libraryDatastore })
class Author {
// ...
@PrimaryColumn()
public emailAddress: string;
// ...
}
An example of a singleton class where you do not need a PrimaryColumn, might be a global configuration Entity where you store application secrets (e.g. API keys).
An IndexableColumn can be used to mark a property as one which you may wish to later lookup with. For example, in SQL, you might perform the following query: SELECT * FROM Author WHERE birthYear = 1564
. In kv-orm, you can lookup Entity instances with a given IndexableColumn value with a repository's search method
@Entity({ datastore: libraryDatastore })
class Author {
// ...
@IndexableColumn()
public birthYear: number;
// ...
}
IndexableColumn types should be used to store non-unique values.
Columns with unique values can be setup with UniqueColumn. This is more efficient that an IndexableColumn, and the loading mechanism is simpler.
@Entity({ datastore: libraryDatastore })
class Author {
// ...
@UniqueColumn()
public phoneNumber: number;
// ...
}
If your property is particularly complex (i.e. can't be stored natively in the datastore), you may wish to use a property getter/setter for a Column, to allow you to serialize it before saving in the datastore.
For example, let's say you have a complex property on Author
, somethingComplex
:
@Entity({ datastore: libraryDatastore })
class Author {
// ...
@Column()
private _complex: string = ""; // place to store serialized value of somethingComplex
set somethingComplex(value: any) {
// function serialize(value: any): string
this._complex = serialize(value);
}
get somethingComplex(): any {
// function deserialize(serializedValue: string): any
return (async () => deserialize(await this._complex))();
}
// ...
}
To actually interact with the datastore, you'll need a Repository.
import { getRepository } from "@kv-orm/core";
const authorRepository = getRepository(Author);
You can then save Entity instances.
const williamShakespeare = new Author({
firstName: "William",
lastName: "Shakespeare",
emailAddress: "[email protected]",
birthYear: 1564,
phoneNumber: "+1234567890",
});
await authorRepository.save(williamShakepseare);
And subsequently, load them back again. If the Entity has a PrimaryColumn, you can load the specific instance by passing in the PrimaryColumn value.
const loadedWilliamShakespeare = await authorRepository.load(
"[email protected]"
);
console.log(await loadedWilliamShakespeare.nickName); // Bill
If a property has been set as an IndexableColumn (is non-unique), you can search for instances with a saved value.
const searchedAuthors = await authorRepository.search("birthYear", 1564);
for await (const searchedAuthor of searchedAuthors) {
console.log(await searchedAuthor.nickName); // Bill
}
If a property has been set as a UniqueColumn, you can directly load an instance by a saved value. If no results are found, null
is returned.
const foundWilliamShakespeare = await authorRepository.find(
"phoneNumber",
"+1234567890"
);
console.log(await foundWilliamShakespeare?.nickName); // Bill
const foundNonexistent = await authorRepository.find(
"phoneNumber",
"+9999999999"
);
console.log(foundNonexistent); // null
All Relationships must supply the type
of Entity as a function (to allow circular dependencies) and a backRef
(the property name on the inverse-side of the Relationship).
In order to propagate changes automatically on update and on delete, cascade
can be set to true
.
Note: deleting may be disproportionally intensive as it load, and then must edit or delete every related instance.
For * to one Relationships, use the ToOne decorator.
import { Author, authorRepository } from "./Author";
@Entity({ datastore: libraryDatastore })
class Book {
// ...
@ToOne({ type: () => Author, backRef: "books", cascade: true })
public author: Author;
// ...
}
const bookRepository = getRepository(Book);
const williamShakespeare = await authorRepository.load(
"[email protected]"
);
const hamlet = new Book({ title: "Hamlet", author: williamShakespeare });
console.log((await hamlet.author) === williamShakespeare); // true
// And because `cascade` was set to `true`, the Author instance has been updated as well
const books = await williamShakespeare.books;
for await (const book of books) {
console.log(await book.title); // Hamlet
}
await bookRepository.save(hamlet); // Will also save williamShakespeare
ToMany Relationships are slightly different to how Columns and ToOne Relationships work. Instead of setting its value and awaiting a Promise of its value, ToMany Relationships set an array of values, and await a Promise of an AsyncGenerator which yields its values.
@Entity({ datastore: libraryDatastore })
class Author {
// ...
@ToMany({ type: () => Book, backRef: "author", cascade: true }) // More on this later
public books: Book[] = [];
// ...
}
// ...
const hamlet = new Book({ title: "Hamlet" });
williamShakespeare.books = [hamlet];
const books = await williamShakespeare.books;
for await (const book of books) {
console.log(await book.title); // Hamlet
}
// And because `cascade` was set to `true`, the Book instance has been updated as well
console.log((await hamlet.author) === williamShakespeare); // true - because `cascade` is set to true
await authorRepository.save(williamShakespeare); // Will also save hamlet
Note: The order of returned Entities by the AsyncGenerator is not guaranteed. In fact, as items are loaded into memory, they will be pushed to the front to ensure tha other Entities are loaded only as needed.
To simplify interacting with ToMany
Relationships, some helpers are available.
Note: saving must still be completed afterwards to persist changes.
addTo
simplifies pushing an element to a ToMany relaionship:
import { addTo } from "@kv-orm/core";
addTo(williamShakespeare, "books", hamlet);
removeFrom
simplifies splicing an element from a ToMany relaionship:
import { removeFrom } from "@kv-orm/core";
removeFrom(williamShakespeare, "books", hamlet);
When defining an Entity in TypeScript, you must provide types for the properties. For example, firstName
and lastName
are set as string
s:
import { Column, Entity } from "@kv-orm/core";
@Entity({ datastore: libraryDatastore })
class Author {
@Column()
public firstName: string;
@Column()
public lastName: string;
// ...
}
However, when reading these values with kv-orm, in fact a Promise<string>
is returned. Simply switching these type definitions to Promise<T>
is not valid either, as writing values is done synchronously.
Therefore, to improve the developer experience and prevent TypeScript errors such as 'await' has no effect on the type of this expression. ts(80007)
and Property 'then' does not exist on type 'T'. ts(2339)
, when declaring the Entity, wrap the property types in the columnType
helper. In the same example:
import { columnType } from "@kv-orm/core";
@Entity({ datastore: libraryDatastore })
class Author {
@Column()
public firstName: columnType<string>;
@Column()
public lastName: columnType<string>;
// ...
}
columnType
is simply an alias: type columnType<T> = T | Promise<T>
.
Similarly, for ToOne Relationships:
import { toOneType } from "@kv-orm/core";
@Entity({ datastore: libraryDatastore })
class Book {
// ...
@ToOne({ type: () => Author, backRef: "books", cascade: true })
public author: toOneType<Author>;
// ...
}
And ToMany Relationships:
import { toManyType } from "@kv-orm/core";
@Entity({ datastore: libraryDatastore })
class Author {
// ...
@ToMany({ type: () => Book, backRef: "author", cascade: true })
public books: toManyType<Book>;
// ...
}
git clone [email protected]:kv-orm/core.git
npm install
npm run lint # 'npm run lint:fix' will automatically fix most problems
npm test
Copyright © 2019 Greg Brimble.
This project is MIT licensed.
If you're using a preprocessor that minifies class names, such as Babel, the class constructors names often get shortened. kv-orm will always use this class name, so, either disable minification in the preprocessor, or manually set the key
value when creating an Entity e.g.
@Entity({ key: "MyClass" })
class MyClass {
// ...
}
Thank you for trying out kv-orm in it's alpha period! Thanks to a generous sponsor, I have been able to complete work to elevate kv-orm to a more featureful beta. Unfortunately, this has meant a couple of minor breaking changes.
-
PrimaryColumn
,IndexableColumn
andUniqueColumn
have been introduced to deprecate theisPrimary
,isIndexable
andisUnique
options on theColumn
decorator. -
Repository
'sfind
method has been renamted tosearch
, and a different, new functionfind
been added.This is probably the most confusing and frustrating breaking change (apologies—I will take efforts to make sure this doesn't happen again). With the introduction of
UniqueColumn
as a more specific type ofIndexableColumn
, we needed a way to take advantage of its simpler loading mechanism. SinceUniqueColumn
has unique-values, there should be only ever one instance with a given value (or none), and sofind
seemed a more appropriate verb for its loading. WhereasIndexableColumn
can have non-unique values,search
seemed more appropriate a verb when returning anAsyncGenerator
of instances. -
ToOne
&ToMany
have been formally introduced (renamed from their draftedManyToOne
&ManyToMany
names—the parent's cardinality does not matter). -
Various bugfixes and improvements. None should have unexpected breaking changes, but if I've missed a use-case, please file a GitHub Issue and tell me about it.