The library is designed to introduce fault-tolerance into the upload of large files through HTTP. This is done by splitting each file into small chunks. Then, whenever the upload of a chunk fails, uploading is retried until the procedure completes. This allows uploads to automatically resume uploading after a network connection is lost either locally or to the server. Additionally, it allows for users to pause, resume and even recover uploads without losing state because only the currently uploading chunks will be aborted, not the entire upload.
Flow.js does not have any external dependencies other than the
HTML5 File API. This is relied on for the ability to chunk files into smaller pieces. Currently, this means that support is limited to Firefox 4+, Chrome 11+, Safari 6+ and Internet Explorer 10+.
Samples and examples are available in the
samples/ folder. Please push your own as Markdown to help document the project.
Can I see a demo?
JQuery and node.js backend demo https://github.com/flowjs/flow.js/tree/master/samples/Node.js
How can I install it?
Download a latest build from https://github.com/flowjs/flow.js/releases
it contains development and minified production files in
or use bower:
bower install flow.js#~2
or use git clone
How can I use it?
Flow object is created with information of what and where to post:
var flow =target:'/api/photo/redeem-upload-token'query:upload_token:'my_token';// Flow.js isn't supported, fall back on a different methodif!flowsupport locationhref = '/some-old-crappy-uploader';
To allow files to be either selected and drag-dropped, you'll assign drop target and a DOM item to be clicked for browsing:
After this, interaction with Flow.js is done by listening to events:
How do I set it up with my server?
Most of the magic for Flow.js happens in the user's browser, but files still need to be reassembled from chunks on the server side. This should be a fairly simple task and can be achieved in any web framework or language, which is able to receive file uploads.
To handle the state of upload chunks, a number of extra parameters are sent along with all requests:
flowChunkNumber: The index of the chunk in the current upload. First chunk is
1(no base-0 counting here).
flowTotalChunks: The total number of chunks.
flowChunkSize: The general chunk size. Using this value and
flowTotalSizeyou can calculate the total number of chunks. Please note that the size of the data received in the HTTP might be lower than
flowChunkSizeof this for the last chunk for a file.
flowTotalSize: The total file size.
flowIdentifier: A unique identifier for the file contained in the request.
flowFilename: The original file name (since a bug in Firefox results in the file name not being transmitted in chunk multipart posts).
flowRelativePath: The file's relative path when selecting a directory (defaults to file name in all browsers except Chrome).
You should allow for the same chunk to be uploaded more than once; this isn't standard behaviour, but on an unstable network environment it could happen, and this case is exactly what Flow.js is designed for.
For every request, you can confirm reception in HTTP status codes (can be change through the
202: The chunk was accepted and correct. No need to re-upload.
501: The file for which the chunk was uploaded is not supported, cancel the entire upload.
- Anything else: Something went wrong, but try reuploading the file.
Handling GET (or
testChunks option will allow uploads to be resumed after browser restarts and even across browsers (in theory you could even run the same file upload across multiple tabs or different browsers). The
POST data requests listed are required to use Flow.js to receive data, but you can extend support by implementing a corresponding
GET request with the same parameters:
- If this request returns a
202HTTP code, the chunks is assumed to have been completed.
- If request returns a permanent error status, upload is stopped.
- If request returns anything else, the chunk will be uploaded in the standard fashion.
After this is done and
testChunks enabled, an upload can quickly catch up even after a browser restart by simply verifying already uploaded chunks that do not need to be uploaded again.
The object is loaded with a configuration options:
var r = opt1:'val' ...;
Available configuration options are:
targetThe target URL for the multipart POST request. This can be a string or a function. If a function, it will be passed a FlowFile, a FlowChunk and isTest boolean (Default:
singleFileEnable single file upload. Once one file is uploaded, second file will overtake existing one, first one will be canceled. (Default: false)
chunkSizeThe size in bytes of each uploaded chunk of data. The last uploaded chunk will be at least this size and up to two the size, see Issue #51 for details and reasons. (Default:
forceChunkSizeForce all chunks to be less or equal than chunkSize. Otherwise, the last chunk will be greater than or equal to
simultaneousUploadsNumber of simultaneous uploads (Default:
fileParameterNameThe name of the multipart POST parameter to use for the file chunk (Default:
queryExtra parameters to include in the multipart POST with data. This can be an object or a function. If a function, it will be passed a FlowFile, a FlowChunk object and a isTest boolean (Default:
headersExtra headers to include in the multipart POST with data. If a function, it will be passed a FlowFile, a FlowChunk object and a isTest boolean (Default:
withCredentialsStandard CORS requests do not send or set any cookies by default. In order to include cookies as part of the request, you need to set the
withCredentialsproperty to true. (Default:
methodMethod to use when POSTing chunks to the server (
testMethodHTTP method to use when chunks are being tested. If set to a function, it will be passed a FlowFile and a FlowChunk arguments. (Default:
uploadMethodHTTP method to use when chunks are being uploaded. If set to a function, it will be passed a FlowFile and a FlowChunk arguments. (Default:
allowDuplicateUploadsOnce a file is uploaded, allow reupload of the same file. By default, if a file is already uploaded, it will be skipped unless the file is removed from the existing Flow object. (Default:
prioritizeFirstAndLastChunkPrioritize first and last chunks of all files. This can be handy if you can determine if a file is valid for your service from only the first or last chunk. For example, photo or video meta data is usually located in the first part of a file, making it easy to test support from only the first chunk. (Default:
testChunksMake a GET request to the server for each chunks to see if it already exists. If implemented on the server-side, this will allow for upload resumes even after a browser crash or even a computer restart. (Default:
preprocessOptional function to process each chunk before testing & sending. To the function it will be passed the chunk as parameter, and should call the
preprocessFinishedmethod on the chunk when finished. (Default:
initFileFnOptional function to initialize the fileObject. To the function it will be passed a FlowFile and a FlowChunk arguments.
readFileFnOptional function wrapping reading operation from the original file. To the function it will be passed the FlowFile, the startByte and endByte, the fileType and the FlowChunk.
generateUniqueIdentifierOverride the function that generates unique identifiers for each file. (Default:
maxChunkRetriesThe maximum number of retries for a chunk before the upload is failed. Valid values are any positive integer and
undefinedfor no limit. (Default:
chunkRetryIntervalThe number of milliseconds to wait before retrying a chunk on a non-permanent error. Valid values are any positive integer and
undefinedfor immediate retry. (Default:
progressCallbacksIntervalThe time interval in milliseconds between progress reports. Set it to 0 to handle each progress callback. (Default:
speedSmoothingFactorUsed for calculating average upload speed. Number from 1 to 0. Set to 1 and average upload speed wil be equal to current upload speed. For longer file uploads it is better set this number to 0.02, because time remaining estimation will be more accurate. This parameter must be adjusted together with
progressCallbacksIntervalparameter. (Default 0.1)
successStatusesResponse is success if response status is in this list (Default:
permanentErrorsResponse fails if response status is in this list (Default:
[404, 415, 500, 501])
.supportA boolean value indicator whether or not Flow.js is supported by the current browser.
.supportDirectoryA boolean value, which indicates if browser supports directory uploads.
.optsA hash object of the configuration of the Flow.js instance.
.filesAn array of
FlowFilefile objects added by the user (see full docs for this object type below).
.assignBrowse(domNodes, isDirectory, singleFile, attributes)Assign a browse action to one or more DOM nodes.
domNodesarray of dom nodes or a single node.
trueto allow directories to be selected (Chrome only, support can be checked with
singleFileTo prevent multiple file uploads set this to true. Also look at config parameter
attributesPass object of keys and values to set custom attributes on input fields. For example, you can set
image/*. This means that user will be able to select only images. Full list of attributes: https://www.w3.org/wiki/HTML/Elements/input/file
Note: avoid using
buttontags as file upload buttons, use span instead.
.assignDrop(domNodes)Assign one or more DOM nodes as a drop target.
.unAssignDrop(domNodes)Unassign one or more DOM nodes as a drop target.
.on(event, callback)Listen for event from Flow.js (see below)
.off()All events are removed.
.off(event)Remove all callbacks of specific event.
.off(event, callback)Remove specific callback of event.
callbackshould be a
.upload()Start or resume uploading.
.cancel()Cancel upload of all
FlowFileobjects and remove them from the list.
.progress()Returns a float between 0 and 1 indicating the current upload progress of all files.
.isUploading()Returns a boolean indicating whether or not the instance is currently uploading anything.
.addFile(file)Add a HTML5 File object to the list of files.
.removeFile(file)Cancel upload of a specific
FlowFileobject on the list from the list.
.getFromUniqueIdentifier(uniqueIdentifier)Look up a
FlowFileobject by its unique identifier.
.getSize()Returns the total size of the upload in bytes.
.sizeUploaded()Returns the total size uploaded of all files in bytes.
.timeRemaining()Returns remaining time to upload all files in seconds. Accuracy is based on average speed. If speed is zero, time remaining will be equal to positive infinity
.fileSuccess(file, message, chunk)A specific file was completed. First argument
fileis instance of
FlowFile, second argument
messagecontains server response. Response is always a string. Third argument
chunkis instance of
FlowChunk. You can get response status by accessing xhr object
.fileProgress(file, chunk)Uploading progressed for a specific file.
.fileAdded(file, event)This event is used for file validation. To reject this file return false. This event is also called before file is added to upload queue, this means that calling
flow.upload()function will not start current file upload. Optionally, you can use the browser
eventobject from when the file was added.
.filesAdded(array, event)Same as fileAdded, but used for multiple file validation.
.filesSubmitted(array, event)Same as filesAdded, but happens after the file is added to upload queue. Can be used to start upload of currently added files.
.fileRemoved(file)The specific file was removed from the upload queue. Combined with filesSubmitted, can be used to notify UI to update its state to match the upload queue.
.fileRetry(file, chunk)Something went wrong during upload of a specific file, uploading is being retried.
.fileError(file, message, chunk)An error occurred during upload of a specific file.
.uploadStart()Upload has been started on the Flow object.
.error(message, file, chunk)An error, including fileError, occurred.
.catchAll(event, ...)Listen to all the events listed above with the same callback function.
FlowFile constructor can be accessed in
.flowObjA back-reference to the parent
.fileThe correlating HTML5
.nameThe name of the file.
.relativePathThe relative path to the file (defaults to file name if relative path doesn't exist)
.sizeSize in bytes of the file.
.uniqueIdentifierA unique identifier assigned to this file object. This value is included in uploads to the server for reference, but can also be used in CSS classes etc when building your upload UI.
.averageSpeedAverage upload speed, bytes per second.
.currentSpeedCurrent upload speed, bytes per second.
.chunksAn array of
FlowChunkitems. You shouldn't need to dig into these.
.pausedIndicated if file is paused.
.errorIndicated if file has encountered an error.
.progress(relative)Returns a float between 0 and 1 indicating the current upload progress of the file. If
true, the value is returned relative to all files in the Flow.js instance.
.pause()Pause uploading the file.
.resume()Resume uploading the file.
.cancel()Abort uploading the file and delete it from the list of files to upload.
.retry()Retry uploading the file.
.bootstrap()Rebuild the state of a
FlowFileobject, including reassigning chunks and XMLHttpRequest instances.
.isUploading()Returns a boolean indicating whether file chunks is uploading.
.isComplete()Returns a boolean indicating whether the file has completed uploading and received a server response.
.sizeUploaded()Returns size uploaded in bytes.
.timeRemaining()Returns remaining time to finish upload file in seconds. Accuracy is based on average speed. If speed is zero, time remaining will be equal to positive infinity
.getExtension()Returns file extension in lowercase.
.getType()Returns file type.
To ensure consistency throughout the source code, keep these rules in mind as you are working:
All features or bug fixes must be tested by one or more specs.
- To clone your Github repository, run:
git clone firstname.lastname@example.org:<github username>/flow.js.git
- To go to the Flow.js directory, run:
- To add node.js dependencies
Our unit and integration tests are written with Jasmine and executed with Karma. To run all of the tests on Chrome run:
Or choose other browser
grunt karma:watch --browsers=Firefox,Chrome
Browsers should be comma separated and case sensitive.
To re-run tests just change any source or test file.
Automated tests is running after every commit at travis-ci.
Running test on sauceLabs
- Connect to sauce labs https://saucelabs.com/docs/connect
grunt test --sauce-local=true --sauce-username=**** --sauce-access-key=***
other browsers can be used with
--browsers flag, available browsers: sl_opera,sl_iphone,sl_safari,sl_ie10,sl_chrome,sl_firefox
Flow.js was inspired by and evolved from https://github.com/23/resumable.js. Library has been supplemented with tests and features, such as drag and drop for folders, upload speed, time remaining estimation, separate files pause, resume and more.