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.
foreach
vignette.foreach
vignette to respect the
CRAN
canonical package URL format.foreach.Rmd
to illustrate how
parabar
can be used in conjunction with the
foreach
package. The implementation for the
%dopar%
operator is provided by the doParabar
package. The foreach.Rmd
vignette is locally build from the
foreach.Rmd.orig
file.AsyncBackend
to allow forceful stopping of the
cluster. This is controlled by the package option
stop_forceful
, which, when set to TRUE
,
permits the backend to stop the cluster even if tasks are running or if
tasks have finished but their results have not yet been read into the
main session. The stopping logic is now based on the newly added class
SessionState
. Closes #59.SessionState
class to easily check if an
asynchronous backend is ready for certain operations.SessionState
and
TaskState
classes, and other improvements to the existing
tests.SyncBackend
and
AsyncBackend
finalizers. These tests don’t perform real
assertions but instead verify that finalizers are invoked when backend
instances are garbage collected. As of now, there appears to be no
reliable way to programmatically capture conditions handled within a
finalizer triggered by manual gc()
invocation.SyncBackend
and AsyncBackend
destructors in favour of R
finalizers registered via
base::reg.finalizer
.Service
to BackendService
for
clarity and consistency. While the name Service
makes sense
in the context of the parabar
package, it can be confusing
when using parabar
in conjunction with other packages
(e.g., in the context of the powerly
package). The new name
BackendService
is hopefully more descriptive and less
likely to clash with other packages.comparison.Rmd
vignette to refresh the
comparison between parabar
and pbapply
packages.task_is_running
test helper to accept more
context types.Specification
to allow backends with a single
core. Closes #71.base::reg.finalizer
mechanism, with finalizers being
automatically registered when a backend is created. This ensures that
the cluster is stopped and resources are released even if the user
forgets to call stop_backend
explicitly. Closes #61.sink
redirect to use the correct destination on
Windows
(i.e., nul
instead of
/dev/null
). Closes #74.README.md
per
CRAN
checks.roxygen
checks for R6
documentation
in DESCRIPTION
.Context
to implement the Service
interface.ProgressTrackingContext.R
class. Closes #44.UserApiConsumer
class.
Closes #41.UML
design diagram. Closes #45.apply
operation to README
documentation.Service$apply
operation for
Backend
classes to validate the provided
margin
argument before running the parallel operation.Helper$check_array_margins
to validate the
margins provided to the Service$apply
operation.Exception$array_margins_not_compatible
for using improper margins in the Service$apply
operation.Exception$primitive_as_task_not_allowed
for trying to decorate primitive functions with progress tracking in the
ProgressTrackingContext
class.Helper$is_of_class
to check if an object is
of a given class.get_output
operation of
SyncBackend
for consistency.Service$lapply
and
Service$apply
operations for all classes that implement the
Service
interface.par_lapply
and par_apply
functions to
the user API
. These functions can be used to run tasks in
parallel akin to parallel::parLapply
and
parallel::parApply
, respectively.UserApiConsumer
R6
class that provides
an opinionated wrapper around the developer API
of the
parabar
package. All parallel operations (e.g.,
par_sapply
and par_lapply
) follow more or less
the same pattern. The UserApiConsumer
encapsulates this
pattern and makes it easier to extend parabar
with new
parallel functions (e.g., par_apply
) while avoiding code
duplication. The UserApiConsumer
class can also be used as
a standalone class for parallel operations, however, its primary purpose
is to be used by the parallel task execution functions in the user
API
.x
argument of task
execution functions in ProgressTrackingContext
class.parabar
in file name.file.create
in
ProgressTrackingContext
class. This warning is superfluous
since the code handles creation failures.par_sapply
to use the
UserApiConsumer
class.API
R6
classes to
implement the lapply
parallel operation.AsyncBackend
. Closes #35.CC BY 4.0
license icons in
README.md
for the package website..decorate
method of
ProgressTrackingContext
to be more flexible. More
specifically, the method will now throw when primitive functions are
provided for decoration. The method can now handle both inline functions
(i.e., function(x) x
) and functions that have a body
defined in terms of compound expressions (i.e.,
function(x) { x }
). Closes #32.export
operation in the
SyncBackend
and Context
classes to fallback to
the parent environment if the argument environment
is not
provided.URL
in package NEWS.md
per
CRAN
request.URLs
in package documentation per
CRAN
request.v0.x.x
to
v1.x.x
.CC BY 4.0
license for package documentation,
vignettes, and website content.GitHub
workflow via codecov
and badge in README
.comparison.Rmd
to compare
parabar
to the pbapply
package, and provide
rough benchmarks. The comparison.Rmd
vignette is locally
build from the comparison.Rmd.orig
file (i.e., see this
resource for more information).Options$progress_log_path
to handle
generation of temporary files for tracking the execution progress of
tasks ran in parallel. Using a custom path (e.g., for debugging) is also
possible by setting this active binding to a desired path.Specification
for testing purposes.\dontrun{}
statements in examples with
try()
calls.Options
class to feature the
progress_log_path
active binding..decorate
method of ProgressTrackingContext
to use
bquote
instead of substitute
.ProgressDecorator
to ProgressTrackingContext
to be more consistent with the idea of backends that run in
contexts....
optional arguments to signature of
get_output
method in Service
interface..make_log
of
ProgressDecorator
to use the progress_log_path
option.UML
diagram to include missing classes and
changed methods. Also updated the corresponding diagram figure in the
package documentation.Specification
to prevent incompatible cluster
types (e.g., FORK
) on Windows
platforms. For
such cases, a warning is issues and the cluster type defaults to
PSOCK
.make_logo
can be ran on all platforms.README
to add CRAN
installation
instructions and new badges.\dontrun{}
from
make_logo
function examples as per CRAN
request in commit 87678fe
. However, this results in the
examples failing the R-CMD-check
workflow. Reverted the
change in commit a9d11ac
.LOGO
from
v1.x.x
to v0.x.x
.SyncBackend
class.pkgdown
reference
section.clear
: to clean a provided backend instance.export
: to export variables from a given environment to
the .GlobalEnv
of the backend instance.peek
: to list the variables names available on the
backend instance.evaluate
: to evaluate arbitrary expressions on the
backend instance.Helper$check_object_type
method checks if the type of
an object matches an expected type. If that is not the case, the helper
throws an error (i.e., Exception$type_not_assignable
)..scss
styles to override the table column width for
the exported wrappers table in the pkgdown
website.README
and package documentation to mention new
exported wrappers.pkgdown
.roxygen2
@examples
for exported
wrappers. The code for the examples is located in the documentation for
the start_backend
function. All other exported wrappers
(i.e., clear
, export
, peek
,
evaluate
, and par_sapply
) inherit the
@examples
section from start_backend
.@seealso
documentation
sections.backend
argument of par_sapply
to
backend = NULL
. This implies, that par_sapply
without a backend behaves identically to base::sapply
.export
method to use the .GlobalEnv
as fallback when exporting variables.extra.scss
to improve
documentation website.S3
print method for the LOGO
object.get_option
,
set_option
, and set_default_options
.README
. More specifically, added description
for Service
interface methods and enabled documentation
linking (i.e., via ?
) for pkgdown
website.Options
class.ProgressBar
use
wait = TRUE
when fetching the output.evaluate
backend operation. The
expression passed to evaluate
was not correctly passed down
the function chain to parallel::clusterCall
. See this question on
StackOverflow
for clarifications. Closes #9..bar_config
in
ProgressDecorator
class to list
. This way, the
configure_bar()
method of ProgressBar
class
becomes an optional step.filelock
package. Closes #8.DESCRIPTION
.man-roxygen/parabar.R
\html{...}
to
fix HTML
validation errors. Closes #6.DESCRIPTION
file to comply
with CRAN
requirements.README
file.par_sapply
exported function due
to exceeding the CRAN
time limit for running examples.pkgdown
and GitHub
pages.GitHub
workflow to automatically check the package
on several platforms and R
versions.R6
classes as developer API and regenerate
the namespace.README
file.par_sapply
to run tasks in
parallel and display a progress bar if appropriate.stop_backend
to stop a backend
instance.start_backend
to create a backend
instance based on the specified specification.configure_bar
for configuring the
type and behavior of the progress bar.set_option
for
Helper$set_option
. This function is available to end-users
and can be used to set the value of a specific option.Helper$set_option
is a wrapper around
base::getOption
that sets the value of a specific option if
it exists, or throws an error otherwise.DESCRIPTION
file.parabar
logo startup message for interactive
R
sessions.ASCII
template in
inst/assets/logo/parabar-logo.txt
.get_option
for
Helper$get_option
. This function is available to end-users
and can be used to get the value of a specific option or its default
value.Helper$get_option
is a wrapper around
base::getOption
that returns the value of a specific option
if it exists, or the default value set by the Options
class
otherwise.Options
R6
class and
set_default_options
function to manage package options for
parabar
. The documented fields of the Options
class represent the options that can be configured by the user. The
set_default_options
function can be used to set the default
package options and is automatically run at package load time. In a
nutshell, set_default_options
stores an instance of
Options
in the base::.Options
list under the
key parabar
.backend
and context
objects. A
backend
represents a set of operations that can be deployed
on a cluster produced by parallel::makeCluster
. The
backend, therefore, interacts with the cluster via specific operations
defined by the Service
interface. The context
represents the specific conditions in which the backend operations are
invoked. A regular context object simply forwards the call to the
corresponding backend method. However, a more complex context can
augment the operation before invoking the backend operation. One example
of a complex context is the ProgressDecorator
class. This
class extends the regular Context
class and decorates the
backend sapply
operation to provide progress tracking and
display a progress bar.Service
interface, the client
can interact with context objects as if they represents instances of
Backend
type. This release introduces the following
contexts:
Context
: represents a regular context that wraps
backend objects. In other words, all Service
methods calls
implemented by this context are forwarded to the corresponding methods
implemented by the backend object.ProgressDecorator
: represents a progress tracking
context. This context decorates the sapply
method available
on the backend instance to log the progress after each task execution
and display a progress bar.ContextFactory
: represents a blueprint for obtaining
specific context instances.Service
: represents an interface that all
concrete backends must implement. It contains the methods (i.e.,
operations) that can be requested from a given backend instance. These
methods form the main API of parabar
.Backend
: represents an abstract class that implements
the Service
interface. It contains fields and methods
relevant to all concrete backend implementations that extend the
class.SyncBackend
: represents a concrete implementation for a
synchronous backend. When executing a task in parallel via the
sapply
method, the caller process (i.e., usually the
interactive R
session) is blocked until the task finishes
executing.AsyncBackend
: represents a concrete implementation for
an asynchronous backend. After lunching a task in parallel via the
sapply
method, the method returns immediately leaving the
caller process (e.g., the interactive R
session) free. The
computation of the task is offloaded to a permanent background
R
session. One can read the state of the task using the
public field (i.e., active binding) task_state
.
Furthermore, the results can be fetched from the background session
using the get_output
method, which can either block the
main process to wait for the results, or attempt to fetch them
immediately and throw an error if not successful.Specification
: represents an auxiliary class that
encapsulates the logic for determining the type of cluster to create
(i.e., via parallel::makeCluster
), and the number of nodes
(i.e., R
processes) for the cluster.TaskState
: represents an auxiliary class that
encapsulates the logic for determining the state of a task.BackendFactory
: represents a blueprint for obtaining
concrete backend implementations.Helper
, Warning
, and
Exception
R6
classes. These classes contain
static member methods that provide useful utilities, handle warning
messages, and throw informative errors, respectively.UML
diagram for package classes. The classes
provided by parabar
can be split in three categories,
namely (1) backend classes responsible for managing clusters, (2)
context classes that decorate backend objects with additional
functionality (e.g., progress tracking), and (3) progress bar classes
providing a common interface for creating and interacting with various
progress bars.Bar
abstraction for working with progress bars in
R
. Currently, two types of of progress bars are supported
(i.e., BasicBar
and ModernBar
).
BasicBar
uses as engine the
utils::txtProgressBar
, and ModernBar
relies on
the R6
class obtained from
progress::progress_bar
. Specific concrete instances of
these bar types can be requested from the BarFactory
.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.