The hardware and bandwidth for this mirror is donated by dogado GmbH, the Webhosting and Full Service-Cloud Provider. Check out our Wordpress Tutorial.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]dogado.de.

memoise

CRAN status R build status Codecov test coverage

The memoise package makes it easy to memoise R functions. Memoisation (https://en.wikipedia.org/wiki/Memoization) caches function calls so that if a previously seen set of inputs is seen, it can return the previously computed output.

Installation

Install from CRAN with

install.packages("memoise")

Usage

To memoise a function, use memoise():

library(memoise)
f <- function(x) {
  Sys.sleep(1)
  mean(x)
}
mf <- memoise(f)
system.time(mf(1:10))
#>    user  system elapsed
#>   0.002   0.000   1.003
system.time(mf(1:10))
#>    user  system elapsed
#>   0.000   0.000   0.001

You can clear mf’s cache with:

forget(mf)

And you can test whether a function is memoised with is.memoised().

Caches

By default, memoise uses an in-memory cache, using cache_mem() from the cachem package. cachem::cache_disk() allows caching using files on a local filesystem.

Both cachem::cache_mem() and cachem::cache_disk() support automatic pruning by default; this means that they will not keep growing past a certain size, and eventually older items will be removed from the cache. The default size cache_mem() is 512 MB, and the default size for a cache_disk() is 1 GB, but this can be customized by specifying max_size:

# 100 MB limit
cm <- cachem::cache_mem(max_size = 100 * 1024^2)

mf <- memoise(f, cache = cm)

You can also change the maximum age of items in the cache with max_age:

# Expire items in cache after 15 minutes
cm <- cachem::cache_mem(max_age = 15 * 60)

mf <- memoise(f, cache = cm)

By default, a cache_disk() uses a subdirectory the R process’s temp directory, but it is possible to specify the directory. This is useful for persisting a cache across R sessions, sharing a cache among different processes, or even for synchronizing across the network.

# Store in "R-myapp" directory inside of user-level cache directory
cd <- cachem::cache_disk(rappdirs::user_cache_dir("R-myapp"))

# Store in Dropbox
cdb <- cachem::cache_disk("~/Dropbox/.rcache")

A single cache object can be shared among multiple memoised functions. By default, the cache key includes not only the arguments to the function, but also the body of the function. This essentially eliminates the possibility of a cache collision, even if two memoised functions are called with the same arguments.

m <- cachem::cache_mem()

times2 <- memoise(function(x) { x * 2 }, cache = m)
times4 <- memoise(function(x) { x * 4 }, cache = m)

times2(10)
#> [1] 20
times4(10)
#> [1] 40

Cache API

It is possible to use other caching backends with memoise. These caching objects must be key-value stores which use the same API as those from the cachem package. The following methods are required for full compatibiltiy with memoise:

Note that the sentinel value for missing keys can be created by calling cachem::key_missing(), or structure(list(), class = "key_missing").

Old-style cache objects

Before version 2.0, memoise used different caching objects, which did not have automatic pruning and had a slightly different API. These caching objects can still be used, but we recommend using the caching objects from cachem when possible.

With the old-style caching objects, memoise first checks for the existence of a key in the cache, and if present, it fetches the value. This results in a possible race condition (when using caches other than the memory cache): an object could be deleted from the cache after the existence check, but before the value is fetched. With the new cachem-style caching objects, the possibility of a a race condition is eliminated: memoise simply tries to fetch the key, and if it’s not present in the cache, the cache returns a sentinel value indicating that it’s missing. (Note that the caching objects must also be designed to avoid a similar race condition internally.)

The following cache objects do not currently have an equivalent in cachem.

These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.
Health stats visible at Monitor.