Node module deep-dive: Buffer

Howdy there! Time for another installment of my Node module-deep dive series! This week, I’m diving into the Buffer object in Node. I’ll admit that when I opened up the file for an initial look-through I freaked out a little bit. It’s a whooping 1,599 lines of code (give or take some for the lines that consist of comments). But you know what? I’ve done enough of these read-throughs to not be intimidated so onwards I go.

Before I dive into the actual code, it might help to present a quick primer on Buffers. Buffers in Node make it possible for the developer to interact with streams of binary data, this is particularly useful for things like reading and writing to a file on the filesystem. If you’ve used functions in the fs module like fs.createReadStream or fs.createWriteStream, you’ve interacted with buffers. To give an example, here is a Buffer containing a representation of the word “Safia” in Node.

> Buffer.from("Safia")
<Buffer 53 61 66 69 61>

Alright! Time to get into the nitty-gritty and look at the code itself. As usual, here is a permalink to the particular version of the Buffer class that I’ll be looking through. I usually start my code read at the bottom of a file to determine what classes and APIs a particular module exposes. Here’s a look at what the Buffer module exports.

module.exports = exports = {
  Buffer,
  SlowBuffer,
  transcode,
  INSPECT_MAX_BYTES: 50,

  // Legacy
  kMaxLength,
  kStringMaxLength
};

So it looks like it exports two classes, a Buffer and a SlowBuffer. I’m not sure what the specification distinction between them is at the moment, besides the obvious difference that one of them is slow and the other is not. In addition to those class exports, it looks like the module also exports several constants and a function.

The first thing I wanted to do was figure out what a SlowBuffer was and why it existed in the codebase. I headed over to the documentation page on Buffer under the latest version of Node and discovered under the section for the SlowBuffer class that it was actually a deprecated. A SlowBuffer is actually a variant of a Buffer object that is unpooled. An unpooled Buffer is one in which memory has not been initialized for the Buffer instance.

Now that I understood that, I started to look through the code for the Buffer class itself. The class exposes a lot of functions, so I wanted to focus on the few that I used in my day-to-day development work.

First, I wanted to start off by taking a look at the Buffer.from method. Buffer.from allows the developer to create a Buffer object from a string, array, or another Buffer. The method definition requires that the developer provide a value, encodingOrOffeset, and length parameters. The second two parameters only apply if the value that the developer is passing is an array, in which case they represent the index of the first byte in the array that the Buffer object will expose and the total number of bytes in the Buffer object to expose. If the value is a string, the second parameter is the encoding of the string (UTF-8 or ASCII, for example).

Buffer.from = function from(value, encodingOrOffset, length) {

The first couple of lines of code in the function define what to do when the type of the value is a string or an array. The method invokes the fromString and fromArrayBuffer functions accordingly.

if (typeof value === 'string')
  return fromString(value, encodingOrOffset);

if (isAnyArrayBuffer(value))
  return fromArrayBuffer(value, encodingOrOffset, length);

I decided to look at the fromString function first. Its function definition requires a string and an encoding as explained above.

function fromString(string, encoding) {

The function starts by handling potential edge cases in the parameters provided by the developer. For example, if the user doesn’t provide a string or an encoding, the function returns an empty Buffer.

  if (typeof encoding !== 'string' || encoding.length === 0) {
    if (string.length === 0)
      return new FastBuffer();

If the developer doesn’t provide an encoding, the function falls back on UTF-8 as the default encoding. The length variable defines the number of bytes in the string assuming it is encoding in UTF-8.

encoding = 'utf8';
length = byteLengthUtf8(string);

The next if-statement checks to see if the length of the bytes in the string are longer than (Buffer.poolSize >>> 1). I was a little bit confused by the (Buffer.poolSize >>> 1) bit so I did some digging up on it. The value of Buffer.poolSize is 8 * 1024 or 8192 bytes. This number represents the number of bytes that the internal Buffer object utilizes. This value is then shifted 1 bit to the right using a zero-fill right shift. A zero-fill right shift differs from the “standard” right shift (>>) because it does not add in bits from the left as the bits are shifted rightward. As a result, every number that undergoes a zero-filling rightward shift is always a positive number. In essence, the if-statement determines if the string that the user is attempting to create a Buffer from will fit in the 8192 bytes that are pre-allocated in the Buffer by default. If so, it’ll load the string in accordingly.

return createFromString(string, encoding);

On the other hand, if the number of bytes in the string is greater than the number of bytes that are pre-allocated in a Buffer, it’ll go ahead and allocate more space for the string before storing it into the Buffer.

if (length > (poolSize - poolOffset))
  createPool();
var b = new FastBuffer(allocPool, poolOffset, length);
const actual = b.write(string, encoding);
if (actual !== length) {
  // byteLength() may overestimate. That's a rare case, though.
  b = new FastBuffer(allocPool, poolOffset, actual);
}
poolOffset += actual;
alignPool();
return b;

Next, I dived into the fromArrayBuffer function which is executed when the user passes an array buffer to Buffer.from. The function definition for the fromArrayBuffer function takes the array object, the byte offset, and the length of the array buffer.

function fromArrayBuffer(obj, byteOffset, length) {

The function starts by responding to potentially messy parameters passed to the function. It first checks to see if the user didn’t pass a byteOffset to the function, in which case it uses an offset of 0. In other cases, the function ensures that the byteOffset is a positive number.

if (byteOffset === undefined) {
  byteOffset = 0;
} else {
  byteOffset = +byteOffset;
  // check for NaN
  if (byteOffset !== byteOffset)
    byteOffset = 0;
}

The length of the buffer is defined as the length of the input buffer array minus the offset.

const maxLength = obj.byteLength - byteOffset;

If the byteOffset was larger than the length of the input buffer, then the function throws an error.

if (maxLength < 0)
    throw new errors.RangeError('ERR_BUFFER_OUT_OF_BOUNDS', 'offset');

Finally, the function executes some checks to ensure that the length of the new ArrayBuffer is a positive number within the bounds of the newly offset object.

if (length === undefined) {
  length = maxLength;
} else {
  // convert length to non-negative integer
  length = +length;
  // Check for NaN
  if (length !== length) {
    length = 0;
  } else if (length > 0) {
    if (length > maxLength)
      throw new errors.RangeError('ERR_BUFFER_OUT_OF_BOUNDS', 'length');
  } else {
    length = 0;
  }

Then the new Buffer is created using the modified byteOffset and length parameters from the old obj ArrayBuffer.

return new FastBuffer(obj, byteOffset, length);

Going back to the Buffer.from function, it does a few more validation checks to ensure that the value the user is attempting to create a Buffer from is valid.

if (value === null || value === undefined) {
  throw new errors.TypeError(
    'ERR_INVALID_ARG_TYPE',
    'first argument',
    ['string', 'Buffer', 'ArrayBuffer', 'Array', 'Array-like Object'],
    value
  );
}

if (typeof value === 'number')
  throw new errors.TypeError(
    'ERR_INVALID_ARG_TYPE', 'value', 'not number', value
  );

Then the function checks to see if the value passed by the user contains a valueOf function. The valueOf function is defined on the Object prototype in JavaScript and returns a value of a primitive type for a specific object in JavaScript. For example, a developer might create a special Cost object that stores the price of an object and create a valueOf function that returns the price as a Number (which is floating point). In a sense, this bit of the Buffer.from method attempts to extract a primitive type out of any object passed as a value to the function and uses it to generate a new Buffer.

const valueOf = value.valueOf && value.valueOf();
if (valueOf !== null && valueOf !== undefined && valueOf !== value)
  return Buffer.from(valueOf, encodingOrOffset, length);

Then the function attempts to invoke the fromObject function and returns the buffer created by this function (assuming it is non-null).

var b = fromObject(value);
if (b)
  return b;

The next check evaluates if the value passed has a toPrimitive function defined. The toPrimitive function returns a primitive value from a given JavaScript object. The Buffer.from function attempts to create a Buffer from the primitive returned by this function if it is available.

if (typeof value[Symbol.toPrimitive] === 'function') {
  return Buffer.from(value[Symbol.toPrimitive]('string'),
                     encodingOrOffset,
                     length);
}

In all other cases, the function raises a TypeError.

throw new errors.TypeError(
  'ERR_INVALID_ARG_TYPE',
  'first argument',
  ['string', 'Buffer', 'ArrayBuffer', 'Array', 'Array-like Object'],
  value
);

So in essence, the Buffer.from function will attempt to process values that are strings or ArrayBuffers then attempt to process values that are Array-like then attempt to extract a primitive value to create a Buffer from then emit a TypeError to the user in all other cases.

The next function on the Buffer object that I wanted to read through was the write function. The function definition for the Buffer.write function requires that the developer pass the string to write, the number of bytes to skip before writing the string as given by the offset, the number of bytes to write as given by length, and the encoding of the string.

Buffer.prototype.write = function write(string, offset, length, encoding) {

If no offset is given, the function writes the string at the start of the Buffer.

if (offset === undefined) {
  return this.utf8Write(string, 0, this.length);
}

If no offset or length is given, the function starts at an offset of 0 and uses the default length of the Buffer.

// Buffer#write(string, encoding)
} else if (length === undefined && typeof offset === 'string') {
  encoding = offset;
  length = this.length;
  offset = 0;
}

Finally, if the developer provides both an offset and a length, the function ensures that they are valid finite values and calculates the length properly if an offset was given.

} else if (isFinite(offset)) {
  offset = offset >>> 0;
  if (isFinite(length)) {
    length = length >>> 0;
  } else {
    encoding = length;
    length = undefined;
  }

  var remaining = this.length - offset;
  if (length === undefined || length > remaining)
    length = remaining;

  if (string.length > 0 && (length < 0 || offset < 0))
    throw new errors.RangeError('ERR_BUFFER_OUT_OF_BOUNDS', 'length', true);
}

In all other cases, the function assumes that the developer is attempting to use an outdated version of the Buffer.write API and raises an error.

 else {
   // if someone is still calling the obsolete form of write(), tell them.
   // we don't want eg buf.write("foo", "utf8", 10) to silently turn into
   // buf.write("foo", "utf8"), so we can't ignore extra args
   throw new errors.Error(
     'ERR_NO_LONGER_SUPPORTED',
     'Buffer.write(string, encoding, offset[, length])'
   );
 }

Once the function has set the offset and length variables appropriately, it determines what to do depending on the different possible encodings. If no encoding is given, the Buffer.write method assumes UTF-8 by default.

if (!encoding) return this.utf8Write(string, offset, length);

In other cases, the function invokes the appropriate xWrite function where x is an encoding. I found it interesting that the switch statement used to evaluate the potential encodings checked the length of the encoding string then checked the actual value of encoding. In essence, the function evaluates the situation where the encoding is utf8 and utf-8 in different branches of the switch statement.

  switch (encoding.length) {
    case 4: ...
    case 5: ...
    case 7: ...
    case 8: ...
    case 6: ...
    case 3: ...
  }

There’s a few more interesting functions that I was hoping to read through in the Buffer class but I might end up putting those in a part 2 of this blog post. For now, I’ll stop here. If you have any questions or comments about the above, feel free to ask me a question or reach out to me on Twitter.