Indeksd

A DSL for defining IndexedDB schemas.

Generate a type-safe IndexedDB client from a simple schema file.

Installation

npm i -D indeksd

Usage

Define a schema for your database.

database Blog {
  table Posts @item("Post") {
    @autoincrement id: number;
    @index title: string;
    content: string;
    @index author: string;
  }
}

This schema would create a database named "Blog" with a single object store names "Posts". The object store would have an autoincrement primary key "id" and two indexes "title" and "author" to search by.

Codegen can be performed programatically or via the command line.

Via command line:

inkeksd --sourceDir schemas --outDir codegen blog.db

Available options:

• --outDir: Where to save generated files to, relative to current directory. Will be created if it doesn't exist. Defaults to './codegen'.
• --sourceDir: The directory to search for raw schema files.
• --rootDir: Root directory for the project, used to resolve outDir and sourceDir if provided. Defaults to cwd.

Anything passed beyond these will be assumed to be source files. If a given source file can't be found relative to sourceDir an error will be thrown.

Using the Generated Code

The generated code uses no external dependencies. It just requires an environment with indexedDB defined on globalThis.

import {init} from './codegen/blog';

const db = await init();

const postId = await db.posts.add({
  title: 'First Post',
  content: 'This is the first post.',
  author: 'Kevin',
});

Because our "Posts" table was defined with an autoincrement "id" field the input type to the "add" method is defined as Omit<Post, 'id'>. We do get the postId as the result of the add operation.

Then when we want to get, we can use either the primary key, or one of our indexes.

const result1 = await db.posts.get(postId);
const result2 = await db.posts.get({id: postId});
const result3 = await db.posts.get({title: 'First Post'});
const result4 = await db.posts.get({author: 'Kevin'});

Again the argument types here are all generated so we couldn't try to perform a get with a field not defined as an index without getting a type error.

For more clarity the argument type for the get operation of "Posts" object store would be generated as:

export type PostsGetArgs = number | {
  id: number;
} | {
  title: string;
} | {
  author: string;
};

A type is also generated for the objects stored in our object store. By default the type name is the same as the table name. However, if you want to define a different name you can use the @item annotation to define a name. In our example we called the table "Posts", but wanted items in the table to use the singular "Post".

For our example this would be generated as:

export type Post = {
  id: number;
  title: string;
  content: string;
  author: string;
};

You can import this type from the generated file.

import {init, Post} from './codegen/blog';

Annotations

Annotations are used in the schema to further define the database we want.

Supported annotations:

• @version - Defined on the database to supply the version number.
• @item - Used on the table to define what name to use for a single item in a table. For example a "Posts" table could define each item as "Post".
• @index - Used on a field in a database to define that a given field should be an index. Takes an optional name argument that can be leveraged to define compound indexes.
• @autoincrement - Used on a field to say it should be autoincrement primary key.
• @key - Used on a field to define it as the primary key.
• @unique - Used on a field to declare that values of that field should be unique for each entry. This should be removed and added as an argument to the @index annotation.

Example Schema using above

database Blog @version(1) {
  table Posts @item("Post") {
    @autoincrement id: number;
    @index @index("title_author") @unique title: string;
    content: string;
    @index("title_author") author: string;
  }
}

This schema defines a Blog database with version 1. A table "Posts" where every entry in the table is a "Post". The primary key is the "id" field and it is an autoincrement field. Two indexes are defined, one "title" and one "title_author". The "title_author" index is a compound index allowing us to search by both the title and author fields. The title for each entry must also be unique.

Automated Joins

IndexedDB is a document store meaning that it stores objects and not the tables we find in relational databases where join operations are common. In document stores sometimes we store redundant data in the name of keeping things simple. However, sometimes it does make sense to eliminate the redundancy.

Extending from our Blog example... let's say we want to store more information about our authors than name and we want the association between Posts and Authors.

We could create a table for "Authors" and then store the primary key of the Author for each item in our "Posts" table. That would look something like this:

database Blob {
  table Authors {
    @autoincrement id: number;
    firstName: string;
    lastName: string;
  }

  table Posts {
    @autoincrement id: number;
    @index title: string;
    content: string;
    author: number;
  }
}

Our generated client can automate some of this (with the intention of supporting more automation). Let's update our schema to look like this:

database Blob {
  table Authors @item("Author") {
    @autoincrement id: number;
    firstName: string;
    lastName: string;
  }

  table Posts {
    @autoincrement id: number;
    @index title: string;
    content: string;
    author: Author;
  }
}

You'll notice I changed the type of author in the Posts table to be Author. I also added an item annotation to Authors. I could have made the type of author to be Authors, but the singular reads better. By defining the type of a field as another table I'm setting up a relationship between these two tables. When storing data a primary key to the Authors table will be stored in the author field in the Posts table. When performing a get the client will automatically fetch the author and merge the objects so the author field will actually hold the author object.

In the generated code there are two types created for Posts:

type Posts = {
  id: number;
  title: string;
  content: string;
  author: number;
};
type PostsWithJoins = {
  id: number;
  title: string;
  content: string;
  author: Author;
};

By default get operations will return the second object. You can disable this by setting withJoins to false. The get operations take an optional second argument that allows you to define some additional options for your query.

const postWithAuthor: PostsWithJoins = await db.posts.get({id: 1});
const postWithoutAuthor: Posts = await db.posts.get({id: 1}, {withJoins: false});

Automated joins also work for add and put operations. Working still from out Blog example we don't have to already have an Author in the Authors table... we can combine our insert of both a Post and an Author into one operation.

const postWithAuthor: PostsAddArgs = {
  title: 'First Post',
  content: 'Nothing special',
  author: {
    firstName: 'Kevin',
    lastName: 'Greene',
  }
}

const result: Posts = await db.posts.add(postWithAuthor);

Note the return type of the add operation. It is the Posts type. That means the author field on the returned object with contain the primary key of the Author that was added to the Authors table.

If there are nested joins used in a single add or put operation in order to get the primary key of objects added down the tree you would need to do those lookups yourself. For example, look up the added Author with the returned id and then look at the primary keys of anything inserted on that Author.

Fetching Multiple Documents

Currently there are two ways to fetch more than one item from the database.

1. Using sortBy to fetch all entries from database sorted by a given index. This can be limited to a given count if you have a large data set.
2. Using RangeQuery to fetch items in a given range. Range queries are exposed through the where method.

Sorted Get

The sortBy method takes as an argument an index in the given store and will return all entries sorted by that index.

To get all Posts sorted by title.

const posts: Array<PostsWithJoins> = await db.posts.sortBy('title');

The sortBy method takes an optional second parameter that will allow you to limit the number of items returned.

const posts: Array<PostsWithJoins> = await db.posts.sortBy('title', { count: 10 });

Range Queries

The ability to get multiple items from the database is exposed through range queries. Range queries are performed by calling the where method.

const results: Array<PostsWithJoins> = await db.posts.where('id').isGreaterThan(4);
const results: Array<PostsWithJoins> = await db.posts.where('id').isGreaterThanOrEqualTo(4);
const results: Array<PostsWithJoins> = await db.posts.where('id').isLessThan(4);
const results: Array<PostsWithJoins> = await db.posts.where('id').isLessThanOrEqualTo(4);
const results: Array<PostsWithJoins> = await db.posts.where('id').isBetween({from: 1, to: 7});
const results: Array<PostsWithJoins> = await db.posts.where('id').isEqualTo(4);

The where method takes an index (or key) from the database and returns an object that allows you to construct range queries on that index.

The supported methods are:

• isGreaterThan - Get all elements where given index is greater than provided argument.
• isGreaterThanOrEqualTo - Get all elements where given index is greater than or equal to the provided argument.
• isLessThan - Get all elements where given index is less than provided argument.
• isLessThanOrEqualTo - Get all elements where given index is less than or equal to the provided argument.
• isBetween - Get all elements where given index is between two arguments.
• isEqualTo - Get all elements where given index is equal to the given argument.

The returned methods are type safe based on the index name provided. For example, if the index you provide is of type string you can't pass a number.

Subscriptions

Often parts of your app will want to respond to database changes made in other parts of your app. To help facilitate this a subscribe method is exposed.

The subscribe method takes two arguments:

1. eventName - What kind of update are you interested in.
   • change - Subscribe to all changes to a table.
   • add - Subscribe when an object is added to a table.
   • put - Subscribe when an object is updated in the table.
   • delete - Subscribe when an object is deleted from the table.
2. callback - A function to call when the updates you are interested in happen. This callback receives a SubscriptionEvent. The SubscriptionEvent has the following type:

type SubscriptionEvent<ItemType, PrimaryKeyType> = {
  type: 'delete';
  data: PrimaryKeyType;
} | {
  type: 'add' | 'put';
  data: ItemType;
}

An example usage:

db.posts.subscribe('add', (e) => {
  const newListItem = renderPostListItem(e.data);
  postList.append(newListItem);
});

Client Operations

Calling the exported init function returns our database client. This client currently only supports a subset of IndexedDB features.

add

const addedPost = await db.posts.add({
  title: 'Test',
  content: 'Some content',
  author: 1,
})

put

const updatedPost = await db.posts.put({
  id: 4
  title: 'New Title',
  content: 'Some content',
  author: 1,
})

delete

await db.posts.delete({
  id: 4,
})

get

const postWithJoins = await db.posts.get({
  id: 4,
})

where

const postWithJoins = await db.posts.where('title').isEqualTo('New Title')

transaction

This is just a wrappter around the built-in db.transaction function that provides a bit of typesafety by ensuring you only ask for transactions for tables defined in your schema.