Memoisation of Functions

Cache the results of a function so that when you call it again with the same arguments it returns the pre-computed value.

Travis-CI Build Status Coverage Status


If a function is called multiple times with the same input, you can often speed things up by keeping a cache of known answers that it can retrieve. This is called memoisation The memoise package provides a simple syntax

mf <- memoise(f)

to create mf(), a memoised wrapper around f(). You can clear mf's cache with


and you can test whether a function is memoised with

is.memoised(mf) # TRUE
is.memoised(f)  # FALSE



External Caches

memoise also supports external caching in addition to the default in-memory caches.

  • cache_filesystem() allows caching using files on a local filesystem. You can point this to a shared file such as dropbox or google drive to share caches between systems.
  • cache_s3() allows caching on Amazon S3


Use cache_s3() to cache objects using s3 storage. Requires you to specify a bucket using cache_name. When creating buckets, they must be unique among all s3 users when created.

Sys.setenv("AWS_ACCESS_KEY_ID" = "<access key>",
           "AWS_SECRET_ACCESS_KEY" = "<access secret>")
mrunif <- memoise(runif, cache = cache_s3("<unique bucket name>"))
mrunif(10) # First run, saves cache
mrunif(10) # Loads cache, results should be identical


cache_filesystem can be used for a file system cache. This is useful for preserving the cache between R sessions as well as sharing between systems when using a shared or synced files system such as Dropbox or Google Drive.

fc <- cache_filesystem("~/.cache")
mrunif <- memoise(runif, cache = fc)
mrunif(20) # Results stored in local file
dbc <- cache_filesystem("~/Dropbox/.rcache")
mrunif <- memoise(runif, cache = dbc)
mrunif(20) # Results stored in Dropbox .rcache folder which will be synced between computers.
gdc <- cache_filesystem("~/Google Drive/.rcache")
mrunif <- memoise(runif, cache = gdc)
mrunif(20) # Results stored in Google Drive .rcache folder which will be synced between computers.


Version 1.1.0

  • Caches now hash the function body along with the arguments, to ensure functions with identical arguments use a separate file-system cache. (#38)
  • Handle missing arguments in memoised functions for simple cases not using non-standard-evaluation (#19).
  • memoise() gains a cache= argument to specify an external cache. Two types of caches are available, cache_s3() for amazon S3 and cache_filesystem() for a file system cache (#25, @danielecook).

Version 1.0.0

  • memoise() now signals an error if an already memoised function is used as input (#4, @richierocks).
  • has_cache() function added which returns a boolean depending on if the given call is cached or not (#10, @dkesh).
  • Memoised functions now have a print method which displays the original function definition, rather than the memoisation code (#15, @jimhester).
  • A memoised function now has the same interface as the original function, if the original function is known when memoise is called. (Otherwise, the old behavior is invoked, with a warning.) (#14, @krlmlr)
  • The enclosing environment of the memoised function is specified explicitly, defaults to parent.frame().
  • is.memoised now checks if the argument is a function.
  • Testing infrastructure, full test coverage.

Version 0.2.1

  • Update to fix outstanding R CMD check issues.

Version 0.2 (2010-11-11)

New features

  • Memoised functions now have an attribute memoised=TRUE, and is.memoised() tests whether a function is memoised. (Contributed by Sietse Brouwer.)


  • Documentation is now more elaborate, and hopefully more accessible to newcomers. Thanks to Sietse Brouwer for the verbosity.

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.