Home > Articles > Open Source > Python

  • Print
  • + Share This
Like this article? We recommend

Summary of the Microthread Library

The uthread module offers a rich set of classes and functions. Some items familiar from other threading libraries are missing, notably some way to catch an exception that has been thrown inside a thread, but they will no doubt be implemented. In the small example just discussed, we didn't need everything; certainly we didn't need to use locks or semaphores.

Here I give an overview of almost all that the uthread library contains, with docstrings or my own comments.

Scheduler Functions

These functions are not bound to a class; they work on any thread that has been created:

  • new(func, *args, **kwargs): Creates a new thread object for function func. Returns the thread.

  • newresistant(func, *args, **kwargs): Create a new resistant thread object for function func and return it.

  • run(): Starts all created threads and runs them until they're completed.

  • runAndContinue*(): Creates a Main thread and starts all created threads from there. Returns immediately.

  • switchContext(): Forces a context switch from the current thread to the next.

  • block(): Blocks the currently running thread.

  • wait(duration): Suspends the current thread for duration seconds.

  • waitUntil(until): SUSPENDs a thread until time.time() >= until.

  • exit(): Exits the current thread with SystemExit.

  • exitAll(reallyAll=FALSE): Exits all threads; if reallyAll==TRUE even resistant threads are exited.

  • exitOthers(reallyAll=FALSE): Exits all threads except the current thread.

  • getCurrentThread(): Returns the thread that's currently running.

  • getActiveThreads(): Gets a list of all active threads.

  • getBlockedThreads(): Gets a list of all blocked threads.

  • getAllThreads(): Gets a list of all threads.

  • microThreadsRunning(): This is TRUE if there are threads running.

Utility Functions

These are module-level functions that don't directly influence the scheduler:

  • atomic(func, *args, **kwargs): Performs a function call as a microthread-safe operation. No context switches will occur while this code is running. It can therefore completely mess up your application.

  • startCritical and endCritical(): This is a bit like atomic. Whatever happens between a call to startCritical and endCritical will not be interrupted by context switches.

  • getExceptionString(): Gets the (probably) last exception that was raised. This is not entirely reliable.

  • external(scheduler, function, *args, **kwargs): Performs a function within a microthread for an external thread.

  • random(x=1): Performs an atomic random. Note that the standard Python random module uses all kinds of state and is therefore not thread-safe.

  • uniqueId(): Returns an integer that is unique within the complete session.

Blockable Classes

The microthread library provides a rich selection of classes that implement blocking behavior.


Blockable is the root of all classes that block threads:

  • wait(): Waits until the blockable object is released.

  • notify(): Notifies the next thread that's waiting for the block.

  • notifyAll(): Notifies all threads.


This is just a mix-in that provides a mechanism for an external process to release a blocked thread.


This is a primitive lock object.

  • acquire(): Tries to take ownership of the lock. Blocks until success.

  • release(): Releases ownership of the lock.


A re-entrant lock object. This means that the same thread can acquire the lock more than once.

  • acquire(): Acquires the lock. Blocks until success.

  • release(): Releases the lock.


The Condition is familiar from the standard Python threads. You can use it to regulate access to a shared state.

  • __init__(lock=None): Creates the Condition.

  • wait(timeout=None): Waits until the timeout is reached.

  • notify(self, n=1)

  • notifyAll()


Semaphores protect globally accessible resources from the effects of context switching.

  • acquire()

  • release()


A Queue is a microthread-safe mechanism for creating a first-in-first-out (FIFO) sequence object. Queues allow for processing messages in order when they're delivered (without the need for constantly polling to see whether new messages are available).

  • put(x): Puts a value in the queue.

  • get(): Gets a value from the queue.

  • unget(x): Puts a value at the first place in the queue.


See ExternalBlockable and Queue.


A SnailQueue is an intentionally slow Queue. It allows for modeling communications latencies. Each new message arriving in the queue is delayed for either a set period (if randomize = 0), or a randomized fraction of that period (if randomize is true). See also Queue.

  • __init__(t=.1, randomize=FALSE)

  • mythread(): This is more or less an internal function used to create delays.

  • put: Adds a value to the queue.

  • close: Closes the queue. Nothing can be retrieved from it; all contents are lost.


A synchronizer provides a method for synchronizing threads. All the threads block on the sync() method, except for the last one. When it reaches the sync() method, all the others are unblocked. The synchronizer must be initialized with the expected number of threads.

  • __init__(maxcount): The number of threads that are to be synchronized.

  • sync(): Call this to block your thread until the last one is ready.


An ActiveTimer blocks a thread for a given amount of time, and then releases the threads, hogging the processor—it doesn't yield. If you want to reduce load on the processor, you'd better use a plain Timer.

  • __init__(interval=1): Creates the timer.

  • wait(): Blocks everything until the timer is finished. If it never finishes (because the interval is negative), your app will hang completely.

  • start(): Starts the timer.


Timers are object-oriented interfaces to the wait/waitUntil functions. Timers block the current thread for a given period of time, measured from the time when the timer's start method is called.

The primary benefit of the Timer versus the ActiveTimer is that the Timer allows for yielding the processor for other microthreads or operating system threads, preventing them from "hogging" the processor.

See the global functions wait and waitUntil for procedural mechanisms with similar functionality.

  • __init__(interval): Creates the timer with the time it has to block in seconds.

  • start(): Starts the timer.

  • wait(): Blocks the thread until the timer has finished.


Thread objects are mostly created with the new() factory function. There are a few things lacking, in comparison with the standard Thread module that comes with Python. The most important is setDemonic—you cannot create daemon threads. There is some initial support for joining threads, but it's commented out in the source.

The standard Python threading doesn't support the gentle stopping of threads—something like thread.stop() should be very handy. However, it's not likely that this will ever be built in. Java had this feature, but it has been deprecated for a long time.

  • __init__(name, func, *args, **kwargs): Creates a named thread with the function func.

  • setResistant(onOff=TRUE): A resistant thread won't stop when an exception occurs in another thread.

  • isResistant(): Checks whether this thread is resistant to exceptions in other threads.

  • isActive(): Checks whether the thread is active.

  • isBlocked(): Checks whether the thread is blocked.

  • start(task=None, scheduler=None): Starts or restarts the thread.

  • block(): Blocks the thread until it's restarted with start().

  • wait(duration): Blocks the thread for duration seconds.

  • waitUntil(until): Blocks the thread until a certain time has been reached.

  • exit(): Exits the thread by raising a SystemExit exception.

  • postException(exc=SystemExit, val=None): Posts an exception to this thread.

  • handleException(exception): This method gets called whenever an exception occurs in the thread. You can override this; by default, it just prints a stacktrace.


You only have to call the Scheduler class explicitly if you combine regular threading with microthreading; then there is one Scheduler object for every system thread.

  • + Share This
  • 🔖 Save To Your Account