sharp/docs/install.md

Installation

npm install sharp

yarn add sharp

Prerequisites

• Node.js v10+

Prebuilt binaries

Ready-compiled sharp and libvips binaries are provided for use with Node.js v10+ on the most common platforms:

• macOS x64 (>= 10.13)
• Linux x64 (glibc >= 2.17, musl >= 1.1.24)
• Linux ARM64 (glibc >= 2.29, musl >= 1.1.24)
• Windows x64
• Windows x86

An ~8MB tarball containing libvips and its most commonly used dependencies is downloaded via HTTPS and stored within node_modules/sharp/vendor during npm install.

This provides support for the JPEG, PNG, WebP, AVIF, TIFF, GIF (input) and SVG (input) image formats.

The following platforms have prebuilt libvips but not sharp:

• macOS ARM64
• Linux ARMv6
• Linux ARMv7 (glibc >= 2.28)
• Windows ARM64

The following platforms require compilation of both libvips and sharp from source:

• Linux x86
• Linux x64 (glibc <= 2.16, includes RHEL/CentOS 6)
• Linux ARM64 (glibc <= 2.28)
• Linux PowerPC
• FreeBSD
• OpenBSD

Common problems

The architecture and platform of Node.js used for npm install must be the same as the architecture and platform of Node.js used at runtime.

When using npm v6 or earlier, the npm install --unsafe-perm flag must be used when installing as root or a sudo user.

When using npm v7, the user running npm install must own the directory it is run in.

The npm install --ignore-scripts=false flag must be used when npm has been configured to ignore installation scripts.

Check the output of running npm install --verbose sharp for useful error messages.

Apple M1

Prebuilt libvips binaries are provided for macOS on ARM64 (since sharp v0.28.0).

During npm install sharp will be built locally, which requires Xcode and Python - see building from source.

When this new ARM64 CPU is made freely available to open source projects via a CI service then prebuilt sharp binaries can also be provided.

Custom libvips

To use a custom, globally-installed version of libvips instead of the provided binaries, make sure it is at least the version listed under config.libvips in the package.json file and that it can be located using pkg-config --modversion vips-cpp.

For help compiling libvips from source, please see https://libvips.github.io/libvips/install.html#building-libvips-from-a-source-tarball.

The use of a globally-installed libvips is unsupported on Windows.

Building from source

This module will be compiled from source at npm install time when:

• a globally-installed libvips is detected (set the SHARP_IGNORE_GLOBAL_LIBVIPS environment variable to skip this),
• prebuilt sharp binaries do not exist for the current platform, or
• when the npm install --build-from-source flag is used.

Building from source requires:

• C++11 compiler
• node-gyp and its dependencies

Custom prebuilt binaries

This is an advanced approach that most people will not require.

To install the prebuilt sharp binaries from a custom URL, set the sharp_binary_host npm config option or the npm_config_sharp_binary_host environment variable.

To install the prebuilt libvips binaries from a custom URL, set the sharp_libvips_binary_host npm config option or the npm_config_sharp_libvips_binary_host environment variable.

The version subpath and file name are appended to these. There should be tarballs available that are compressed with both gzip and Brotli, as the format downloaded will vary depending on whether the user's version of Node supports Brotli decompression (Node.js v10.16.0+)

For example, if sharp_libvips_binary_host is set to https://hostname/path and the libvips version is 1.2.3 then the resultant URL will be https://hostname/path/v1.2.3/libvips-1.2.3-platform-arch.tar.br or https://hostname/path/v1.2.3/libvips-1.2.3-platform-arch.tar.gz.

See the Chinese mirror below for a further example.

Chinese mirror

Alibaba provide a mirror site based in China containing binaries for both sharp and libvips.

To use this either set the following configuration:

npm config set sharp_binary_host "https://npm.taobao.org/mirrors/sharp"
npm config set sharp_libvips_binary_host "https://npm.taobao.org/mirrors/sharp-libvips"
npm install sharp

or set the following environment variables:

npm_config_sharp_binary_host="https://npm.taobao.org/mirrors/sharp" \
  npm_config_sharp_libvips_binary_host="https://npm.taobao.org/mirrors/sharp-libvips" \
  npm install sharp

FreeBSD

The vips package must be installed before npm install is run.

pkg install -y pkgconf vips

cd /usr/ports/graphics/vips/ && make install clean

Linux memory allocator

The default memory allocator on most glibc-based Linux systems (e.g. Debian, Red Hat) is unsuitable for long-running, multi-threaded processes that involve lots of small memory allocations.

For this reason, by default, sharp will limit the use of thread-based concurrency when the glibc allocator is detected at runtime.

To help avoid fragmentation and improve performance on these systems, the use of an alternative memory allocator such as jemalloc is recommended.

Those using musl-based Linux (e.g. Alpine) and non-Linux systems are unaffected.

Heroku

Add the jemalloc buildpack to reduce the effects of memory fragmentation.

Set NODE_MODULES_CACHE to false when using the yarn package manager.

AWS Lambda

Set the Lambda runtime to nodejs12.x.

The binaries in the node_modules directory of the deployment package must be for the Linux x64 platform.

When building your deployment package on machines other than Linux x64 (glibc), run the following commands:

macOS:

rm -rf node_modules/sharp
SHARP_IGNORE_GLOBAL_LIBVIPS=1 npm install --arch=x64 --platform=linux sharp

Windows:

rmdir /s /q node_modules/sharp
npm install --arch=x64 --platform=linux sharp

Alternatively a Docker container closely matching the Lambda runtime can be used:

rm -rf node_modules/sharp
docker run -v "$PWD":/var/task lambci/lambda:build-nodejs12.x npm install sharp

To get the best performance select the largest memory available. A 1536 MB function provides ~12x more CPU time than a 128 MB function.

Webpack

Ensure sharp is added to the externals configuration.

externals: {
  'sharp': 'commonjs sharp'
}

Worker threads

The main thread must call require('sharp') before worker threads are created to ensure shared libraries remain loaded in memory until after all threads are complete.

Known conflicts

Canvas and Windows

The prebuilt binaries provided by canvas for Windows depend on the unmaintained GTK 2, last updated in 2011.

These conflict with the modern, up-to-date binaries provided by sharp.

If both modules are used in the same Windows process, the following error will occur:

The specified procedure could not be found.