ArrayBuffer, binary arrays

In web-development we meet binary data mostly while dealing with files (create, upload, download). Another typical use case is image processing.

Thats all possible in JavaScript, and binary operations are high-performant.

Although, theres a bit of confusion, because there are many classes. To name a few:

• ArrayBuffer, Uint8Array, DataView, Blob, File, etc.

Binary data in JavaScript is implemented in a non-standard way, compared to other languages. But when we sort things out, everything becomes fairly simple.

The basic binary object is ArrayBuffer a reference to a fixed-length contiguous memory area.

We create it like this:

let buffer = new ArrayBuffer(16); // create a buffer of length 16
alert(buffer.byteLength); // 16

This allocates a contiguous memory area of 16 bytes and pre-fills it with zeroes.

ArrayBuffer is a memory area. Whats stored in it? It has no clue. Just a raw sequence of bytes.

To manipulate an ArrayBuffer, we need to use a view object.

A view object does not store anything on its own. Its the eyeglasses that give an interpretation of the bytes stored in the ArrayBuffer.

For instance:

• Uint8Array treats each byte in ArrayBuffer as a separate number, with possible values from 0 to 255 (a byte is 8-bit, so it can hold only that much). Such value is called a 8-bit unsigned integer.
• Uint16Array treats every 2 bytes as an integer, with possible values from 0 to 65535. Thats called a 16-bit unsigned integer.
• Uint32Array treats every 4 bytes as an integer, with possible values from 0 to 4294967295. Thats called a 32-bit unsigned integer.
• Float64Array treats every 8 bytes as a floating point number with possible values from 5.0x10-324 to 1.8x10308.

So, the binary data in an ArrayBuffer of 16 bytes can be interpreted as 16 tiny numbers, or 8 bigger numbers (2 bytes each), or 4 even bigger (4 bytes each), or 2 floating-point values with high precision (8 bytes each).

[]

ArrayBuffer is the core object, the root of everything, the raw binary data.

But if were going to write into it, or iterate over it, basically for almost any operation we must use a view, e.g:

let buffer = new ArrayBuffer(16); // create a buffer of length 16

let view = new Uint32Array(buffer); // treat buffer as a sequence of 32-bit integers

alert(Uint32Array.BYTES_PER_ELEMENT); // 4 bytes per integer

alert(view.length); // 4, it stores that many integers
alert(view.byteLength); // 16, the size in bytes

// let's write a value
view[0] = 123456;

// iterate over values
for(let num of view) {
  alert(num); // 123456, then 0, 0, 0 (4 values total)
}

TypedArray

The common term for all these views (Uint8Array, Uint32Array, etc) is TypedArray. They share the same set of methods and properities.

Please note, theres no constructor called TypedArray, its just a common umbrella term to represent one of views over ArrayBuffer: Int8Array, Uint8Array and so on, the full list will soon follow.

When you see something like new TypedArray, it means any of new Int8Array, new Uint8Array, etc.

Typed arrays behave like regular arrays: have indexes and are iterable.

A typed array constructor (be it Int8Array or Float64Array, doesnt matter) behaves differently depending on argument types.

There are 5 variants of arguments:

new TypedArray(buffer, [byteOffset], [length]);
new TypedArray(object);
new TypedArray(typedArray);
new TypedArray(length);
new TypedArray();

1. If an ArrayBuffer argument is supplied, the view is created over it. We used that syntax already.

   Optionally we can provide byteOffset to start from (0 by default) and the length (till the end of the buffer by default), then the view will cover only a part of the buffer.

2. If an Array, or any array-like object is given, it creates a typed array of the same length and copies the content.

   We can use it to pre-fill the array with the data:

   let arr = new Uint8Array([0, 1, 2, 3]);
   alert( arr.length ); // 4, created binary array of the same length
   alert( arr[1] ); // 1, filled with 4 bytes (unsigned 8-bit integers) with given values

3. If another TypedArray is supplied, it does the same: creates a typed array of the same length and copies values. Values are converted to the new type in the process, if needed.

   let arr16 = new Uint16Array([1, 1000]);
   let arr8 = new Uint8Array(arr16);
   alert( arr8[0] ); // 1
   alert( arr8[1] ); // 232, tried to copy 1000, but can't fit 1000 into 8 bits (explanations below)

4. For a numeric argument length creates the typed array to contain that many elements. Its byte length will be length multiplied by the number of bytes in a single item TypedArray.BYTES_PER_ELEMENT:

   let arr = new Uint16Array(4); // create typed array for 4 integers
   alert( Uint16Array.BYTES_PER_ELEMENT ); // 2 bytes per integer
   alert( arr.byteLength ); // 8 (size in bytes)

5. Without arguments, creates an zero-length typed array.

We can create a TypedArray directly, without mentioning ArrayBuffer. But a view cannot exist without an underlying ArrayBuffer, so gets created automatically in all these cases except the first one (when provided).

To access the ArrayBuffer, there are properties:

• arr.buffer references the ArrayBuffer.
• arr.byteLength the length of the ArrayBuffer.

So, we can always move from one view to another:

let arr8 = new Uint8Array([0, 1, 2, 3]);

// another view on the same data
let arr16 = new Uint16Array(arr8.buffer);

Heres the list of typed arrays:

• Uint8Array, Uint16Array, Uint32Array for integer numbers of 8, 16 and 32 bits.
  • Uint8ClampedArray for 8-bit integers, clamps them on assignment (see below).
• Int8Array, Int16Array, Int32Array for signed integer numbers (can be negative).
• Float32Array, Float64Array for signed floating-point numbers of 32 and 64 bits.

Out-of-bounds behavior

What if we attempt to write an out-of-bounds value into a typed array? There will be no error. But extra bits are cut-off.

For instance, lets try to put 256 into Uint8Array. In binary form, 256 is 100000000 (9 bits), but Uint8Array only provides 8 bits per value, that makes the available range from 0 to 255.

For bigger numbers, only the rightmost (less significant) 8 bits are stored, and the rest is cut off:

[]

So well get zero.

For 257, the binary form is 100000001 (9 bits), the rightmost 8 get stored, so well have 1 in the array:

[]

In other words, the number modulo 28 is saved.

Heres the demo:

let uint8array = new Uint8Array(16);

let num = 256;
alert(num.toString(2)); // 100000000 (binary representation)

uint8array[0] = 256;
uint8array[1] = 257;

alert(uint8array[0]); // 0
alert(uint8array[1]); // 1

Uint8ClampedArray is special in this aspect, its behavior is different. It saves 255 for any number that is greater than 255, and 0 for any negative number. That behavior is useful for image processing.

TypedArray methods

TypedArray has regular Array methods, with notable exceptions.

We can iterate, map, slice, find, reduce etc.

There are few things we cant do though:

• No splice we cant delete a value, because typed arrays are views on a buffer, and these are fixed, contiguous areas of memory. All we can do is to assign a zero.
• No concat method.

There are two additional methods:

• arr.set(fromArr, [offset]) copies all elements from fromArr to the arr, starting at position offset (0 by default).
• arr.subarray([begin, end]) creates a new view of the same type from begin to end (exclusive). Thats similar to slice method (thats also supported), but doesnt copy anything just creates a new view, to operate on the given piece of data.

These methods allow us to copy typed arrays, mix them, create new arrays from existing ones, and so on.

DataView

DataView is a special super-flexible untyped view over ArrayBuffer. It allows to access the data on any offset in any format.

• For typed arrays, the constructor dictates what the format is. The whole array is supposed to be uniform. The i-th number is arr[i].
• With DataView we access the data with methods like .getUint8(i) or .getUint16(i). We choose the format at method call time instead of the construction time.

The syntax:

new DataView(buffer, [byteOffset], [byteLength])

• buffer the underlying ArrayBuffer. Unlike typed arrays, DataView doesnt create a buffer on its own. We need to have it ready.
• byteOffset the starting byte position of the view (by default 0).
• byteLength the byte length of the view (by default till the end of buffer).

For instance, here we extract numbers in different formats from the same buffer:

// binary array of 4 bytes, all have the maximal value 255
let buffer = new Uint8Array([255, 255, 255, 255]).buffer;

let dataView = new DataView(buffer);

// get 8-bit number at offset 0
alert( dataView.getUint8(0) ); // 255

// now get 16-bit number at offset 0, it consists of 2 bytes, together interpreted as 65535
alert( dataView.getUint16(0) ); // 65535 (biggest 16-bit unsigned int)

// get 32-bit number at offset 0
alert( dataView.getUint32(0) ); // 4294967295 (biggest 32-bit unsigned int)

dataView.setUint32(0, 0); // set 4-byte number to zero, thus setting all bytes to 0

DataView is great when we store mixed-format data in the same buffer. For example, when we store a sequence of pairs (16-bit integer, 32-bit float), DataView allows to access them easily.

Summary

ArrayBuffer is the core object, a reference to the fixed-length contiguous memory area.

To do almost any operation on ArrayBuffer, we need a view.

• It can be a TypedArray:
  • Uint8Array, Uint16Array, Uint32Array for unsigned integers of 8, 16, and 32 bits.
  • Uint8ClampedArray for 8-bit integers, clamps them on assignment.
  • Int8Array, Int16Array, Int32Array for signed integer numbers (can be negative).
  • Float32Array, Float64Array for signed floating-point numbers of 32 and 64 bits.
• Or a DataView the view that uses methods to specify a format, e.g. getUint8(offset).

In most cases we create and operate directly on typed arrays, leaving ArrayBuffer under cover, as a common denominator. We can access it as .buffer and make another view if needed.

There are also two additional terms, that are used in descriptions of methods that operate on binary data:

• ArrayBufferView is an umbrella term for all these kinds of views.
• BufferSource is an umbrella term for ArrayBuffer or ArrayBufferView.

Well see these terms in the next chapters. BufferSource is one of the most common terms, as it means any kind of binary data an ArrayBuffer or a view over it.

Heres a cheatsheet:

[]