Mutually exclusive (mutex) locks are used to control access to shared
resources.
An exclusive lock grants permission to one process at a time, for
example to update the contents of a database file. While an exclusive lock
is active, no other exclusive or shared locks will be granted.
Multiple shared locks can be held by different processes at the same
time, for example to read a database file. While a shared lock is active, no
exclusive locks will be granted.
Arguments
- name
Unique ID. Alphanumeric, starting with a letter.
- assert
Apply an additional constraint.
'create'
- Error if the mutex already exists.'exists'
- Error if the mutex doesn't exist.NULL
- No constraint; create the mutex if it doesn't exist.
- cleanup
Remove the mutex when the R session exits. If
FALSE
, the mutex will persist until$remove()
is called or the operating system is restarted.- file
Use a hash of this file/directory path as the mutex name. The file itself will not be read or modified, and does not need to exist.
- data
A
mutex
object.- expr
Expression to evaluate if the mutex is acquired.
- alt_expr
Expression to evaluate if
timeout_ms
is reached.If
FALSE
(the default) an exclusive lock is returned. IfTRUE
, a shared lock is returned instead. See description.- timeout_ms
Maximum time (in milliseconds) to block the process while waiting for the operation to succeed. Use
0
orInf
to return immediately or only when successful, respectively.- ...
Not used.
Value
mutex()
returns a mutex
object with the following methods:
$name
Returns the mutex's name (scalar character).
$lock(shared = FALSE, timeout_ms = Inf)
Returns
TRUE
if the lock is acquired, orFALSE
if the timeout is reached.
$unlock(warn = TRUE)
Returns
TRUE
if successful, orFALSE
(with optional warning) if the mutex wasn't locked.
$remove()
Returns
TRUE
on success, orFALSE
if the mutex wasn't found.
with()
returns eval(expr)
if the lock was acquired, or eval(alt_expr)
if the timeout is reached.
Error Handling
The with()
wrapper automatically unlocks the mutex if an error stops
evaluation of expr
. If you are directly calling lock()
, be sure that
unlock()
is registered with error handlers or added to on.exit()
.
Otherwise, the lock will persist until the process terminates.
Duplicate Mutexes
Mutex locks are per-process. If a process already has a lock, it can not attempt to acquire a second lock on the same mutex.
Examples
tmp <- tempfile()
mut <- interprocess::mutex(file = tmp)
print(mut)
#> <mutex> "C8iOKsEEg8E"
# Exclusive lock to write the file
with(mut, writeLines('some data', tmp))
# Use a shared lock to read the file
with(mut,
shared = TRUE,
timeout_ms = 0,
expr = readLines(tmp),
alt_expr = warning('Mutex was locked. Giving up.') )
#> [1] "some data"
# Directly lock/unlock with safeguards
if (mut$lock(timeout_ms = 0)) {
local({
on.exit(mut$unlock())
writeLines('more data', tmp)
})
} else {
warning('Mutex was locked. Giving up.')
}
mut$remove()
unlink(tmp)