We're looking for a user researcher to understand the needs of developers and designers. Is this you or someone you know? Check out the post: https://mzl.la/2IGzdXS

IndexedDB는 사용자의 브라우저에 데이터를 영구적으로 저장할 수 있는 방법 중 하나입니다. IndexedDB를 사용하여 네트워크 상태에 상관없이 풍부한 쿼리 기능을 이용할 수 있는 웹 어플리케이션을 만들 수 있기 때문에, 여러분의 웹 어플리케이션은 온라인과 오프라인 환경에서 모두 동작할 수 있습니다. 

이 문서에 대하여

여러분은 이 튜토리얼에서 IndexedDB의 비동기 방식(asynchronous) API에 대해 훑어볼 수 있습니다. 만약 IndexedDB가 생소하다면, Basic Concepts About IndexedDB 를 먼저 읽어보는 것이 좋습니다.

IndexedDB API에 대한 참조(reference) 문서를 원한다면, IndexedDB 항목과 하위 페이지를 보십시오. 이 문서에서는 IndexedDB에서 사용되는 객체의 종류와, 동기식(synchrounous), 비동기식(asynchronous) API에 대해서 기술하고 있습니다.

기본 패턴

IndexedDB가 권장하는 기본 패턴은 다음과 같습니다:

  1. 데이터베이스를 엽니다.
  2. 객체 저장소(Object store)를 생성합니다. 
  3. 트랜젝션(Transaction)을 시작하고, 데이터를 추가하거나 읽어들이는 등의 데이터베이스 작업을 요청합니다.
  4. DOM 이벤트 리스너를 사용하여 요청이 완료될때까지 기다립니다.
  5. (요청 객체에서 찾을 수 있는) 결과를 가지고 무언가를 합니다.

그러면 이제, 이런 큰 개념을 익히면 더 구체적인 것을 할 수 있습니다.

저장소를 생성하고 구성하기

표준이 아직 완성되어가는 중이기 때문에, IndexedDB의 구현은 브라우저 접두사(Prefix) 밑에 숨겨져 있습니다. 표준 IndexedDB API가 확정될때까지 브라우저 제조사들은 서로 다른 IndexedDB의 구현을 가지고 있을 수도 있습니다. 하지만 표준에 대한 합의가 이루어지고 나면 제조사들은 접두사 없이 이것을 구현할 것입니다. 사실 다음과 같은 몇몇 구현은 접두사를 제거했습니다: Internet Explorer 10, Firefox 16, Chrome 24. 만약 제조사가 접두사를 사용한다면, 게코 기반의(Gecko-based) 브라우저는 moz 접두사를 사용하고, 웹킷 기반의(WebKit-based) 브라우저는 webkit 접두사를 사용합니다.

Indexed DB 의 실험적인 버전을 사용해보기

이번 건에서 당신은 prefix 를 계속 사용하여 당신의 프로그래밍 코드를 브라우져에서 테스트하기를 원합니다, 이 경우 당신은 밑의 코드를 쓸 수있습니다:  

// In the following line, you should include the prefixes of implementations you want to test.
window.indexedDB = window.indexedDB || window.mozIndexedDB || window.webkitIndexedDB || window.msIndexedDB;
// DON'T use "var indexedDB = ..." if you're not in a function.
// Moreover, you may need references to some window.IDB* objects:
window.IDBTransaction = window.IDBTransaction || window.webkitIDBTransaction || window.msIDBTransaction;
window.IDBKeyRange = window.IDBKeyRange || window.webkitIDBKeyRange || window.msIDBKeyRange
// (Mozilla has never prefixed these objects, so we don't need window.mozIDB*)

Beware that implementations that use prefix may be buggy, or incomplete, or following an old version of the specification. Therefore, it is not recommended to use it in production code. It may be preferable to not support a browser than to claim to support it and fail:

if (!window.indexedDB) {
    window.alert("Your browser doesn't support a stable version of IndexedDB. Such and such feature will not be available.")

데이터베이스 열기

우리는 밑의 프로그래밍 코드로 시작할 것입니다:

// 내 데이터 베이스를 열도록 요청하자
var request = window.indexedDB.open("MyTestDatabase");

보셨나요? 데이터베이스 접속은 다른 연살들과 비슷합니다 — 당신은 "요청(request)" 하면 됩니다.

The IndexedDB object has just a single method, open(), which, when called, opens the database called "MyTestDatabase." All IndexedDB databases are stored in the same origin, so mozilla.com might have a database named "binky" and mozilla.org might have a totally distinct database named "binky". If the database does not exist, then it is created; if the database already exists, then it is simply opened.

The open request doesn't open the database or start the transaction right away.  The call to the open() function returns an IDBOpenDBRequest object with a result (success) or error value that you handle as an event. Most other asynchronous functions in IndexedDB do the same thing - return an IDBRequest object with the result or error. The result for the open function is an instance of an IDBDatabase.

The open method accepts a second parameter, the version of the database. It allows you to update the schema of the database, i.e. creating or deleting object stores, if this one is not up-to-date. In that case, you implement the onupgradeneeded handler, with a versionchange transaction that allows you to deal with object stores - more on this later in Updating the version of the database, below. An example of opening of the database with the expected version indicated is as follows:

var request = indexedDB.open("MyTestDatabase", 3);

제어 객체 생성

첫번째로 당신이 하려는 모든 요청에 대해 성공했을 때 그리고 에러가 발생했을 때 제어를 할 객체를 요청해야 됩니다:

request.onerror = function(event) {
  // request.errorCode 에 대해 무언가를 한다!
request.onsuccess = function(event) {
  // request.result 에 대해 무언가를 한다!

Which of the two functions, onsuccess() or onerror(), gets called? If everything succeeds, a success event (that is, a DOM event whose type property is set to "success") is fired with request as its target. Once it is fired, the onsuccess() function on request is triggered with the success event as its argument. Otherwise, if everything didn't succeed, an error event (that is, a DOM event whose type property is set to "error") is fired at request. This triggers the onerror() function with the error event as its argument.

The IndexedDB API is designed to minimize the need for error handling, so you're not likely to see many error events (at least, not once you're used to the API!). In the case of opening a database, however, there are some common conditions that generate error events. The most likely problem is that the user decided not to allow your web app access to create a database. One of the main design goals of IndexedDB is to allow large amounts of data to be stored for offline use. (To learn more about how much storage you can have for each browser, see Storage limits).  

Obviously, browsers do not want to allow some advertising network or malicious website to pollute your computer, so browsers prompt the user the first time any given web app attempts to open an IndexedDB storage. The user can choose to allow or deny access. Also, IndexedDB is completely disabled in the privacy modes of browsers (Private Browsing mode for Firefox and Incognito mode for Chrome). The whole point of private browsing is to leave no footprints, so attempting to open a database fails while in this mode.

Now, assuming that the user allowed your request to create a database, and you've received a success event to trigger the success callback; What next? The request here was generated with a call to indexedDB.open(), so request.result is an instance of IDBDatabase, and you definitely want to save that for later. Your code might look something like this:

var db;
var request = indexedDB.open("MyTestDatabase");
request.onerror = function(event) {
  alert("Why didn't you allow my web app to use IndexedDB?!");
request.onsuccess = function(event) {
  db = request.result;

Handling Errors

As mentioned above, error events bubble. Error events are targeted at the request that generated the error, then the event bubbles to the transaction, and then finally to the database object. If you want to avoid adding error handlers to every request, you can instead add a single error handler on the database object, like so:

db.onerror = function(event) {
  // Generic error handler for all errors targeted at this database's
  // requests!
  alert("Database error: " + event.target.errorCode);

One of the common possible errors when opening a database is VER_ERR. It indicates that the version of the database stored on the disk is greater than the version that you are trying to open. This is an error case that must always be handled by the error handler.

Creating or updating the version of the database

To update the schema of the database, i.e. creating or deleting object stores, you implement the onupgradeneeded handler, which will be called as part of a versionchange transaction that allows you to deal with object stores.

// This event is only implemented in recent browsers
request.onupgradeneeded = function(event) { 
   // Update object stores and indices .... 

This versionchange transaction is created when a database is opened for the first time or when the specified version number is greater than the version number of the database that is currently persisted.

버전은 unsigned long long number 타입으로 간단하게 매우 큰 숫자입니다.

It also means that you can't use a float, otherwise it will be converted to the closest lower integer and the transaction may not start, and the upgradeneeded event neither as a result. eg don't use 2.4 as a version:

var request = indexedDB.open("MyTestDatabase", 2.4); // don't do this, as the version will be rounded to 2

WebKit supports the current version of the spec but only Chrome 23 ships it. Other and older implementations don't implement the current version of the spec, and thus does not support the indexedDB.open(name, version).onupgradeneeded signature yet. For more information on how to upgrade the version of the database in older Webkit, see the IDBDatabase reference article.

데이터베이스 구성

Now to structure the database. IndexedDB uses object stores rather than tables, and a single database can contain any number of object stores. Whenever a value is stored in an object store, it is associated with a key. There are several different ways that a key can be supplied depending on whether the object store uses a key path or a key generator.

The following table shows the different ways the keys are supplied. 

Key Path Key Generator Description
No No This object store can hold any kind of value, even primitive values like numbers and strings. You must supply a separate key argument whenever you want to add a new value.
Yes No This object store can only hold JavaScript objects. The objects must have a property with the same name as the key path.
No Yes This object store can hold any kind of value. The key is generated for you automatically, or you can supply a separate key argument if you want to use a specific key.
Yes Yes This object store can only hold JavaScript objects. Usually a key is generated and the value of the generated key is stored in the object in a property with the same name as the key path. However, if such a property already exists, the value of that property is used as key rather than generating a new key.

You can also create indices on any object store, provided the object store holds objects, not primitives. An index lets you look up the values stored in an object store using the value of a property of the stored object, rather than the object's key.

Additionally, indexes have the ability to enforce simple constraints on the stored data. By setting the unique flag when creating the index, the index ensures that no two objects are stored with both having the same value for the index's key path. So, for example, if you have an object store which holds a set of people, and you want to ensure that no two people have the same email address, you can use an index with the unique flag set to enforce this.

That may sound confusing, but this simple example should illustrate the concepts:

// This is what our customer data looks like.
const customerData = [
  { ssn: "444-44-4444", name: "Bill", age: 35, email: "bill@company.com" },
  { ssn: "555-55-5555", name: "Donna", age: 32, email: "donna@home.org" }
const dbName = "the_name";

var request = indexedDB.open(dbName, 2);

request.onerror = function(event) {
  // Handle errors.
request.onupgradeneeded = function(event) {
  var db = event.target.result;

  // Create an objectStore to hold information about our customers. We're
  // going to use "ssn" as our key path because it's guaranteed to be
  // unique.
  var objectStore = db.createObjectStore("customers", { keyPath: "ssn" });

  // Create an index to search customers by name. We may have duplicates
  // so we can't use a unique index.
  objectStore.createIndex("name", "name", { unique: false });

  // Create an index to search customers by email. We want to ensure that
  // no two customers have the same email, so use a unique index.
  objectStore.createIndex("email", "email", { unique: true });

  // Store values in the newly created objectStore.
  for (var i in customerData) {

As mentioned previously, onupgradeneeded is the only place where you can alter the structure of the database. In it, you can create and delete object stores and build and remove indices.

Object stores are created with a single call to createObjectStore(). The method takes a name of the store, and a parameter object. Even though the parameter object is optional, it is very important, because it lets you define important optional properties and refine the type of object store you want to create. In our case, we've asked for an object store named "customers" and  defined a keyPath that is the property that makes an individual object in the store unique. That property in this example is "ssn" since a social security number is guaranteed to be unique. "ssn" must be present on every object that is stored in the objectStore. 

We've also asked for an index named "name" that looks at the name property of the stored objects. As with createObjectStore(), createIndex() takes an optional options object that refines the type of  index that you want to create. Adding objects that don't have a name property still succeeds, but the object won't appear in the "name" index.

We can now retrieve the stored customer objects using their ssn from the object store directly, or using their name by using the index. To learn how this is done, see the section on using an index.

데이터 추가 및 제거

Before you can do anything with your new database, you need to start a transaction. Transactions come from the database object, and you have to specify which object stores you want the transaction to span. Also, you need to decide if you're going to make changes to the database or if you just need to read from it.  Although transactions have three modes (read-only, read/write, and versionchange), you're better off using a read-only transaction when you can, because they can run concurrently

데이터베이스에 데이터 추가

If you've just created a database, then you probably want to write to it. Here's what that looks like:

var transaction = db.transaction(["customers"], "readwrite");
// Note: Older experimental implementations use the deprecated constant IDBTransaction.READ_WRITE instead of "readwrite".
// In case you want to support such an implementation, you can just write: 
// var transaction = db.transaction(["customers"], IDBTransaction.READ_WRITE);

The transaction() function takes three arguments (though two are optional) and returns a transaction object. The first argument is a list of object stores that the transaction will span. You can pass an empty array if you want the transaction to span all object stores. If you don't specify anything for the second argument, you get a read-only transaction. Since you want to write to it here you need to pass the "readwrite" flag.

Now that you have a transaction you need to understand its lifetime. Transactions are tied very closely to the event loop. If you make a transaction and return to the event loop without using it then the transaction will become inactive. The only way to keep the transaction active is to make a request on it. When the request is finished you'll get a DOM event and, assuming that the request succeeded, you'll have another opportunity to extend the transaction during that callback. If you return to the event loop without extending the transaction then it will become inactive, and so on. As long as there are pending requests the transaction remains active. Transaction lifetimes are really very simple but it might take a little time to get used to. A few more examples will help, too. If you start seeing TRANSACTION_INACTIVE_ERR error codes then you've messed something up.

Transactions can receive DOM events of three different types: error, abort, and complete. We've talked about the way that error events bubble, so a transaction  receives error events from any requests that are generated from it. A more subtle point here is that the default behavior of an error is to abort the transaction in which it occurred. Unless you handle the error by calling preventDefault() on the error event, the entire transaction is rolled back. This design forces you to  think about and handle errors, but you can always add a catchall error handler to the database if fine grained error handling is too cumbersome. If you don't handle an error event or if you call abort() on the transaction, then the transaction is rolled back and an abort event is fired on the transaction. Otherwise, after all pending requests have completed, you'll get a complete event. If you're doing lots of database operations, then tracking the transaction rather than individual requests can certainly aide your sanity.

Now that you have a transaction, you'll need to get the object store from it. Transactions only let you have an object store that you specified when creating the transaction. Then you can add all the data you need.

// Do something when all the data is added to the database.
transaction.oncomplete = function(event) {
  alert("All done!");

transaction.onerror = function(event) {
  // Don't forget to handle errors!

var objectStore = transaction.objectStore("customers");
for (var i in customerData) {
  var request = objectStore.add(customerData[i]);
  request.onsuccess = function(event) {
    // event.target.result == customerData[i].ssn

The result of a request generated from a call to add() is the key of the value that was added. So in this case, it should equal the ssn property of the object that was added, since the object store uses the ssn property for the key path. Note that the add() function requires that no object already be in the database with the same key. If you're trying to modify an existing entry, or you don't care if one exists already, use the put() function.

Removing data from the database

Removing data is very similar:

var request = db.transaction(["customers"], "readwrite")
request.onsuccess = function(event) {
  // It's gone!

Getting data from the database

Now that the database has some info in it, you can retrieve it in several ways. First, the simple get(). You need to provide the key to retrieve the value, like so:

var transaction = db.transaction(["customers"]);
var objectStore = transaction.objectStore("customers");
var request = objectStore.get("444-44-4444");
request.onerror = function(event) {
  // Handle errors!
request.onsuccess = function(event) {
  // Do something with the request.result!
  alert("Name for SSN 444-44-4444 is " + request.result.name);

That's a lot of code for a "simple" retrieval. Here's how you can shorten it up a bit, assuming that you handle errors at the database level:

db.transaction("customers").objectStore("customers").get("444-44-4444").onsuccess = function(event) {
  alert("Name for SSN 444-44-4444 is " + event.target.result.name);

See how this works? Since there's only one object store, you can avoid passing a list of object stores you need in your transaction and just pass the name as a string. Also, you're only reading from the database, so you don't need a "readwrite" transaction. Calling transaction() with no mode specified gives you a "readonly" transaction. Another subtlety here is that you don't actually save the request object to a variable. Since the DOM event has the request as its target you can use the event to get to the result property. Easy, right?!

Using a cursor

Using get() requires that you know which key you want to retrieve. If you want to step through all the values in your object store, then you can use a cursor. Here's what it looks like:

var objectStore = db.transaction("customers").objectStore("customers");

objectStore.openCursor().onsuccess = function(event) {
  var cursor = event.target.result;
  if (cursor) {
    alert("Name for SSN " + cursor.key + " is " + cursor.value.name);
  else {
    alert("No more entries!");

The openCursor() function takes several arguments. First, you can limit the range of items that are retrieved by using a key range object that we'll get to in a minute. Second, you can specify the direction that you want to iterate. In the above example, we're iterating over all objects in ascending order. The success callback for cursors is a little special. The cursor object itself is the result of the request (above we're using the shorthand, so it's event.target.result). Then the actual key and value can be found on the key and value properties of the cursor object. If you want to keep going, then you have to call continue() on the cursor. When you've reached the end of the data (or if there were no entries that matched your openCursor() request) you still get a success callback, but the result property is undefined.

One common pattern with cursors is to retrieve all objects in an object store and add them to an array, like this:

var customers = [];

objectStore.openCursor().onsuccess = function(event) {
  var cursor = event.target.result;
  if (cursor) {
  else {
    alert("Got all customers: " + customers);
Warning: The following function is not part of the IndexedDB standard!

Mozilla has also implemented getAll() to handle this case. It isn't part of the IndexedDB standard, so it may disappear in the future. We've included it because we think it's useful. The following code does precisely the same thing as above:

objectStore.getAll().onsuccess = function(event) {
  alert("Got all customers: " + event.target.result);

There is a performance cost associated with looking at the value property of a cursor, because the object is created lazily. When you use getAll(), Gecko must create all the objects at once. If you're just interested in looking at each of the keys, for instance, it is much more efficient to use a cursor than to use getAll(). If you're trying to get an array of all the objects in an object store, though, use getAll().

Using an index

Storing customer data using the SSN as a key is logical since the SSN uniquely identifies an individual. (Whether this is a good idea for privacy is a different question, outside the scope of this article.) If you need to look up a customer by name, however, you'll need to iterate over every SSN in the database until you find the right one. Searching in this fashion would be very slow, so instead you can use an index.

var index = objectStore.index("name");
index.get("Donna").onsuccess = function(event) {
  alert("Donna's SSN is " + event.target.result.ssn);

The "name" cursor isn't unique, so there could be more than one entry with the name set to "Donna". In that case you always get the one with the lowest key value.

If you need to access all the entries with a given name you can use a cursor. You can open two different types of cursors on indexes. A normal cursor maps the index property to the object in the object store. A key cursor maps the index property to the key used to store the object in the object store. The differences are illustrated here:

index.openCursor().onsuccess = function(event) {
  var cursor = event.target.result;
  if (cursor) {
    // cursor.key is a name, like "Bill", and cursor.value is the whole object.
    alert("Name: " + cursor.key + ", SSN: " + cursor.value.ssn + ", email: " + cursor.value.email);

index.openKeyCursor().onsuccess = function(event) {
  var cursor = event.target.result;
  if (cursor) {
    // cursor.key is a name, like "Bill", and cursor.value is the SSN.
    // No way to directly get the rest of the stored object.
    alert("Name: " + cursor.key + ", "SSN: " + cursor.value);

Specifying the range and direction of cursors

If you would like to limit the range of values you see in a cursor, you can use a key range object and pass it as the first argument to openCursor() or openKeyCursor(). You can make a key range that only allows a single key, or one the has a lower or upper bound, or one that has both a lower and upper bound. The bound may be "closed" (i.e., the key range includes the given value) or "open" (i.e., the key range does not include the given value). Here's how it works:

// Only match "Donna"
var singleKeyRange = IDBKeyRange.only("Donna");

// Match anything past "Bill", including "Bill"
var lowerBoundKeyRange = IDBKeyRange.lowerBound("Bill");

// Match anything past "Bill", but don't include "Bill"
var lowerBoundOpenKeyRange = IDBKeyRange.lowerBound("Bill", true);

// Match anything up to, but not including, "Donna"
var upperBoundOpenKeyRange = IDBKeyRange.upperBound("Donna", true);

//Match anything between "Bill" and "Donna", but not including "Donna"
var boundKeyRange = IDBKeyRange.bound("Bill", "Donna", false, true);

index.openCursor(boundKeyRange).onsuccess = function(event) {
  var cursor = event.target.result;
  if (cursor) {
    // Do something with the matches.

Sometimes you may want to iterate in descending order rather than in ascending order (the default direction for all cursors). Switching direction is accomplished by passing prev to the openCursor() function:

objectStore.openCursor(null, IDBCursor.prev).onsuccess = function(event) {
  var cursor = event.target.result;
  if (cursor) {
    // Do something with the entries.

Since the "name" index isn't unique, there might be multiple entries where name is the same. Note that such a situation cannot occur with object stores since the key must always be unique. If you wish to filter out duplicates during cursor iteration over indexes, you can pass nextunique (or prevunique if you're going backwards) as the direction parameter. When nextunique or prevunique is used, the entry with the lowest key is always the one returned.

index.openKeyCursor(null, IDBCursor.nextunique).onsuccess = function(event) {
  var cursor = event.target.result;
  if (cursor) {
    // Do something with the entries.

Version changes while a web app is open in another tab

When your web app changes in such a way that a version change is required for your database, you need to consider what happens if the user has the old version of your app open in one tab and then loads the new version of your app in another. When you call open() with a greater version than the actual version of the database, all other open databases must explicitly acknowledge the request before you can start making changes to the database. Here's how it works:

var openReq = mozIndexedDB.open("MyTestDatabase", 2);

openReq.onblocked = function(event) {
  // If some other tab is loaded with the database, then it needs to be closed
  // before we can proceed.
  alert("Please close all other tabs with this site open!");
openReq.onupgradeneeded = function(event) {
  // All other databases have been closed. Set everything up.
  db.createObjectStore(/* ... */);
openReq.onsuccess = function(event) {
  var db = event.target.result;

function useDatabase(db) {
  // Make sure to add a handler to be notified if another page requests a version
  // change. We must close the database. This allows the other page to upgrade the database.
  // If you don't do this then the upgrade won't happen until the user close the tab.
  db.onversionchange = function(event) {
    alert("A new version of this page is ready. Please reload!");

  // Do stuff with the database.


IndexedDB uses the same-origin principle, which means that it ties the store to the origin of the site that creates it (typically, this is the site domain or subdomain), so it cannot be accessed by any other origin.

It's important to note that IndexedDB doesn't work for content loaded into a frame from another site (either <frame> or <iframe>. This is a security measure. Details as to why this matters are forthcoming. See bug 595307.

Next step

If you want to start tinkering with the API, jump in to the reference documentation and checking out the different methods.

See also



Related articles


문서 태그 및 공헌자

이 페이지의 공헌자: yongbam, WKBae, jhlee1979
최종 변경: yongbam,