Working with IndexedDB

A guide to the basics of IndexedDB.

Jan 1, 2016 — Updated Aug 9, 2022

This text guides you through the basics of the IndexedDB API. We are using Jake Archibald's IndexedDB Promised library, which is very similar to the IndexedDB API, but uses promises rather than events. This simplifies the API while maintaining its structure, so anything you learn using this library can be applied to the IndexedDB API directly.

What is IndexedDB? #

IndexedDB is a large-scale, NoSQL storage system. It lets you store just about anything in the user's browser. In addition to the usual search, get, and put actions, IndexedDB also supports transactions. Here is the definition of IndexedDB on MDN:

IndexedDB is a low-level API for client-side storage of significant amounts of structured data, including files/blobs. This API uses indexes to enable high performance searches of this data. While DOM Storage is useful for storing smaller amounts of data, it is less useful for storing larger amounts of structured data. IndexedDB provides a solution.

Each IndexedDB database is unique to an origin (typically, this is the site domain or subdomain), meaning it cannot access or be accessed by any other origin. Data storage limits are usually quite large, if they exist at all, but different browsers handle limits and data eviction differently. See the Further reading section for more information.

IndexedDB terms #

Database: The highest level of IndexedDB. It contains the object stores, which in turn contain the data you would like to persist. You can create multiple databases with whatever names you choose.
Object store: An individual bucket to store data. You can think of object stores as being similar to tables in traditional relational databases. Typically, there is one object store for each type (not JavaScript data type) of data you are storing. For example, given an app that persists blog posts and user profiles, you could imagine two object stores. Unlike tables in traditional databases, the actual JavaScript data types of data within the store do not need to be consistent (for example, if there are three people in the people object store, their age properties could be 53, 'twenty-five', and unknown ).
Index: A kind of object store for organizing data in another object store (called the reference object store) by an individual property of the data. The index is used to retrieve records in the object store by this property. For example, if you're storing people, you may want to fetch them later by their name, age, or favorite animal.
Operation: An interaction with the database.
Transaction: A wrapper around an operation, or group of operations, that ensures database integrity. If one of the actions within a transaction fails, none of them are applied and the database returns to the state it was in before the transaction began. All read or write operations in IndexedDB must be part of a transaction. This allows for atomic read-modify-write operations without having to worry about other threads acting on the database at the same time.
Cursor: A mechanism for iterating over multiple records in a database.

Checking for IndexedDB support #

IndexedDB is almost universally supported, with the one exception being Opera Mini. It's still a good idea to feature-detect support just in case. The easiest way is to check the window object:

if (!('indexedDB' in window)) {
  console.log("This browser doesn't support IndexedDB");
  return;
}

You simply place this function at the beginning of our scripts and you're ready to use IndexedDB.

Opening a database #

With IndexedDB, you can create multiple databases with any names you choose. To open a database, you use:

idb.open(name, version, upgradeCallback);

This method returns a promise that resolves to a database object. When using idb.open, you provide a name, version number, and an optional callback to set up the database.

Here is an example of idb.open() in context:

// Check for support.
if (!('indexedDB' in window)) {
  console.log("This browser doesn't support IndexedDB.");
  return;
}

const dbPromise = idb.open('test-db1', 1);

You place the check for IndexedDB support at the top of the anonymous function. This exits out of the function if the browser doesn't support IndexedDB. Then you call idb.open() to open a database named "test-db1". In this example, the optional callback function has been left out to keep things simple.

Working with object stores #

Creating object stores #

A database typically contains one or more object stores. Object stores can be thought of as similar to tables in SQL databases and should contain objects of the same type (not JavaScript data type). For example, for a site persisting user profiles and notes, you can imagine a people object store containing person objects, and a notes object store. A well structured IndexedDB database should have one object store for each type of data that needs to be persisted.

To ensure database integrity, object stores can only be created and removed in the callback function in idb.open(). The callback receives an instance of UpgradeDB, a special object in the IDB Promised library that is used to create object stores. Call the createObjectStore() method on UpgradeDB to create the object store:

upgradeDb.createObjectStore('storeName', options);

This method takes the name of the object store as well as a parameter object that lets you define various configuration properties for the object store.

Below is an example of the createObjectStore() method:

// Check for support.
if (!('indexedDB' in window)) {
  console.log("This browser doesn't support IndexedDB.");
  return;
}

const dbPromise = idb.open('test-db2', 1, function (upgradeDb) {
  console.log('Creating a new object store.');
  if (!upgradeDb.objectStoreNames.contains('firstOS')) {
    upgradeDb.createObjectStore('firstOS');
  }
});

Again, you first check the browser for IndexedDB support. This time you include the callback function in idb.open() in order to create the object store. The browser throws an error if you try to create an object store that already exists in the database so you wrap the createObjectStore() method in an if statement that checks if the object store exists. Inside the if block you call createObjectStore() on the UpgradeDB object to create an object store named "firstOS".

Defining primary keys #

When you define object stores, you can define how data is uniquely identified in the store using the primary key. You can define a primary key by either defining a key path, or by using a key generator.

A key path is a property that always exists and contains a unique value. For example, in the case of a people object store, you could choose the email address as the key path.

upgradeDb.createObjectStore('people', { keyPath: 'email' });

This example creates an object store called "people" and assigns the email property as the primary key.

You could also use a key generator, such as autoIncrement. The key generator creates a unique value for every object added to the object store. By default, if you don't specify a key, IndexedDB creates a key and stores it separately from the data.

upgradeDb.createObjectStore('notes', { autoIncrement: true });

This example creates an object store called "notes" and sets the primary key to be assigned automatically as an auto incrementing number.

upgradeDb.createObjectStore('logs', { keyPath: 'id', autoIncrement: true });

This example is similar to the previous example, but this time the auto incrementing value is assigned to a property called "id".

Choosing which method to use to define the key depends on your data. If your data has a property that is always unique, you can make it the keypath to enforce this uniqueness. Otherwise, using an auto incrementing value makes sense.

Let's look at an example:

// Check for support.
if (!('indexedDB' in window)) {
  console.log("This browser doesn't support IndexedDB.");
  return;
}

const dbPromise = idb.open('test-db3', 1, function (upgradeDb) {
  if (!upgradeDb.objectStoreNames.contains('people')) {
    upgradeDb.createObjectStore('people', { keyPath: 'email' });
  }
  if (!upgradeDb.objectStoreNames.contains('notes')) {
    upgradeDb.createObjectStore('notes', { autoIncrement: true });
  }
  if (!upgradeDb.objectStoreNames.contains('logs')) {
    upgradeDb.createObjectStore('logs', { keyPath: 'id', autoIncrement: true });
  }
});

This code creates three object stores demonstrating the various ways of defining primary keys in object stores.

Defining indexes #

Indexes are a kind of object store used to retrieve data from the reference object store by a specified property. An index lives inside the reference object store and contains the same data, but uses the specified property as its key path instead of the reference store's primary key. Indexes must be made when you create your object stores and can also be used to define a unique constraint on your data.

To create an index, call the createIndex() method on an object store instance:

objectStore.createIndex('indexName', 'property', options);

This method creates and returns an index object. The createIndex() method takes the name of the new index as the first argument and the second argument refers to the property on the data you want to index. The final argument lets you define two options that determine how the index operates: unique and multiEntry. If unique is set to true, the index does not allow duplicate values for a single key. Next, multiEntry determines how createIndex() behaves when the indexed property is an array. If it's set to true, createIndex() adds an entry in the index for each array element. Otherwise, it adds a single entry containing the array.

Here is an example:

// Check for support.
if (!('indexedDB' in window)) {
  console.log("This browser doesn't support IndexedDB.");
  return;
}

const dbPromise = idb.open('test-db4', 1, function (upgradeDb) {
  if (!upgradeDb.objectStoreNames.contains('people')) {
    const peopleOS = upgradeDb.createObjectStore('people', { keyPath: 'email' });
    peopleOS.createIndex('gender', 'gender', { unique: false });
    peopleOS.createIndex('ssn', 'ssn', { unique: true });
  }
  if (!upgradeDb.objectStoreNames.contains('notes')) {
    const notesOS = upgradeDb.createObjectStore('notes', { autoIncrement: true });
    notesOS.createIndex('title', 'title', { unique: false });
  }
  if (!upgradeDb.objectStoreNames.contains('logs')) {
    const logsOS = upgradeDb.createObjectStore('logs', { keyPath: 'id', autoIncrement: true });
  }
});

In this example, the "people" and "notes" object stores have indexes. To create the indexes, you first assign the result of createObjectStore() (which is an object store object) to a variable so you can call createIndex() on it.

Working with data #

This section describes how to create, read, update, and delete data. These operations are all asynchronous, using promises where the IndexedDB API uses requests. This simplifies the API. Instead of listening for events triggered by the request, you can simply call .then() on the database object returned from idb.open() to start interactions with the database or await its creation.

All data operations in IndexedDB are carried out inside a transaction. Each operation has this form:

Get database object
Open transaction on database
Open object store on transaction
Perform operation on object store

A transaction can be thought of as a safe wrapper around an operation or group of operations. If one of the actions within a transaction fails, all of the actions are rolled back. Transactions are specific to one or more object stores, which you define when you open the transaction. They can be read-only or read and write. This signifies whether the operations inside the transaction read the data or make a change to the database.

Creating data #

To create data, call the add() method on the object store and pass in the data you want to add. The add() method has an optional second argument that lets you define the primary key for the individual object on creation, but it should only be used if you have not specified the key path in createObjectStore(). Here is a simple example:

someObjectStore.add(data, optionalKey);

The data parameter can be data of any type: a string, number, object, array, and so forth. The only restriction is if the object store has a defined keypath, the data must contain this property and the value must be unique. The add() method returns a promise that resolves once the object has been added to the store.

Each add() call occurs within a transaction, so even if the promise resolves successfully, it doesn't necessarily mean the operation worked. Remember, if one of the actions in the transaction fails, all of the operations in the transaction are rolled back. To be sure that the add operation was carried out, you need to check if the whole transaction has completed using the transaction.complete() method. This is a promise that resolves when the transaction completes and rejects if the transaction errors. Note that this method doesn't actually close the transaction. The transaction completes on its own. You must perform this check for all "write" operations, because it is your only way of knowing that the changes to the database have actually been carried out.

Let's look at an example of the add method:

dbPromise
  .then(function (db) {
    const tx = db.transaction('store', 'readwrite');
    const store = tx.objectStore('store');
    const item = {
      name: 'sandwich',
      price: 4.99,
      description: 'A very tasty sandwich',
      created: new Date().getTime(),
    };
    store.add(item);
    return tx.complete;
  })
  .then(function () {
    console.log('Added item to the store!');
  });

First, you get the database object. You call .then() on dbPromise, which resolves with the database object, and pass this object to the callback function in .then(). Because dbPromise (from idb.open()) is a promise, you can safely assume that when .then() executes, the database is open and all object stores and indexes are ready for use.

The next step is to open a transaction by calling the transaction() method on the database object. This method takes a list of object stores, which defines the scope of the transaction (if it is a single store name, you can pass it directly instead of passing an array of names, as you did in the example where you only wanted the "store" object store). The transaction() method also has an optional second argument for the mode, which can be readonly or readwrite. This option is read-only by default.

You can then open the "store" object store on this transaction and assign it to the store variable. Now when you call store.add(), the add operation occurs within the transaction. Finally, you return tx.complete() and log a success message once the transaction has completed.

Reading data #

To read data, call the get() method on the object store. The get() method takes the primary key of the object you want to retrieve from the store. Here is a basic example:

someObjectStore.get(primaryKey);

As with add(), the get() method returns a promise and must happen within a transaction.

Let's look at an example of the get() method:

dbPromise
  .then(function (db) {
    const tx = db.transaction('store', 'readonly');
    const store = tx.objectStore('store');
    return store.get('sandwich');
  })
  .then(function (val) {
    console.dir(val);
  });

Once again, you start the operation by getting the database object and creating a transaction. Note that this time it is a read-only transaction because you are not writing anything to the database inside the transaction (that is, using put(), add(), or delete()). You then open the object store on the transaction and assign the resulting object store object to the store variable. Finally, you return the result of store.get() and log this object to the console.

Updating data #

To update data, call the put() method on the object store. The put() method is very similar to the add() method and can be used instead of add() to create data in the object store. Like add(), put() takes the data and an optional primary key:

someObjectStore.put(data, optionalKey);

Again, this method returns a promise and occurs inside a transaction. As with add(), you need to be careful to check transaction.complete() if you want to be sure that the operation was actually carried out.

Here is an example using the put method:

dbPromise
  .then(function (db) {
    const tx = db.transaction('store', 'readwrite');
    const store = tx.objectStore('store');
    const item = {
      name: 'sandwich',
      price: 99.99,
      description: 'A very tasty, but quite expensive, sandwich',
      created: new Date().getTime(),
    };
    store.put(item);
    return tx.complete;
  })
  .then(function () {
    console.log('Item updated!');
  });

To update an existing item in the object store, use the put() method on an object containing the same primary key value as the object in the store. Here, the assumption is that the keyPath for the store object store is the "name" property and that you are updating the price and description of the "sandwich" object. The database interaction has the same structure as the create and read operations: get the database object, create a transaction, open an object store on the transaction, and perform the operation on the object store.

Deleting data #

To delete data, call the delete() method on the object store.

someObjectStore.delete(primaryKey);

Once again, this method returns a promise and must be wrapped in a transaction. Here is a simple example:

dbPromise
  .then(function (db) {
    const tx = db.transaction('store', 'readwrite');
    const store = tx.objectStore('store');
    store.delete(key);
    return tx.complete;
  })
  .then(function () {
    console.log('Item deleted.');
  });

The structure of the database interaction is the same as for the other operations. Note that you again check that the whole transaction has completed by returning the tx.complete() method to be sure that the deletion was carried out.

Getting all the data #

So far you have only retrieved objects from the store one at a time. You can also retrieve all of the data (or a subset) from an object store or index using either the getAll() method or using cursors.

Using the getAll() method #

The simplest way to retrieve all of the data is to call the getAll() method on the object store or index, like this:

someObjectStore.getAll(optionalConstraint);

This method returns all the objects in the object store matching the specified key or key range (see Working with ranges and indexes), or all objects in the store if no parameter is given. As with all other database operations, this operation happens inside a transaction. Here is a short example:

dbPromise
  .then(function (db) {
    const tx = db.transaction('store', 'readonly');
    const store = tx.objectStore('store');
    return store.getAll();
  })
  .then(function (items) {
    console.log('Items by name:', items);
  });

Here you are calling getAll() on the "store" object store. This returns all of the objects in the store ordered by the primary key.

Using cursors #

Another way to retrieve all of the data is to use a cursor. A cursor selects each object in an object store or index one by one, letting you do something with the data as it is selected. Cursors, like the other database operations, work within transactions.

You create the cursor by calling the openCursor() method on the object store, like this:

someObjectStore.openCursor(optionalKeyRange, optionalDirection);

This method returns a promise that resolves with a cursor object representing the first object in the object store or undefined if there is no object. To move on to the next object in the object store, you call cursor.continue(). This returns a promise that resolves with the next object, or undefined if there are no more objects. You put this inside a loop to move through all of the entries in the store one by one. The optional key range in the openCursor() method limits the iteration to a subset of the objects in the store. The direction option can be next or prev, specifying forward or backward traversal through the data.

The next example uses a cursor to iterate through all the items in the "store" object store and log them to the console:

dbPromise
  .then(function (db) {
    const tx = db.transaction('store', 'readonly');
    const store = tx.objectStore('store');
    return store.openCursor();
  })
  .then(function logItems(cursor) {
    if (!cursor) {
      return;
    }
    console.log('Cursored at:', cursor.key);
    for (const field in cursor.value) {
      console.log(cursor.value[field]);
    }
    return cursor.continue().then(logItems);
  })
  .then(function () {
    console.log('Done cursoring.');
  });

As usual, you start by getting the database object, creating a transaction, and opening an object store. You call the openCursor() method on the object store and pass the cursor object to the callback function in .then(). This time you name the callback function "logItems" so you can call it from inside the function and make a loop. The line if (!cursor) {return;} breaks the loop if the promise returned by store.openCursor() resolves with undefined, or if the promise returned by cursor.continue()resolves with undefined (indicating that there are no more objects).

The cursor object contains a key property that represents the primary key for the item. It also contains a value property that represents the data. At the end of logItems(), you return cursor.continue().then(logItems). The cursor.continue() method returns a promise that resolves to a cursor object representing the next item in the store or undefined if there are no more objects. This result is passed to the callback function in .then(), which you have chosen to be logItems(), so that the function loops. Consequently, logItems() continues to call itself until no objects remain.

Working with ranges and indexes #

You can get all the data in a couple of different ways, but what if you want only a subset of the data based on a particular property? This is where indexes come in. Indexes let you fetch the data in an object store by a property other than the primary key. You can create an index on any property (which becomes the keypath for the index), specify a range on that property, and get the data within the range using the getAll() method or a cursor.

You define the range using the IDBKeyRange object. This object has five methods that are used to define the limits of the range:

As expected, the upperBound() and lowerBound() methods specify the upper and lower limits of the range.

IDBKeyRange.lowerBound(indexKey);

IDBKeyRange.upperBound(indexKey);

They each take one argument which is the index's keypath value of the item you want to specify as the upper or lower limit.

The bound() method is used to specify both an upper and lower limit, and takes the lower limit as the first argument:

IDBKeyRange.bound(lowerIndexKey, upperIndexKey);

The range for these functions is inclusive by default, but can be specified as exclusive by passing true as the second argument (or the third and fourth in the case of bound(), for the lower and upper limits respectively). An inclusive range includes the data at the limits of the range. An exclusive range does not.

Let's look at an example. For this demo, you have created an index on the "price" property in the "store" object store. You have also added a small form with two inputs for the upper and lower limits of the range. Imagine you are passing in the lower and upper bounds to the function as floating point numbers representing prices:

function searchItems(lower, upper) {
  if (lower === '' && upper === '') {
    return;
  }

  let range;
  if (lower !== '' && upper !== '') {
    range = IDBKeyRange.bound(lower, upper);
  } else if (lower === '') {
    range = IDBKeyRange.upperBound(upper);
  } else {
    range = IDBKeyRange.lowerBound(lower);
  }

  dbPromise
    .then(function (db) {
      const tx = db.transaction(['store'], 'readonly');
      const store = tx.objectStore('store');
      const index = store.index('price');
      return index.openCursor(range);
    })
    .then(function showRange(cursor) {
      if (!cursor) {
        return;
      }
      console.log('Cursored at:', cursor.key);
      for (const field in cursor.value) {
        console.log(cursor.value[field]);
      }
      return cursor.continue().then(showRange);
    })
    .then(function () {
      console.log('Done cursoring.');
    });
}

The code first gets the values for the limits and checks if the limits exist. The next block of code decides which method to use to limit the range based on the values. In the database interaction, you open the object store on the transaction as usual, then you open the "price" index on the object store. The "price" index allows you to search for the items by price. You open a cursor on the index and pass in the range. The cursor now returns a promise representing the first object in the range, or undefined if there is no data within the range. The cursor.continue() method returns a cursor representing the next object and so on through the loop until you reach the end of the range.

Using database versioning #

When you call idb.open(), you can specify the database version number in the second parameter. If this version number is greater than the version of the existing database, the upgrade callback executes, allowing you to add object stores and indexes to the database.

The UpgradeDB object has a special oldVersion property, which indicates the version number of the database existing in the browser. You can pass this version number into a switch statement to execute blocks of code inside the upgrade callback based on the existing database version number. Here's an example:

const dbPromise = idb.open('test-db7', 2, function (upgradeDb) {
  switch (upgradeDb.oldVersion) {
    case 0:
      upgradeDb.createObjectStore('store', { keyPath: 'name' });
    case 1:
      const peopleStore = upgradeDb.transaction.objectStore('store');
      peopleStore.createIndex('price', 'price');
  }
});

In the example you have set the newest version of the database at 2. When this code first executes, and since the database doesn't yet exist in the browser, upgradeDb.oldVersion is 0 and the switch statement starts at case 0. In the example, this results in a "store" object store being added to the database. Usually, in switch statements, there is a break after each case, but this is deliberately not used here. This way, if the existing database is a few versions behind (or if it doesn't exist), the code continues through the rest of the case blocks until it has executed all the latest changes. So in the example, the browser continues executing through case 1, creating a "price" index on the "store" object store. Once this has finished executing, the database in the browser is at version 2 and contains a "store" object store with a "price" index.

Let's say you now want to create a "description" index on the "store" object store. You need to update the version number and add a case, like this:

const dbPromise = idb.open('test-db7', 3, function (upgradeDb) {
  switch (upgradeDb.oldVersion) {
    case 0:
      upgradeDb.createObjectStore('store', { keyPath: 'name' });
    case 1:
      const storeOS = upgradeDb.transaction.objectStore('store');
      storeOS.createIndex('price', 'price');
    case 2:
      const storeOS = upgradeDb.transaction.objectStore('store');
      storeOS.createIndex('description', 'description');
  }
});

Assuming the database you created in the previous example still exists in the browser, when this executes upgradeDb.oldVersion is 2. The case 0 and case 1 are skipped, and the browser executes the code in case 2, which creates a "description" index. Once all this has finished, the browser has a database at version 3 containing a "store" object store with "price" and "description" indexes.

Appendix #

Comparison of IndexedDB API and IndexedDB Promised library #

The IndexedDB Promised library sits on top of the IndexedDB API, translating its requests into promises. The overall structure is the same between the library and the API and, in general, the actual syntax for the database operations is the same and they will act the same way. But there are a few differences because of the differences between requests and promises, which are covered here.

All database interactions in the IndexedDB API are requests and have associated onsuccess and onerror event handlers. These are similar to the .then() and .catch() promise functions. The indexedDB.open() method in the raw API also gets a special event handler, onupgradeneeded, which is used to create the object stores and indexes. This is equivalent to the upgrade callback in idb.open() in the Promised library. In fact, if you look through the Promised library, you will find the upgrade callback is just a convenient wrapper for the onupgradeneeded event handler.

Here's an example of the IndexedDB API. This example opens a database, adds an object store, and adds one item to the object store:

let db;

const openRequest = indexedDB.open('test_db', 1);

openRequest.onupgradeneeded = function (e) {
  db = e.target.result;
  console.log('running onupgradeneeded');
  if (!db.objectStoreNames.contains('store')) {
    const storeOS = db.createObjectStore('store', { keyPath: 'name' });
  }
};
openRequest.onsuccess = function (e) {
  console.log('running onsuccess');
  db = e.target.result;
  addItem();
};
openRequest.onerror = function (e) {
  console.log('onerror!');
  console.dir(e);
};

function addItem() {
  const transaction = db.transaction(['store'], 'readwrite');
  const store = transaction.objectStore('store');
  const item = {
    name: 'banana',
    price: '$2.99',
    description: 'It is a purple banana!',
    created: new Date().getTime(),
  };

  const request = store.add(item);

  request.onerror = function (e) {
    console.log('Error', e.target.error.name);
  };
  request.onsuccess = function (e) {
    console.log('Woot! Did it');
  };
}

This code does something very similar to previous examples in this tutorial except that it doesn't use the Promised library. You can see that the structure of the database interaction hasn't changed. Object stores are created on the database object in the upgrade event handler, and items are added to the object store in the same transaction sequence you've seen before. The difference is that this is done with requests and event handlers rather than promises and promise chains.

Here is a short reference of the differences between the IndexedDB API and the IndexedDB Promised library.

Action	IndexedDB Promised	IndexedDB API
Open database	`idb.open(name, version, upgradeCallback)`	`indexedDB.open(name, version)`
Upgrade database	Inside upgradeCallback	`request.onupgradeneeded`
Success	`.then`	`request.onsuccess`
Error	`.catch`	`request.onerror`

Last updated: Aug 9, 2022 — Improve article