Node.js v14.17.4 documentation

Threadpool usage and performance considerations#

All zlib APIs, except those that are explicitly synchronous, use the Node.js internal threadpool. This can lead to surprising effects and performance limitations in some applications.

Creating and using a large number of zlib objects simultaneously can cause significant memory fragmentation.

const zlib = require('zlib');

const payload = Buffer.from('This is some data');

// WARNING: DO NOT DO THIS!
for (let i = 0; i < 30000; ++i) {
  zlib.deflate(payload, (err, buffer) => {});
}

In the preceding example, 30,000 deflate instances are created concurrently. Because of how some operating systems handle memory allocation and deallocation, this may lead to to significant memory fragmentation.

It is strongly recommended that the results of compression operations be cached to avoid duplication of effort.

Compressing HTTP requests and responses#

The zlib module can be used to implement support for the gzip, deflate and br content-encoding mechanisms defined by HTTP.

The HTTP Accept-Encoding header is used within an http request to identify the compression encodings accepted by the client. The Content-Encoding header is used to identify the compression encodings actually applied to a message.

The examples given below are drastically simplified to show the basic concept. Using zlib encoding can be expensive, and the results ought to be cached. See Memory usage tuning for more information on the speed/memory/compression tradeoffs involved in zlib usage.

// Client request example
const zlib = require('zlib');
const http = require('http');
const fs = require('fs');
const { pipeline } = require('stream');

const request = http.get({ host: 'example.com',
                           path: '/',
                           port: 80,
                           headers: { 'Accept-Encoding': 'br,gzip,deflate' } });
request.on('response', (response) => {
  const output = fs.createWriteStream('example.com_index.html');

  const onError = (err) => {
    if (err) {
      console.error('An error occurred:', err);
      process.exitCode = 1;
    }
  };

  switch (response.headers['content-encoding']) {
    case 'br':
      pipeline(response, zlib.createBrotliDecompress(), output, onError);
      break;
    // Or, just use zlib.createUnzip() to handle both of the following cases:
    case 'gzip':
      pipeline(response, zlib.createGunzip(), output, onError);
      break;
    case 'deflate':
      pipeline(response, zlib.createInflate(), output, onError);
      break;
    default:
      pipeline(response, output, onError);
      break;
  }
});

// server example
// Running a gzip operation on every request is quite expensive.
// It would be much more efficient to cache the compressed buffer.
const zlib = require('zlib');
const http = require('http');
const fs = require('fs');
const { pipeline } = require('stream');

http.createServer((request, response) => {
  const raw = fs.createReadStream('index.html');
  // Store both a compressed and an uncompressed version of the resource.
  response.setHeader('Vary', 'Accept-Encoding');
  let acceptEncoding = request.headers['accept-encoding'];
  if (!acceptEncoding) {
    acceptEncoding = '';
  }

  const onError = (err) => {
    if (err) {
      // If an error occurs, there's not much we can do because
      // the server has already sent the 200 response code and
      // some amount of data has already been sent to the client.
      // The best we can do is terminate the response immediately
      // and log the error.
      response.end();
      console.error('An error occurred:', err);
    }
  };

  // Note: This is not a conformant accept-encoding parser.
  // See https://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.3
  if (/\bdeflate\b/.test(acceptEncoding)) {
    response.writeHead(200, { 'Content-Encoding': 'deflate' });
    pipeline(raw, zlib.createDeflate(), response, onError);
  } else if (/\bgzip\b/.test(acceptEncoding)) {
    response.writeHead(200, { 'Content-Encoding': 'gzip' });
    pipeline(raw, zlib.createGzip(), response, onError);
  } else if (/\bbr\b/.test(acceptEncoding)) {
    response.writeHead(200, { 'Content-Encoding': 'br' });
    pipeline(raw, zlib.createBrotliCompress(), response, onError);
  } else {
    response.writeHead(200, {});
    pipeline(raw, response, onError);
  }
}).listen(1337);

By default, the zlib methods will throw an error when decompressing truncated data. However, if it is known that the data is incomplete, or the desire is to inspect only the beginning of a compressed file, it is possible to suppress the default error handling by changing the flushing method that is used to decompress the last chunk of input data:

// This is a truncated version of the buffer from the above examples
const buffer = Buffer.from('eJzT0yMA', 'base64');

zlib.unzip(
  buffer,
  // For Brotli, the equivalent is zlib.constants.BROTLI_OPERATION_FLUSH.
  { finishFlush: zlib.constants.Z_SYNC_FLUSH },
  (err, buffer) => {
    if (err) {
      console.error('An error occurred:', err);
      process.exitCode = 1;
    }
    console.log(buffer.toString());
  });

This will not change the behavior in other error-throwing situations, e.g. when the input data has an invalid format. Using this method, it will not be possible to determine whether the input ended prematurely or lacks the integrity checks, making it necessary to manually check that the decompressed result is valid.

Memory usage tuning#

For zlib-based streams#

From zlib/zconf.h, modified for Node.js usage:

The memory requirements for deflate are (in bytes):

(1 << (windowBits + 2)) + (1 << (memLevel + 9))

That is: 128K for windowBits = 15 + 128K for memLevel = 8 (default values) plus a few kilobytes for small objects.

For example, to reduce the default memory requirements from 256K to 128K, the options should be set to:

const options = { windowBits: 14, memLevel: 7 };

This will, however, generally degrade compression.

The memory requirements for inflate are (in bytes) 1 << windowBits. That is, 32K for windowBits = 15 (default value) plus a few kilobytes for small objects.

This is in addition to a single internal output slab buffer of size chunkSize, which defaults to 16K.

The speed of zlib compression is affected most dramatically by the level setting. A higher level will result in better compression, but will take longer to complete. A lower level will result in less compression, but will be much faster.

In general, greater memory usage options will mean that Node.js has to make fewer calls to zlib because it will be able to process more data on each write operation. So, this is another factor that affects the speed, at the cost of memory usage.

For Brotli-based streams#

There are equivalents to the zlib options for Brotli-based streams, although these options have different ranges than the zlib ones:

• zlib’s level option matches Brotli’s BROTLI_PARAM_QUALITY option.
• zlib’s windowBits option matches Brotli’s BROTLI_PARAM_LGWIN option.

See below for more details on Brotli-specific options.

Flushing#

Calling .flush() on a compression stream will make zlib return as much output as currently possible. This may come at the cost of degraded compression quality, but can be useful when data needs to be available as soon as possible.

In the following example, flush() is used to write a compressed partial HTTP response to the client:

const zlib = require('zlib');
const http = require('http');
const { pipeline } = require('stream');

http.createServer((request, response) => {
  // For the sake of simplicity, the Accept-Encoding checks are omitted.
  response.writeHead(200, { 'content-encoding': 'gzip' });
  const output = zlib.createGzip();
  let i;

  pipeline(output, response, (err) => {
    if (err) {
      // If an error occurs, there's not much we can do because
      // the server has already sent the 200 response code and
      // some amount of data has already been sent to the client.
      // The best we can do is terminate the response immediately
      // and log the error.
      clearInterval(i);
      response.end();
      console.error('An error occurred:', err);
    }
  });

  i = setInterval(() => {
    output.write(`The current time is ${Date()}\n`, () => {
      // The data has been passed to zlib, but the compression algorithm may
      // have decided to buffer the data for more efficient compression.
      // Calling .flush() will make the data available as soon as the client
      // is ready to receive it.
      output.flush();
    });
  }, 1000);
}).listen(1337);

Constants#

zlib constants#

All of the constants defined in zlib.h are also defined on require('zlib').constants. In the normal course of operations, it will not be necessary to use these constants. They are documented so that their presence is not surprising. This section is taken almost directly from the zlib documentation.

Previously, the constants were available directly from require('zlib'), for instance zlib.Z_NO_FLUSH. Accessing the constants directly from the module is currently still possible but is deprecated.

Allowed flush values.

• zlib.constants.Z_NO_FLUSH
• zlib.constants.Z_PARTIAL_FLUSH
• zlib.constants.Z_SYNC_FLUSH
• zlib.constants.Z_FULL_FLUSH
• zlib.constants.Z_FINISH
• zlib.constants.Z_BLOCK
• zlib.constants.Z_TREES

Return codes for the compression/decompression functions. Negative values are errors, positive values are used for special but normal events.

• zlib.constants.Z_OK
• zlib.constants.Z_STREAM_END
• zlib.constants.Z_NEED_DICT
• zlib.constants.Z_ERRNO
• zlib.constants.Z_STREAM_ERROR
• zlib.constants.Z_DATA_ERROR
• zlib.constants.Z_MEM_ERROR
• zlib.constants.Z_BUF_ERROR
• zlib.constants.Z_VERSION_ERROR

Compression levels.

• zlib.constants.Z_NO_COMPRESSION
• zlib.constants.Z_BEST_SPEED
• zlib.constants.Z_BEST_COMPRESSION
• zlib.constants.Z_DEFAULT_COMPRESSION

Compression strategy.

• zlib.constants.Z_FILTERED
• zlib.constants.Z_HUFFMAN_ONLY
• zlib.constants.Z_RLE
• zlib.constants.Z_FIXED
• zlib.constants.Z_DEFAULT_STRATEGY

Brotli constants#

There are several options and other constants available for Brotli-based streams:

Flush operations#

The following values are valid flush operations for Brotli-based streams:

• zlib.constants.BROTLI_OPERATION_PROCESS (default for all operations)
• zlib.constants.BROTLI_OPERATION_FLUSH (default when calling .flush())
• zlib.constants.BROTLI_OPERATION_FINISH (default for the last chunk)
• zlib.constants.BROTLI_OPERATION_EMIT_METADATA
  • This particular operation may be hard to use in a Node.js context, as the streaming layer makes it hard to know which data will end up in this frame. Also, there is currently no way to consume this data through the Node.js API.

Compressor options#

There are several options that can be set on Brotli encoders, affecting compression efficiency and speed. Both the keys and the values can be accessed as properties of the zlib.constants object.

The most important options are:

• BROTLI_PARAM_MODE
  • BROTLI_MODE_GENERIC (default)
  • BROTLI_MODE_TEXT, adjusted for UTF-8 text
  • BROTLI_MODE_FONT, adjusted for WOFF 2.0 fonts
• BROTLI_PARAM_QUALITY
  • Ranges from BROTLI_MIN_QUALITY to BROTLI_MAX_QUALITY, with a default of BROTLI_DEFAULT_QUALITY.
• BROTLI_PARAM_SIZE_HINT
  • Integer value representing the expected input size; defaults to 0 for an unknown input size.

The following flags can be set for advanced control over the compression algorithm and memory usage tuning:

• BROTLI_PARAM_LGWIN
  • Ranges from BROTLI_MIN_WINDOW_BITS to BROTLI_MAX_WINDOW_BITS, with a default of BROTLI_DEFAULT_WINDOW, or up to BROTLI_LARGE_MAX_WINDOW_BITS if the BROTLI_PARAM_LARGE_WINDOW flag is set.
• BROTLI_PARAM_LGBLOCK
  • Ranges from BROTLI_MIN_INPUT_BLOCK_BITS to BROTLI_MAX_INPUT_BLOCK_BITS.
• BROTLI_PARAM_DISABLE_LITERAL_CONTEXT_MODELING
  • Boolean flag that decreases compression ratio in favour of decompression speed.
• BROTLI_PARAM_LARGE_WINDOW
  • Boolean flag enabling “Large Window Brotli” mode (not compatible with the Brotli format as standardized in RFC 7932).
• BROTLI_PARAM_NPOSTFIX
  • Ranges from 0 to BROTLI_MAX_NPOSTFIX.
• BROTLI_PARAM_NDIRECT
  • Ranges from 0 to 15 << NPOSTFIX in steps of 1 << NPOSTFIX.

Decompressor options#

These advanced options are available for controlling decompression:

• BROTLI_DECODER_PARAM_DISABLE_RING_BUFFER_REALLOCATION
  • Boolean flag that affects internal memory allocation patterns.
• BROTLI_DECODER_PARAM_LARGE_WINDOW
  • Boolean flag enabling “Large Window Brotli” mode (not compatible with the Brotli format as standardized in RFC 7932).

Class: Options#