Changelog
Source:NEWS.md
crew 0.10.2
- Eliminate spurious
launch_max
error from underutilized workers (#189). - Deprecate
launch_max
in favor ofcrashes_error
(#189). - Look for crashes of all workers in
rotate()
instead of looking for crashes of a specific worker inlaunch()
(#189). - Add a
crashes()
launcher method to allow plugins to detect and respond to crashes more easily. - Change default
seconds_idle
to 300. - Move
autometric
toSuggests:
.
crew 0.10.1
CRAN release: 2024-11-03
- Instrument
crew_eval()
withautometric::log_phase_set()
andautometric::log_phase_reset()
.
crew 0.10.0
CRAN release: 2024-10-11
- Give full worker name in the worker launch error message (@multimeric).
- Start the controller automatically in more methods.
- Add an
r_arguments
argument to supply command line arguments to R for workers (#175, @rpruim). - Add a
cancel()
controller method to cancel one or more tasks. - Call
cancel(all = TRUE)
fromterminate()
as a safeguard. - Make
client$dispatcher
aps::ps_handle()
handle and add a newclient$client
handle for the current process. - To passively log memory usage when
log_resources
is given, the controller now callslog()
as a side effect in most controller methods, with throttling to preserve speed. - Return a status and status code from
pop()
etc. - New internal function
as_monad()
makes error reporting more consistent. - Use
path.expand()
on local log files. - Switch to
Rscript
. - Print session info from
crew_worker()
before attempting to dial intomirai
. - Add
asyncdial = FALSE
back tomirai::daemon()
call. - Deprecate
local_log_directory
andlocal_log_join
in favor ofcrew_options_local()
and theoptions_local
argument. - Add
crew_options_metrics()
and theoptions_metrics
argument for recording resource metrics (#178). - Add a new
logging.Rmd
vignette to explain best practices for logging and resource usage metrics (#178).
crew 0.9.4
CRAN release: 2024-06-20
- Do not use extended tasks in Shiny vignette.
- Add a new
retry_tasks
argument with defaultTRUE
(#170). - Avoid Base64 encoding functions from
nanonext
crew 0.9.2
CRAN release: 2024-04-24
- Use
.args
rather than...
inmirai::mirai()
to make sure arguments continue to be passed as local variables inmirai
>= 0.13.1.9012. - Add new controller methods
autoscale()
,descale()
, andstarted()
to facilitate different kinds of Shiny apps. - Deprecate the
scale
andthrottle
methods ofcontroller$promise()
.promise()
now always callsautoscale()
to make sure one and only one auto-scaling loop is running asynchronously. Auto-scaling thus continues even after the promise resolves. - Add a second example vignette that simulates coin flips.
- Add a new
error
argument tocollect()
(#166).
crew 0.9.1
CRAN release: 2024-03-25
- Rewrite the async Shiny vignette with
crew
promises and Shiny extended tasks (#157, @jcheng5). - Clarify the intent of
controller$promise(mode = "one")
in the vignette on promises (@jcheng5). - Implement an
error
argument inpop()
which may help with integration withExtendedTask
(@jcheng5). - Handle task errors in the Shiny vignette (@jcheng5).
crew 0.9.0
CRAN release: 2024-02-08
- Require
nanonext
>= 0.12.0 andmirai
>= 0.12.0. - Return to always re-launching backlogged inactive workers (#79, https://github.com/shikokuchuo/mirai/discussions/95).
- Implement
push_backlog()
andpop_backlog()
to manage cases when it is not desirable to push to saturated controllers (https://github.com/ropensci/targets/issues/1220). - Invisibly return the
mirai
object of a task pushed to the controller. This allows users to interact with the task directly, e.g. to create a promise object withpromises::as.promise()
(#146, @jcheng5). - Add a new
walk()
method for controllers and controller groups to submit tasks in batch and return control immediately without waiting for any task to complete (#148, @jcheng5). - Revive the
collect()
method for popping multiple tasks at once (#148, @jcheng5). - Add controller group methods
nonempty()
,resolved()
,unresolved()
, andunpopped()
to help with #148. - Make the
mirai
dispatcher error message extremely verbose.
crew 0.8.0
CRAN release: 2024-01-09
- Configure workers to send themselves a termination signal if the connection to the dispatcher is broken (#141, @psychelzh). Huge thanks to @shikokuchuo for the support through https://github.com/shikokuchuo/mirai/issues/87, https://github.com/shikokuchuo/mirai/pull/88, and https://github.com/shikokuchuo/nanonext/pull/25! The signal itself is platform-dependent and determined by the new function
crew_terminate_signal()
. - Implement
crew_monitor_local()
to help users monitor and terminate local R processes created bycrew
andmirai
. - Implement new utility function
crew_terminate_process()
to terminate a process manually without resorting toSIGKILL
on Windows. - Throw a warning from
controller$map()
if at least one task threw one.warnings = FALSE
suppresses this behavior. - Set
output = TRUE
indaemon()
sostdout
andstderr
streams print. - Add new arguments
local_log_directory
andlocal_log_join
to write to local log files.
crew 0.7.0
CRAN release: 2023-12-11
- Migrate from
asyncdial
toautoexit
. - Use
Sys.info()[["user"]]
to get the user increw_clean()
(#135, @luwidmer). - Use condition variables to wait for tasks more efficiently (#108).
- Because of #108,
controller$map()
can no longer be used if there are tasks waiting to be popped. - Use a
cli
progress bar inmap()
. - Encapsulate non-function
R6
members in theprivate
list and add active bindings where interfaces are necessary (#137). Ad hoc tests that absolutely need to modify private objects can useobject$.__enclos_env__$private
trick, but these tests should be skipped on CRAN in case there is a change toR6
that breaks this. - Drop
.signal
frommirai::mirai()
since allmirai
tasks signal as ofmirai
version 0.11.2.9025. - Implement
crew_throttle()
, a decoupled mechanism for throttling that can be applied to scaling. - Bring back the
throttle
argument, powered bycrew_throttle()
. - Retry
mirai::status()
again indaemons_info()
and make it configurable usingseconds_interval
andseconds_timeout
in both the client and the launcher (#128).
crew 0.6.0
CRAN release: 2023-10-12
- Migrate checks to enforce features in version 0.5.0 for reverse dependencies.
- Drop check of backlogged workers, c.f. https://github.com/shikokuchuo/mirai/discussions/63#discussioncomment-7051889 (#79, #124, @shikokuchuo).
- Deprecate
seconds_exit
becauseexitlinger
inmirai
is now obsolete (#125, @shikokuchuo). - Use
mirai::nextget("cv")
to count unresolved tasks instead of looping through all the task objects (#131). - Remove throttling and
collect()
in auto-scaling. Simplifies much of the code. Made possible by the efficiency gains in #131. - Simplify
wait()
. -
seconds_interval
inmap()
no longer defaults tocontroller$client$seconds_interval
. -
launcher$terminate_workers()
terminates one or more workers, andlauncher$terminate
terminates the whole launcher. - Add infrastructure to let custom launcher plugins launch and terminate workers asynchronously (#133). Launchers can set a positive number in the
processes
field to set the number of localmirai
daemons for asynchronous requests to launch and terminate the serious workers. Then,launch_worker()
andterminate_worker()
can optionally make use oflauncher$async$eval()
to send these asynchronous calls.
crew 0.5.0
CRAN release: 2023-09-18
- Suppress interactive browser on Windows which launched on each worker previously (@psychelzh).
- Migrate to the new host/daemon nomenclature in
mirai
0.9.1 (#96). - Suppress
status()
retries (@shikokuchuo, #100). - Implement
launch_max
to error out if workers repeatedly launch without completing any tasks (#101, @shikokuchuo, @multimeric). - Detect discovered workers more robustly in
launcher$done()
. - Add a new
algorithm
argument to thepush()
,shove()
, andmap()
methods of controllers and controller groups, as well as arguments/fields increw_eval()
and its return value (#113, @shikokuchuo). - As a default for pseudo-random number generation, leverage widely-spaced L’Ecuyer streams as supported by
mirai::nextstream()
(#115, @shikokuchuo). - Move README documentation to vignettes.
- Add a new
crew_tls()
function for TLS configuration (#120). - Deprecate the
tls_enable
andtls_config
arguments ofcrew_client()
etc. in favor of atls
argument which acceptscrew_tls()
objects (#120).
crew 0.4.0
CRAN release: 2023-07-10
Highlights
- Use the TLS capabilities of
nanonext
andmirai
. - Add a
map()
method to the controller class for functional programming (#93). - Fix bug in
scale()
which launched too many workers whenactive - demand
was less than 0. - Add
crew_clean()
to help clean up any lingering dispatchers and workers from previous sessions.
crew 0.3.0
CRAN release: 2023-06-26
- Track warnings properly (#81, @brendanf).
- Optimize heavily (#81, #83, @shikokuchuo, @brendanf).
- Use CRAN
nanonext
0.9.0. - Delegate the task pushing and collection logic to a new
R6
schedule class (#84). - Delegate auto-scaling to the launcher (#84).
- Drastically simplify internal logic (#84).
- Rename “router” to “client” and change the function signature of
launch_worker()
(#84). Unfortunately neither of these changes is back-compatible for third-party launcher plugins. However, the controller interface is still back-compatible, sotargets
remains unaffected. - Terminate “lost” workers on websocket rotation.
- Allow
NULL
result inpop()$result[[1L]]
. - Delegate summaries among the controller, schedule, launcher, and client.
- Default to the loopback address for the local controller.
- Use already cumulative
assigned
andcomplete
stats frommirai
0.8.7.9029. - Reorganize
pkgdown
reference andRd
families to clarify which parts are for end-users and which parts are for launcher plugin developers. - Add
controller$shove()
andschedule$list()
to support fast extensions for functional programming. - Depend on
mirai
0.9.0 or greater.
crew 0.2.0
- Use
.args
instead of...
inmirai::mirai()
. - Use
set.seed()
instead ofwithr::local_seed()
(#67, @shikokuchuo). - Add a vignette to showcase integration with Shiny, co-authored with @dwoodie and with helpful suggestions by @shikokuchuo.
- Drop
rlang::duplicate()
(#72, @shikokuchuo). - Do not poll dispatcher availability when starting router objects (@shikokuchuo).
- Implement a
saturated()
controller method to supporttargets
. - Add a
worker_index
column to controller summaries. - When relaunching workers, prioritize those with unresolved tasks stuck at the NNG level (#75, @shikokuchuo).
- Always relaunch backlogged inactive workers (#79).
- Deprecate the
auto_scale
argument/field of controllers in favor of thescale
argument ofpush()
,pop()
, andwait()
. - Throttle auto-scaling with interval
self$router$seconds_interval
(#76). - Remove
clean()
and instead terminate lost workers on launch. - Fix examples.
- Add a
collect
argument topop()
.
crew 0.1.0
CRAN release: 2023-04-19
- Remove
asyncdial
from call tomirai::daemons()
. - Explain vision in the launcher plugin vignette.
- Remove the superfluous
clean()
controller method. - Clean up logic for
scale()
andlaunch()
controller methods. - Add a new
inactive()
launcher method to help with the above. - Eliminate superfluous worker launches: clean up the controller logic, and check worker discoveries before worker connections (#51).
- Simplify the launcher plugin interface (#54).
- Commit to using the fast
nanonext::.unresolved()
(#49). - Rename the
callr
launcher to the local process launcher (#56). - Switch from SHA224 to SHA1 as the hash for random tokens and names (#59, @shikokuchuo).
- Open and confirm bus socket listeners asynchronously in the launcher abstract class (#58).
- Use condition variables to test the connection status of worker instances (#57, @shikokuchuo).
- Allow the dispatcher process to not exist so that the router will keep working if
mirai
implements the dispatcher another way (#60). - Add a
substitute
argument topush()
(#63). - Remove dependency on custom NNG bus sockets, which significantly simplifies the internals, improves robustness, and smooths the integration with
mirai
(#61). - Use
processx
instead ofcallr
for the local process launcher (#64”). - Speed up
crew_eval()
by e.g. relying onnanonext::mclock()
for the time. - Add an
empty()
method to check if the controller or controller group is empty. - Rely on
mirai
for globals cleanup, garbage collection, and global option resets (#65). - Expose the package unloading capabilities of
mirai
(#65). - Make monads classed environments and leave tibbles to the last minute in
pop()
. - Rename
crew_wait()
tocrew_retry()
. - Export
crew_assert()
andcrew_random_name()
to help launcher plugins. - Rename the
listen()
method tostart()
in the router class. - The
poll()
method now retriesmirai::daemons()
for robustness. -
crew_retry()
gains amax_tries
argument. - Bug fix: stop setting
router$daemons
toNULL
inpoll()
when the connection has a momentary problem. - Simplify
listening()
.
crew 0.0.5
CRAN release: 2023-03-23
- Write a
call()
method to make it easier to write custom launcher plugins. - Add robust documentation and tests for custom launcher plugins.
- Use
as.character()
to make sure thelist
argument supplied torm()
is safe. Fixes a test that was failing on R-devel withcrew
version 0.0.4. - Remove the
seconds_poll_high
andseconds_poll_low
fields and arguments for compatibility withmirai
development and release versions. - Hard-code the
async_dial
field in the router for compatibility withmirai
development and release versions.