EBML stands for Extensible Binary Meta-Language and is somewhat of a binary version of XML. It's used for container formats like WebM or MKV.
this version fixes just the encoder
This is for version 3.0.0 and up, which has undergone a massive rewrite and
now builds with RollupJS.
Version 2.2.4 is the last version to have guaranteed legacy semantics.
Install via NPM or Yarn:
npm install ebml --save
# or
yarn add ebmlThe Decoder() class is implemented as a Node Transform stream.
As input it takes EBML. As output it emits a sequence of chunks: two-element
arrays looking like this example.
[
'tag',
{
name: 'TimecodeScale',
type: 'u',
value: 1000000,
},
];The first element of the array is a short text string. For tags containing
values, like this example, the string is 'tag'. ebml also has nesting tags.
The opening of those tags has the string 'start' and the closing has the
string 'end'. Integers stored in 6 bytes or less are represented as numbers,
and longer integers are represented as hexadecimal text strings.
The second element of the array is an object with these members, among others:
nameis the Matroska Element Name.typeis the data type.u: unsigned integer. Some of these are UIDs, coded as 128-bit numbers.i: signed integer.f: IEEE-754 floating point number.s: printable ASCII text string.8: printable utf-8 Unicode text string.d: a 64-bit signed timestamp, in nanoseconds after (or before)2001-01-01T00:00UTC.bbinary data, otherwise uninterpreted.
valueis the value of the data in the element, represented as a number or a string.datais the binary data of the entire element stored in aUint8Array.
Elements with the Block and SimpleBlock types
get special treatment. They have these additional members:
payloadis the coded information in the element, stored in aUint8Array.trackis an unsigned integer indicating the payload's track.keyframeis a Boolean value set to true if the payload starts an I frame (SimpleBlocksonly).discardableis a Boolean value showing the value of the element's Discardable flag. (SimpleBlocksonly).
And the value member shows the block's Timecode value.
This example reads a media file into memory and decodes it. The decoder
invokes its data event for each Element.
const fs = require('fs');
const { Decoder } = require('./lib/ebml.js');
const decoder = new Decoder();
decoder.on('data', (chunk) => console.log(chunk));
fs.readFile('media/test.webm', (err, data) => {
if (err) {
throw err;
}
decoder.write(data);
});This example does the same thing, but by piping the file stream into the decoder (a Transform stream).
const { Decoder } = require('./lib/ebml.js');
const ebmlDecoder = new Decoder();
const counts = {};
require('fs')
.createReadStream('media/test.webm')
.pipe(ebmlDecoder)
.on('data', (chunk) => {
const { name } = chunk[1];
if (!counts[name]) {
counts[name] = 0;
}
counts[name] += 1;
})
.on('finish', () => console.log(counts));Parsing should work. If it doesn't, please create an issue.
d-type elements (timestamps) are not yet decoded to Javascript timestamp
values.
Thanks to @chrisprice we got an encoder!
(in alphabetical order)