duplicity.dup_threading module

Duplicity specific but otherwise generic threading interfaces and utilities.

(Not called “threading” because we do not want to conflict with the standard threading module, and absolute imports require at least python 2.5.)

class duplicity.dup_threading.Value(value=None)[source]

A thread-safe container of a reference to an object (but not the object itself).

In particular this means it is safe to:

value.set(1)

But unsafe to:

value.get()[‘key’] = value

Where the latter must be done using something like:

def _setprop():
value.get()[‘key’] = value

with_lock(value, _setprop)

Operations such as increments are best done as:

value.transform(lambda val: val + 1)
acquire()[source]

Acquire this Value for mutually exclusive access. Only ever needed when calling code must perform operations that cannot be done with get(), set() or transform().

get()[source]

Returns the value protected by this Value.

release()[source]

Release this Value for mutually exclusive access.

set(value)[source]

Resets the value protected by this Value.

transform(fn)[source]

Call fn with the current value as the parameter, and reset the value to the return value of fn.

During the execution of fn, all other access to this Value is prevented.

If fn raised an exception, the value is not reset.

Returns the value returned by fn, or raises the exception raised by fn.

duplicity.dup_threading.async_split(fn)[source]

Splits the act of calling the given function into one front-end part for waiting on the result, and a back-end part for performing the work in another thread.

Returns (waiter, caller) where waiter is a function to be called in order to wait for the results of an asynchronous invokation of fn to complete, returning fn’s result or propagating it’s exception.

Caller is the function to call in a background thread in order to execute fn asynchronously. Caller will return (success, waiter) where success is a boolean indicating whether the function suceeded (did NOT raise an exception), and waiter is the waiter that was originally returned by the call to async_split().

duplicity.dup_threading.interruptably_wait(cv, waitFor)[source]

cv - The threading.Condition instance to wait on test - Callable returning a boolean to indicate whether

the criteria being waited on has been satisfied.

Perform a wait on a condition such that it is keyboard interruptable when done in the main thread. Due to Python limitations as of <= 2.5, lock acquisition and conditions waits are not interruptable when performed in the main thread.

Currently, this comes at a cost additional CPU use, compared to a normal wait. Future implementations may be more efficient if the underlying python supports it.

The condition must be acquired.

This function should only be used on conditions that are never expected to be acquired for extended periods of time, or the lock-acquire of the underlying condition could cause an uninterruptable state despite the efforts of this function.

There is no equivalent for acquireing a lock, as that cannot be done efficiently.

Example:

Instead of:

cv.acquire() while not thing_done:

cv.wait(someTimeout)

cv.release()

do:

cv.acquire() interruptable_condwait(cv, lambda: thing_done) cv.release()
duplicity.dup_threading.require_threading(reason=None)[source]

Assert that threading is required for operation to continue. Raise an appropriate exception if this is not the case.

Reason specifies an optional reason why threading is required, which will be used for error reporting in case threading is not supported.

duplicity.dup_threading.thread_module()[source]

Returns the thread module, or dummy_thread if threading is not supported.

duplicity.dup_threading.threading_module()[source]

Returns the threading module, or dummy_thread if threading is not supported.

duplicity.dup_threading.threading_supported()[source]

Returns whether threading is supported on the system we are running on.

duplicity.dup_threading.with_lock(lock, fn)[source]

Call fn with lock acquired. Guarantee that lock is released upon the return of fn.

Returns the value returned by fn, or raises the exception raised by fn.

(Lock can actually be anything responding to acquire() and release().)