dastash
dastash is a persistent disk cache that every R process on a machine can
share, in the style of Python’s
diskcache. Keys map to
values that outlive the session, expire on a clock, carry tags for bulk
invalidation, and are evicted when the cache grows past a size limit.
Small values live in a transactional
libmdbx database, through the
mdbx package; larger ones become
read-only files named by their content, stored once however many keys
hold them.
Installation
install.packages("dastash")
The development version is on GitHub:
# install.packages("pak")
pak::pak("pedrobtz/dastash")
A tour
library(dastash)
s <- stash(file.path(tempdir(), "prices"))
# Keys map to values, with expiry and tags.
s |> stash_set("XSWX/2026-08-29", c(101.2, 101.9), expire = 3600, tags = "XSWX")
stash_get(s, "XSWX/2026-08-29")
#> [1] 101.2 101.9
stash_get(s, "missing", default = NULL)
#> NULL
# A key can be any combination of values; field order does not matter.
stash_set(s, list(exchange = "XSWX", date = as.Date("2026-08-29")), "quotes")
stash_get(s, list(date = as.Date("2026-08-29"), exchange = "XSWX"))
#> [1] "quotes"
# A function, cached across sessions and processes.
slow_square <- function(x) {
Sys.sleep(1)
x^2
}
square <- stash_memoise(slow_square, s)
system.time(square(12))
#> user system elapsed
#> 0.003 0.000 1.034
system.time(square(12))
#> user system elapsed
#> 0.000 0.000 0.001
# The catalogue is a data frame.
stash_entries(s)[, c("key", "bytes", "codec", "expires")]
#> key bytes codec expires
#> 1 XSWX/2026-08-29 47 rds 2026-09-29 19:59:14
#> 2 slow_square/v1/{x=i:12} 39 rds <NA>
#> 3 {date=d:2026-08-29,exchange=s:XSWX} 6 raw <NA>
stash_close(s)
Where it works
Many R processes on one machine, over a local filesystem:
interactive sessions, scheduled jobs and parallel workers can use one
stash at once. Reads never wait; writes take turns. A stash on a network
filesystem (NFS, SMB, cloud file shares) is not supported, because its
locks cannot be relied on.
A stash cannot be carried into a forked child such as a
parallel::mclapply() worker: open it inside the worker.
Learn more
