Locally caching 'npm install'. Useful in CI-like situations.
npmi works by hashing package-lock.json / npm-shrinkwrap.json / package.json to get a reasonably unique id and using that as a cache key.
The cache key is something like
v8.9.1-darwin-x64-prod-0a8ca5554ece9332de1bb64096d04a57
.
The cache key is determined from node version, platform, cpu architecture, NODE_ENV and the content of one of the package files mentioned above.
On the first run npmi will run 'npm install' as usual, storing whatever ends up in node_modules to $CACHEDIR/$HASH.tgz.
Later npmi will simply extract node_modules from the cache without hitting the network at all.
It can also use a Redis server to share cached files between hosts.
node_modules may contain compiled/binary files and npmi should only be run in the same environment as the actual code to prevent nasty surprises.
$ npmi -h
NPMI v6.0.0 - a caching 'npm install'
Usage: npmi [-hcefprtv]
-h Display this help
-c Use specified cache dir: Default /tmp/npmi
-e Cache existing node_modules AKA 'Oops I forgot to npmi' mode
-f Force install from NPM and update cache
-p Run given command before packages are cached
-r Use specified Redis server for shared cache
-t Specify Redis TTL in seconds. Default: 259200
-v Verbose output
The following configuration options are available.
Key | Value |
---|---|
CACHEDIR | Local cache directory. Default: $TMPDIR/npmi . |
VERBOSE | Verbose output. Default: 0 |
FORCE | Force reinstall. Default: 0 |
CACHE_EXISTING | Cache whatever is currently in node_modules. Default: 0 |
REDIS_SERVER | Host/IP for Redis server. Default: "" (Don't use Redis) |
REDIS_PORT | What port to use when connecting to Redis. Default: 6379 |
REDIS_TTL | How many seconds should modules be cached. Default: 86400 (24h) |
REDIS_PREFIX | Key prefix for Redis. Default: NPMI4 |
USE_NPM_CI | Use npm ci when available. Default: 1 |
NPMI can be configured using a file, .npmirc
located in the working
directory.
The file format is KEY=VALUE
separated by newlines.
All configuration can be given as environment variables if necessary.
The options are same for both .npmirc
but are prefixed with NPMI_
.
To run NPMI on verbose mode one would use NPMI_VERBOSE=1 npmi
.
NPM 5.7.1 added 'npm ci' for faster installs. NPMI now uses that for installation when possible to speed things up even more. If you want to use the classic 'npm install' method, see the USE_NPM_CI configuration option.
Even if using a local NPM cache there are often packages which need to compile something or download an external dependency. NPMI can skip these steps if nothing has changed and generally is something like 50-100x faster than running npm install.
Clone some project with a decent amount of dependencies
$ git clone https://github.com/request/request
Install deps once to fill local caches etc.
$ npm install
# Downloading, compiling, extracting, doing stuff...
Remove the deps and install again with timing
$ rm -rf node_modules && time npm install
# npm install 29.38s user 13.26s system 158% cpu 26.960 total
Install with NPMI
$ npmi -v
# - HASH e3102336c4165f1ca58ef03f09e1a9fd determined from package.json (v4.2.1-darwin-x64)
# - Cleaning node_modules...
# - Attempt to install from cache
# - Modules not found in cache
# - Installing modules from NPM
# - Modules cached successfully
Install from NPMI cache
$ time npmi -v
# - HASH e3102336c4165f1ca58ef03f09e1a9fd determined from package.json (v4.2.1-darwin-x64)
# - Cleaning node_modules...
# - Attempt to install from cache
# - Modules installed successfully
# npmi -v 0.57s user 3.65s system 94% cpu 4.486 total
Any post-installation script of NPM will NOT get run when installing from cache. This includes at least the following:
install
postinstall
For instance a postinstall
block in package.json which does something
outside of node_modules will only be run on a cache MISS. If it merely creates
a file in node_modules etc. those will get cached.