Skip to content

Latest commit

 

History

History
714 lines (518 loc) · 19.3 KB

README.md

File metadata and controls

714 lines (518 loc) · 19.3 KB

IndexedDB with usability and remote syncing

This is a fork of the awesome idb library, which adds the ability to sync an IndexedDB database with a remote REST API.

Video of two clients syncing their IndexedDB database

The source code for the example above can be found here.

Bundle size: ~3.48 kB brotli'd

Based on [email protected] (Jun 2022): e04104a...HEAD

Table of content

Features

All the usability improvements from the idb library

Since it is a fork of the idb library, synceddb shares the same Promise-based API:

import { openDB, SyncManager } from 'synceddb';

const db = await openDB('my-awesome-database');

const transaction = db.transaction('items', 'readwrite');
await transaction.store.add({ id: 1, label: 'Dagger' });

// short version
await db.add('items', { id: 1, label: 'Dagger' });

Async iterators are supported too (please notice the specific import):

import { openDB, SyncManager } from 'synceddb/with-async-ittr';

const tx = db.transaction('items');

for await (const cursor of tx.store) {
  // ...
}

More information here.

Sync with a remote REST API

Every change is tracked in a store. The SyncManager then sync these changes with the remote REST API when the connection is available, making it easier to build offline-first applications.

import { openDB, SyncManager } from 'synceddb';

const db = await openDB('my-awesome-database');
const manager = new SyncManager(db, 'https://example.com');

manager.start();

// will result in the following HTTP request: POST /items
await db.add('items', { id: 1, label: 'Dagger' });

// will result in the following HTTP request: DELETE /items/2
await db.delete('items', 2);

See also: Expectations for the REST API

Auto-reloading queries

The LiveQuery provides a way to run a query every time the underlying stores are updated:

import { openDB, LiveQuery } from 'synceddb';

const db = await openDB('my awesome database');

let result;

const query = new LiveQuery(['items'], async () => {
  // result will be updated every time the 'items' store is modified
  result = await db.getAll('items');
});

// trigger the liveQuery
await db.put('items', { id: 2, label: 'Long sword' });

// or manually run it
await query.run();

Inspired from Dexie.js liveQuery.

Computed stores

A computed store is a bit like a materialized view in PostgreSQL, it is updated every time one of the source object stores is updated:

import { openDB, createComputedStore } from 'synceddb/with-async-ittr';

const db = await openDB('my awesome database');

await createComputedStore(db, 'invoices-with-customer', 'invoices', ['customers'], async (tx, change) => {
  const computedStore = tx.objectStore('invoices-with-customer');

  if (change.storeName === 'invoices') {
    if (change.operation === 'add' || change.operation === 'put') {
      const invoice = change.value;
      // fetch the customer object
      invoice.customer = await tx.objectStore('customers').get(invoice.customerId);
      // update the computed store
      computedStore.put(invoice);
    } else { // change.operation === 'delete'
      computedStore.delete(change.key);
    }
  }

  if (change.storeName === 'customers') {
    if (change.operation === 'put') {
      const customer = change.value;
      // update all invoices with the given customer in the computed store
      for await (const cursor of computedStore.index('by-customer').iterate(change.key)) {
        const invoice = cursor.value;
        if (invoice.customerId === customer.id) {
          invoice.customer = customer;
          cursor.update(invoice);
        }
      }
    }
  }
});

This feature is great when you need to:

  • apply some filters on an aggregation of multiple object stores (no need to compute the JOIN in memory, and then apply the filters)
  • compute statistics over historical data (as the data is computed incrementally)

Disclaimer

  • no version history

Only the last version of each entity is kept on the client side.

  • basic conflict management

The last write wins (though you can customize the behavior in the onpusherror handler).

Installation

npm install synceddb

Then:

import { openDB, SyncManager, LiveQuery } from 'synceddb';

async function doDatabaseStuff() {
  const db = await openDB('my awesome database');

  // sync your database with a remote server
  const manager = new SyncManager(db, 'https://example.com');

  manager.start();
  
  // create an auto-reloading query
  let result;
  const query = new LiveQuery(['items'], async () => {
    // result will be updated every time the 'items' store is modified
    result = await db.getAll('items');
  });
}

API

For database-related operations, please see the idb documentation.

SyncManager

import { openDB, SyncManager } from 'synceddb';

const db = await openDB('my-awesome-database');
const manager = new SyncManager(db, 'https://example.com');

manager.start();

Options

fetchOptions

Additional options for all HTTP requests.

import { openDB, SyncManager } from 'synceddb';

const db = await openDB('my-awesome-database');
const manager = new SyncManager(db, 'https://example.com', {
  fetchOptions: {
    headers: {
      'accept': 'application/json'
    },
    credentials: 'include'
  }
});

manager.start();

Reference: https://developer.mozilla.org/en-US/docs/Web/API/fetch

fetchInterval

The number of ms between two fetch requests for a given store.

Default value: 30000

import { openDB, SyncManager } from 'synceddb';

const db = await openDB('my-awesome-database');
const manager = new SyncManager(db, 'https://example.com', {
  fetchInterval: 10000
});

manager.start();
buildPath

A function that allows to override the request path for a given request.

import { openDB, SyncManager } from 'synceddb';

const db = await openDB('my-awesome-database');
const manager = new SyncManager(db, 'https://example.com', {
  buildPath: (operation, storeName, key) => {
    if (storeName === 'my-local-store') {
      if (key) {
        return `/the-remote-store/${key[1]}`;
      } else {
        return '/the-remote-store/';
      }
    }
    // defaults to `/${storeName}/${key}`
  }
});

manager.start();
buildFetchParams

A function that allows to override the query params of the fetch requests.

Defaults to ?sort=updated_at:asc&size=100&after=2000-01-01T00:00:00.000Z,123.

import { openDB, SyncManager } from 'synceddb';

const db = await openDB('my-awesome-database');
const manager = new SyncManager(db, 'https://example.com', {
  buildFetchParams: (storeName, offset) => {
    const searchParams = new URLSearchParams({
      sort: '+updatedAt',
      size: '10',
    });
    if (offset) {
      searchParams.append('after', `${offset.updatedAt}+${offset.id}`);
    }
    return searchParams;
  }
});

manager.start();
updatedAtAttribute

The name of the attribute that indicates the last updated date of the entity.

Default value: updatedAt

import { openDB, SyncManager } from 'synceddb';

const db = await openDB('my-awesome-database');
const manager = new SyncManager(db, 'https://example.com', {
  updatedAtAttribute: 'lastUpdateDate'
});

manager.start();
withoutKeyPath

List entities from object stores without keyPath.

import { openDB, SyncManager } from 'synceddb';

const db = await openDB('my-awesome-database');
const manager = new SyncManager(db, 'https://example.com', {
  withoutKeyPath: {
    common: [
      'user',
      'settings'
    ]
  },
  buildPath: (_operation, storeName, key) => {
    if (storeName === 'common') {
      if (key === 'user') {
        return '/me';
      } else if (key === 'settings') {
        return '/settings';
      }
    }
  }
});

manager.start();

await db.put('common', { firstName: 'john' }, 'user');

Methods

start()

Starts the sync process with the remote server.

import { openDB, SyncManager } from 'synceddb';

const db = await openDB('my-awesome-database');
const manager = new SyncManager(db, 'https://example.com');

manager.start();
stop()

Stops the sync process.

import { openDB, SyncManager } from 'synceddb';

const db = await openDB('my-awesome-database');
const manager = new SyncManager(db, 'https://example.com');

manager.stop();
clear()

Clears the local stores.

import { openDB, SyncManager } from 'synceddb';

const db = await openDB('my-awesome-database');
const manager = new SyncManager(db, 'https://example.com');

manager.clear();
hasLocalChanges()

Returns whether a given entity currently has local changes that are not synced yet.

import { openDB, SyncManager } from 'synceddb';

const db = await openDB('my-awesome-database');
const manager = new SyncManager(db, 'https://example.com');

await db.put('items', { id: 1 });

const hasLocalChanges = await manager.hasLocalChanges('items', 1); // true
onfetchsuccess

Called after some entities are successfully fetched from the remote server.

import { openDB, SyncManager } from 'synceddb';

const db = await openDB('my-awesome-database');
const manager = new SyncManager(db, 'https://example.com');

manager.onfetchsuccess = (storeName, entities, hasMore) => {
  // ...
}
onfetcherror

Called when something goes wrong when fetching the changes from the remote server.

import { openDB, SyncManager } from 'synceddb';

const db = await openDB('my-awesome-database');
const manager = new SyncManager(db, 'https://example.com');

manager.onfetcherror = (err) => {
  // ...
}
onpushsuccess

Called after a change is successfully pushed to the remote server.

import { openDB, SyncManager } from 'synceddb';

const db = await openDB('my-awesome-database');
const manager = new SyncManager(db, 'https://example.com');

manager.onpushsuccess = ({ operation, storeName, key, value }) => {
  // ...
}
onpusherror

Called when something goes wrong when pushing a change to the remote server.

import { openDB, SyncManager } from 'synceddb';

const db = await openDB('my-awesome-database');
const manager = new SyncManager(db, 'https://example.com');

manager.onpusherror = (change, response, retryAfter, discardLocalChange, overrideRemoteChange) => {
  // this is the default implementation
  switch (response.status) {
    case 403:
    case 404:
      return discardLocalChange();
    case 409:
      // last write wins by default
      response.json().then((content) => {
        const version = content[VERSION_ATTRIBUTE];
        change.value[VERSION_ATTRIBUTE] = version + 1;
        overrideRemoteChange(change.value);
      });
      break;
    default:
      return retryAfter(DEFAULT_RETRY_DELAY);
  }
}

LiveQuery

The first argument is an array of stores. Every time one of these stores is updated, the function provided in the 2nd argument will be called.

import { openDB, LiveQuery } from 'synceddb';

const db = await openDB('my awesome database');

let result;

const query = new LiveQuery(['items'], async () => {
  // result will be updated every time the 'items' store is modified
  result = await db.getAll('items');
});

Example with React

import { openDB, LiveQuery } from 'synceddb';
import { useEffect, useState } from 'react';

export default function MyComponent() {
  const [items, setItems] = useState([]);

  useEffect(() => {
    let query;

    openDB('test', 1, {
      upgrade(db) {
        db.createObjectStore('items', { keyPath: 'id' });
      },
    }).then(db => {
      query = new LiveQuery(['items'], async () => {
        setItems(await db.getAll('items'));
      });

      query.run();
    });

    return () => {
      // !!! IMPORTANT !!! This ensures the query stops listening to the database updates and does not leak memory.
      query?.close();
    }
  }, []);

  return (
    <div>
      <!-- ... -->
    </div>
  );
}

Example with Vue.js

<script setup>
import { openDB, LiveQuery } from 'synceddb';
import { ref, onBeforeUnmount } from 'vue';

const items = ref([]);

const db = await openDB('test', 1, {
  upgrade(db) {
    db.createObjectStore('items', { keyPath: 'id' });
  },
})

const query = new LiveQuery(['items'], async () => {
  items.value = await db.getAll('items');
});

await query.run();

onBeforeUnmount(() => {
  // !!! IMPORTANT !!! This ensures the query stops listening to the database updates and does not leak memory.
  query.close();
});
</script>

createComputedStore()

Arguments:

  • db: the database object
  • computedStoreName: the name of the computed store
  • mainStoreName: the name of main source store (used to init the computed store)
  • secondaryStoreNames: the names of any additional source stores
  • onChange: the handler for the change
    • tx: the transaction
    • change: the change (fields: storeName, operation, key, value)

Example:

import { openDB, createComputedStore } from 'synceddb/with-async-ittr';

const db = await openDB('my awesome database');

await createComputedStore(db, 'invoices-with-customer', 'invoices', ['customers'], async (tx, change) => {
  const computedStore = tx.objectStore('invoices-with-customer');

  if (change.storeName === 'invoices') {
    if (change.operation === 'add' || change.operation === 'put') {
      const invoice = change.value;
      // fetch the customer object
      invoice.customer = await tx.objectStore('customers').get(invoice.customerId);
      // update the computed store
      computedStore.put(invoice);
    } else { // change.operation === 'delete'
      computedStore.delete(change.key);
    }
  }

  if (change.storeName === 'customers') {
    if (change.operation === 'put') {
      const customer = change.value;
      // update all invoices with the given customer in the computed store
      for await (const cursor of computedStore.index('by-customer').iterate(change.key)) {
        const invoice = cursor.value;
        if (invoice.customerId === customer.id) {
          invoice.customer = customer;
          cursor.update(invoice);
        }
      }
    }
  }
});

In the example above, the invoices-with-customer store will be updated every time the invoices or the customers store is updated, either by a manual update from the user or when fetching updates from the server.

Expectations for the REST API

Fetching changes

Changes are fetched from the REST API with GET requests:

GET /<storeName>?sort=updated_at:asc&size=100&after=2000-01-01T00:00:00.000Z&after_id=123

Explanations:

  • sort=updated_at:asc indicates that we want to sort the entities based on the date of last update
  • size=100 indicates that we want 100 entities max
  • after=2000-01-01T00:00:00.000Z&after_id=123 indicates the offset (with an update date above 2000-01-01T00:00:00.000Z, excluding the entity 123)

The query parameters can be customized with the buildFetchParams option.

Expected response:

{
  data: [
    {
      id: 1,
      version: 1,
      updatedAt: '2000-01-01T00:00:00.000Z',
      label: 'Dagger'
    },
    {
      id: 2,
      version: 12,
      updatedAt: '2000-01-02T00:00:00.000Z',
      label: 'Long sword'
    },
    {
      id: 3,
      version: -1, // tombstone
      updatedAt: '2000-01-03T00:00:00.000Z',
    }
  ],
  hasMore: true
}

A fetch request will be sent for each store of the database, every X seconds (see the fetchInterval option).

Pushing changes

Each successful readwrite transaction will be translated into an HTTP request, when the connection is available:

Operation HTTP request Body
db.add('items', { id: 1, label: 'Dagger' }) POST /items { id: 1, version: 1, label: 'Dagger' }
db.put('items', { id: 2, version: 2, label: 'Long sword' }) PUT /items/2 { id: 2, version: 3, label: 'Long sword' }
db.delete('items', 3) DELETE /items/3
db.clear('items') one DELETE request per item

Success must be indicated by an HTTP 2xx response. Any other response status means the change was not properly synced. You can customize the error handling behavior with the onpusherror method.

Please see the Express server there for reference.

Alternatives

Here are some alternatives that you might find interesting:

Miscellaneous