codem-isoboxer

• https://github.com/madebyhiro/codem-isoboxer

Description

codem-isoboxer is a lightweight JavaScript MP4 (MPEG-4, ISOBMFF) parser. It is meant to be small, fast and efficient. A typical use-case would be inclusion in a new player framework (for emerging standards such as MPEG-DASH which rely on ISOBMFF for most situations, or HLS using fragmented MP4) or to extract metadata from MPEG-4 files:

• Parsing emsg boxes for in-band events;
• Parsing mdat boxes for extracting subtitles;
• Validating ISOBMFF segments before playing them back;
• [etc.]

Currently a limited set of ISOBMFF boxes is supported:

ISO/IEC 14496-12:2012 (ISOBMFF)

• ardi
• dinf
• edts
• elst
• elng
• free / skip
• ftyp / styp
• grpl
• hdlr
• imda
• kind
• labl
• mdat
• mdia
• mdhd
• meco
• mehd
• meta
• mfhd
• mfra
• mfro
• minf
• moov / moof
• mp4a / enca
• mvex
• mvhd / mfhd
• prsl
• sidx
• ssix
• stbl
• stsd
• strk
• subs
• tfhd / tkhd
• tfdt
• tfra
• traf / trak
• tref
• trex
• trun
• udta

ISO/IEC 23009-1:2014 (MPEG-DASH)

• emsg

ISO/IEC 14496-30:2014 (Timed text and other visual overlays in ISOBMFF)

• vttC
• vttc
• vtte
• vlab
• payl
• sttg

ISO/IEC 14496-15:2014 (Carriage of network abstraction layer (NAL) unit structured video in ISO base media file format)

• avc1/2/3/4 / hev1 / hvc1 / encv

Support for more boxes can easily be added by adding additional box parsers in src/parsers. Some utility functions are included to help with reading the various ISOBMFF data types from the raw file. Also, see the Box Support page on the Wiki for a full list.

Requirements

A modern web browser with support for:

• ArrayBuffer
• DataView
• TextDecoder (optional)

Usage

Include one of the files in the dist folder (regular or minified) in your web page/application:

Then, you can parse a file by calling the parseBuffer function:

var parsedFile = ISOBoxer.parseBuffer(arrayBuffer);
console.log(parsedFile.boxes);

The arrayBuffer can be obtained for example by issuing an XMLHttpRequest with responsetype set to arrayBuffer, or by using the FileReader API to read a local file.

codem-isoboxer makes no assumptions on the validity of the given file and/or segment. It also does minimal handling of the data types and provides mostly a raw interface. Some frequently used attributes are parsed to easier-to-use types, such as the major brand and list of compatible brands in the ftyp box.

For traversing the box structure you can use the _parent property. It returns exactly what you expect: the parent of the current box. The opposite is the boxes property (only available on container boxes such as moov), which gives you its children. Every box also has a _root property which returns the top-level (file) container.

Another way to use the software is to only retrieve the boxes you are interested in. This way you don't have to traverse the box structure yourself:

var parsedFile = ISOBoxer.parseBuffer(arrayBuffer); // Parse the file
var ftyp       = parsedFile.fetch('ftyp');          // Fetch the first box with the specified type (`ftyp`)
var mdats      = parsedFile.fetchAll('mdat');       // Fetch all the boxes with the specified type (`mdat`)

Traversal of the box structure is always depth first. Note that while you jump directly to a box using the fetch commands, the entire parsed structure is still available, so the _parent, _root and boxes properties are still available.

An additional utility method is included to convert DataViews into strings. This uses the TextDecoder interface is available, otherwise it falls back to a naïve implementation (bytes to character codes). If the TextDecoder interface is available you can supply an additional encoding parameter (defaults to utf-8) to the function.

var parsedFile = ISOBoxer.parseBuffer(arrayBuffer); // Parse the file
var mdat       = parsedFile.fetch('mdat');          // Get the first 'mdat' box
var text       = ISOBoxer.Utils.dataViewToString(mdat.data); // Convert the data into a string (e.g. captions)

Basic support for incomplete buffers is also available. Boxes and containers that cannot be fully parsed (due to an empty buffer) will be marked with an _incomplete property. You can simply add new data to the ArrayBuffer when it becomes available and re-parse the buffer.

NodeJS

Does it work in NodeJS? Well, it's really meant to be run in a web browser, but since Node supports most features it shouldn't be a problem. You can install it using NPM:

npm install codem-isoboxer

Then use it in your code (NodeJS v0.10.36 tested, 0.12.x should work as well):

var ISOBoxer    = require('codem-isoboxer'),
    fs          = require('fs');
var arrayBuffer = new Uint8Array(fs.readFileSync('my_test_file.mp4')).buffer;
var parsedFile  = ISOBoxer.parseBuffer(arrayBuffer);

Et voila. It does not support any of the fancy stream stuff from Node. Also, the Node console has some issues with printing objects that contain buffers/dataviews/circular references. It might not look pretty in the console, but it works.

Development

Check out the source from Github. Make changes to the files in /src (not in /dist). We use grunt to build the distribution files. If you add a box parser be sure to include a comment that points toward the relevant section in the specs. And if at all possible add a (small!) file to test/fixtures to provide an example. The test directory is not included in the published package on npmjs.org because of the included MPEG-4 files.

Building

grunt

Using grunt watcher

You can use grunt-contrib-watch to watch for changes to the source and automatically build it:

grunt watch

Tests

Included is a small set of tests which check the code against a known set of ISOBMFF files. The tests use NodeJS with jasmine-node. Make sure you have the grunt watcher running, as it tests against the resulting build of the JavaScript files. Usage:

# npm install jasmine-node -g
# jasmine-node test/spec

When adding new parsers please consider adding an ISOBMFF test file and a relevant test spec.

Please note, for size concerns, tests are only included in the Github repository and not in the released packages on NPM.

Advanced build options

Writing functionality

By default, codem-isoboxer build file contains all the code that manages the parsing AND the writing functionalities. If you are only interested in the parsing functionality, you can generate a build without writing functionality. The syntax for building is:

grunt --no-writing

Modular build

codem-isoboxer now has the option for modular builds. This means you can specify which boxes you are interested in during build time and you will get a generated file containing only the necessary boxes. This can help you further decrease the size of the library if you know you will only need access to some boxes. The syntax for building is:

grunt --boxes=moov,mdat

This will generate a build that only contains the code to parse these boxes and can yield significantly smaller builds. The list needs to be comma-separated. Be sure not to include any white-space in it. Note that some parsers share identical code (e.g. ftyp/styp, free/skip and all the regular container boxes). Including one of those will automatically include the other ones as well, but at no additional build size. See src/parsers for a list of available parsers and see which parsers share code.

codem-isoboxer does not take into account the box hierarchy/dependency when building, you must explicitly specify which boxes you need (e.g. if you want to parse mvhd you must also include moov). Boxes that are not included don't have their properties set, but you can still access their type and size properties and the raw data (._raw).

Demo

Open test/index.html in your browser. Use the file picker to select a local MPEG-4 file to parse it. Results will be in the window.parsedFile variable. Inspect it from your browser's console. Some test files are included in test/fixtures (not included in the package published on npmjs.org).

License

codem-isoboxer is released under the MIT license, see LICENSE.txt.