Non-standard
This feature is non-standard and is not on a standards track. Do not use it on production sites facing the Web: it will not work for every user. There may also be large incompatibilities between implementations and the behavior may change in the future.
The FileHandle API allows for the manipulating of files, including creating files and modifying their content (unlike the File API). Because the files manipulated through that API can be physically stored on the device, the editing part uses a turn-based locking mechanism in order to avoid race issues.
API Overview
This API is based on the following interfaces:
IDBDatabase.mozCreateFileHandle
(was calledIDBDatabase.mozCreateFileHandle
.)IDBMutableFile
(was previouslyFileHandle
.)LockedFile
FileRequest
It also has connections with the File API, especially the File
and Blob
interfaces.
Basic operations
Create a FileHandle
Because the intent is to allow the storage of files through IndexedDB, creating a FileHandle
instance requires an IndexedDB Database.
mozCreateFileHandle()
takes two arguments: a name and an optional type. Both of these are just descriptive and are not used by the database. However, they are important for the FileHandle
object as it can generate File
objects which inherit their own name
and type
from those values. That said, as the name does not match any real filename it can be an empty string, for example, and it doesn't even have to be unique.
Note: the above code only creates a "temporary file" that exists only while you hold the FileHandle
instance. If you want a file to survive a page refresh/app relaunch, you need to store the handle in a more permanent location, like the database itself. See File storage below to learn more about this.
Perform read and write operations
To read or write within a handled file, it is required to get a LockedFile
. The FileHandle.open()
method provides such an object which can be readonly
or readwrite
. Any attempt to perform a write action on a readonly
LockedFile
object will fail.
Writing
There are three possible writing operations on a locked file:
write
: It's an arbitrary writing method which starts writing in the file at theLockedFile.location
byte.append
: This operation always writes content at the end of the file.truncate
: This operation keeps the nth-first bytes of the file and removes the rest.
// Get a LockedFile object from the handle var myFile = myFileHandle.open('readwrite'); // Start a writing operation var writing = myFile.append('Some content'); writing.onsuccess = function () { console.log('Writing operation successful'); } writing.onerror = function () { console.log('Something goes wrong in the writing process: ' + this.error); }
Reading
It's possible to directly write the content of a LockedFile
object without using an intermediate File
object and a FileReader
object. The LockedFile
interface provides a readAsText
method and a readAsArrayBuffer
method.
Those two methods expect a size to indicate how many bytes must be read starting at the LockedFile.location
byte.
To read the whole file, it is required to know its size. This information (as well as the date of its last modification) can be retrieved through the LockedFile.getMetadata()
method.
// Get a LockedFile object from the handle var myFile = myFileHandle.open('readwrite'); // Retrieve the size of the file var getmeta = myFile.getMetadata(); getmeta.onsuccess = function () { var size = this.result.size; // The reading operation will start with the byte at index 0 in the file myFile.location = 0; // Start a reading operation for the whole file content var reading = myFile.readAsText(size); reading.onsuccess = function () { console.log('The content of the file is:'); console.log(this.result); } reading.onerror = function () { console.log('Something goes wrong in the reading process: ' + this.error); } }
File snapshot
In many cases it can be handy to get a snapshot of the file. For example, there are many APIs that expect Blob
or File
objects such as domxref("FileReader")}} (which can be easier to use to read the whole file) or XMLHttpRequest
.
It's possible to get a File
object representing the current state of the file handled by the FileHandle
object by using the getFile
method. Such a File
object is completely desynchronized from the original file, which means any change made to that object will never be reflected to the handled file as well as any change made to the handled file will never be pushed to the File
object.
var mySnapshot = null; var request = myFileHandle.getFile(); request.onsuccess = function () { mySnapshot = this.result; }
Managing progress
All the methods from the LockedFile
interface return a FileRequest
object. Such an object is basically a DOMRequest
with an extra power: it allows to monitor the progress of an operation. Sometimes writing and reading operations can be very long, therefore it is a good idea to monitor the operation to provide feedback to the user. Such monitoring can be done using the FileRequest.onprogress
event handler.
var progress = document.querySelector('progress'); var myFile = myFileHandle.open('readonly'); // Let's read a 1GB file var action = myFile.readAsArrayBuffer(1000000000); action.onprogress = function (event) { if (progress) { progress.value = event.loaded; progress.max = event.total; } } action.onsuccess = function () { console.log('Yeah \o/ Just read a 1GB file'); } action.onerror = function () { console.log('Oups :( Unable to read a 1GB file') }
File storage
When a file handle is created, the associated file only exists as a "temporary file" as long as you hold the FileHandle
instance. If you want a file to survive a page refresh/app relaunch, you need to store the handle in a database (not necessarily the one used to create the FileHandle
object).
var IDBReq = window.indexedDB.open('myFileStorageDataBase'); // If necessary, let's create a datastore for the files IDBReq.onupgradeneeded = function () { this.result.createObjectStore('files'); } IDBReq.onsuccess = function () { var DB = this.result; // Let's create a new file var handleReq = DB.mozCreateFileHandle("test.txt", "plain/text"); handleReq.onsuccess = function () { var myFileHandle = this.result; var store = DB.transaction(['files'], 'readwrite').objectStore('files'); // Let's store the file permanently // HINT: it could be handy to use the file name as the storage key var storeReq = store.add(myFileHandle, myFileHandle.name); storeReq.onsuccess = function () { console.log('The file has been successfully stored and can be retrieved anytime.') } } }
A file stored that way is physically put on the device. The database itself only stores a pointer to that file. It means that if the FileHandle
object is stored several times in several DBs or several data stores, all those objects will reference the same unique file. This is not a problem because to access the file, a LockedFile
object is required and operations on such object are performed in isolation, meaning that once a LockedFile
is active, all operations of this LockedFile
are guaranteed to happen sequentially on the underlying file without being interleaved with operations from other LockedFile
.
Secured write operation
For performance reasons, write (and read) operations are done in memory. Periodically, the results of those operation are asynchronously flushed to the device storage area. If for some reason a problem occurs before that, you can lose the results of some operations. To avoid that problem, you can force the data to be flushed by using the LockedFile.flush()
method. Once this method has been successfully called, you can be sure your change on the file will be safe.
// Get a LockedFile object from the handle var myFile = myFileHandle.open('readwrite'); // Start a writing operation var writing = myFile.append('Some content'); writing.onsuccess = function () { console.log('Writing operation successful'); var saving = myFile.flush(); saving.onsuccess = function () { console.log('The file has been successfully stored'); } } writing.onerror = function () { console.log('Something goes wrong in the writing process: ' + this.error); }
API Compatibility
Why a different API than FileWriter?
The FileWriter specification defines FileWriters, objects aiming at representing editable files. Discussions on public-webapps led to the conclusion that the API would behave poorly in the case of different entities writing concurrently to the same file. The outcome of this discussion is the FileHandle API with its LockedFile and transaction mechanism.
Specifications
A formal specification draft is being written. As it does not fully match the current implementation, be warned that the implementation and/or the specification will be subject to changes.
Specification | Status | Comment |
---|---|---|
FileSystem API | Editor's Draft | Draft proposal |
Browser compatibility
Feature | Chrome | Firefox (Gecko) | Internet Explorer | Opera | Safari |
---|---|---|---|---|---|
Basic support | Not supported | 15 | Not supported | Not supported | Not supported |
Feature | Android | Firefox Mobile (Gecko) | IE Mobile | Opera Mobile | Safari Mobile |
---|---|---|---|---|---|
Basic support | Not supported | 15 | Not supported | Not supported | Not supported |