resourcing related classes and functions
Project description
Release 20190812:
MultiOpenMixin: no longer subclass cs.obj.O.
MultiOpenMixin: remove lock
param support, the mixin has its own lock.
MultiOpen: drop lock
param support, no longer used by MultiOpenMixin.
MultiOpenMixin: do finalise inside the lock for the same reason as shutdown (competition with open/startup).
MultiOpenMixin.close: new unopened_ok=False
parameter intended for callback closes which might fire even if the initial open does not occur.
Resource management classes and functions.
Class ClosedError
MRO: builtins.Exception
, builtins.BaseException
Exception for operations invalid when something is closed.
Class MultiOpen
MRO: MultiOpenMixin
Context manager class that manages a single open/close object
using a MultiOpenMixin.
Method MultiOpen.__init__(self, openable, finalise_later=False)
Initialise: save the openable
and call the MultiOpenMixin initialiser.
Class MultiOpenMixin
A mixin to count open and close calls, and to call .startup on the first .open and to call .shutdown on the last .close.
Recommended subclass implementations do as little as possible during init, and do almost all setup during startup so that the class may perform multiple startup/shutdown iterations.
If used as a context manager calls open()/close() from enter() and exit().
Multithread safe.
This mixin defines ._lock = RLock(); subclasses need not bother, but may supply their own lock.
Classes using this mixin need to define .startup and .shutdown.
Method MultiOpenMixin.__init__(self, finalise_later=False)
Initialise the MultiOpenMixin state.
Parameters:
finalise_later
: do not notify the finalisation Condition on shutdown, require a separate call to .finalise(). This is mode is useful for objects such as queues where the final close prevents further .put calls, but users calling .join may need to wait for all the queued items to be processed.
TODO:
subopens
: if true (default false) then .open will return a proxy object with its own .closed attribute set by the proxy's .close.
Function not_closed(func)
Decorator to wrap methods of objects with a .closed property which should raise when self.closed.
Class Pool
MRO: cs.obj.O
A generic pool of objects on the premise that reuse is cheaper than recreation.
All the pool objects must be suitable for use, so the
new_object
callable will typically be a closure.
For example, here is the init for a per-thread AWS Bucket using a
distinct Session:
def __init__(self, bucket_name):
Pool.__init__(self, lambda: boto3.session.Session().resource('s3').Bucket(bucket_name)
Method Pool.__init__(self, new_object, max_size=None, lock=None)
Initialise the Pool with creator new_object
and maximum size max_size
.
Parameters:
new_object
is a callable which returns a new object for the Pool.max_size
: The maximum size of the pool of available objects saved for reuse. If omitted orNone
, defaults to 4. If 0, no upper limit is applied.lock
: optional shared Lock; if omitted orNone
a new Lock is allocated
Class RunState
A class to track a running task whose cancellation may be requested.
Its purpose is twofold, to provide easily queriable state
around tasks which can start and stop, and to provide control
methods to pronounce that a task has started (.start
),
should stop (.cancel
)
and has stopped (.stop
).
A RunState
can be used a a context manager, with the enter
and exit methods calling .start
and .stop
respectively.
Note that if the suite raises an exception
then the exit method also calls .cancel
before the call to .stop
.
Monitor or daemon processes can poll the RunState
to see when
they should terminate, and may also manage the overall state
easily using a context manager.
Example:
def monitor(self):
with self.runstate:
while not self.runstate.cancelled:
... main loop body here ...
A RunState
has three main methods:
.start()
: set.running
and clear.cancelled
.cancel()
: set.cancelled
.stop()
: clear.running
A RunState
has the following properties:
cancelled
: true if.cancel
has been called.running
: true if the task is running. Further, assigning a true value to it also sets.start_time
to now. Assigning a false value to it also sets.stop_time
to now.start_time
: the time.running
was last set to true.stop_time
: the time.running
was last set to false.run_time
:max(0,.stop_time-.start_time)
stopped
: true if the task is not running.stopping
: true if the task is running but has been cancelled.notify_start
: a set of callables called with theRunState
instance to be called whenever.running
becomes true.notify_end
: a set of callables called with theRunState
instance to be called whenever.running
becomes false.notify_cancel
: a set of callables called with theRunState
instance to be called whenever.cancel
is called.
Class RunStateMixin
Mixin to provide convenient access to a RunState
.
Provides: .runstate
, .cancelled
, .running
, .stopping
, .stopped
.
Method RunStateMixin.__init__(self, runstate=None)
Initialise the RunStateMixin
; sets the .runstate
attribute.
runstate
: RunState
instance or name.
If a str
, a new RunState
with that name is allocated.
Release Log
Release 20190812:
MultiOpenMixin: no longer subclass cs.obj.O.
MultiOpenMixin: remove lock
param support, the mixin has its own lock.
MultiOpen: drop lock
param support, no longer used by MultiOpenMixin.
MultiOpenMixin: do finalise inside the lock for the same reason as shutdown (competition with open/startup).
MultiOpenMixin.close: new unopened_ok=False
parameter intended for callback closes which might fire even if the initial open does not occur.
Release 20190617: RunState.exit: if an exception was raised call .canel() before calling .stop().
Release 20190103: Bugfixes for context managers. MultiOpenMixin fixes and changes. RunState improvements.
Release 20171024: bugfix MultiOpenMixin finalise logic and other small logic fixes and checs new class RunState for tracking or controlling a running task
Release 20160828: Use "install_requires" instead of "requires" in DISTINFO.
Release 20160827: BREAKING CHANGE: rename NestingOpenCloseMixin to MultiOpenMixin. New Pool class for generic object reuse. Assorted minor improvements.
Release 20150115: First PyPI release.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.