Clock object¶

The Clock object allows you to schedule a function call in the future; once or repeatedly at specified intervals. You can get the time elapsed between the scheduling and the calling of the callback via the dt argument:

# dt means delta-time
def my_callback(dt):
    pass

# call my_callback every 0.5 seconds
Clock.schedule_interval(my_callback, 0.5)

# call my_callback in 5 seconds
Clock.schedule_once(my_callback, 5)

# call my_callback as soon as possible (usually next frame.)
Clock.schedule_once(my_callback)

Note

If the callback returns False, the schedule will be canceled and won’t repeat.

If you want to schedule a function to call with default arguments, you can use the functools.partial python module:

from functools import partial

def my_callback(value, key, *largs):
    pass

Clock.schedule_interval(partial(my_callback, 'my value', 'my key'), 0.5)

Conversely, if you want to schedule a function that doesn’t accept the dt argument, you can use a lambda expression to write a short function that does accept dt. For Example:

def no_args_func():
    print("I accept no arguments, so don't schedule me in the clock")

Clock.schedule_once(lambda dt: no_args_func(), 0.5)

Note

You cannot unschedule an anonymous function unless you keep a reference to it. It’s better to add *args to your function definition so that it can be called with an arbitrary number of parameters.

Important

The callback is weak-referenced: you are responsible for keeping a reference to your original object/callback. If you don’t keep a reference, the ClockBase will never execute your callback. For example:

class Foo(object):
    def start(self):
        Clock.schedule_interval(self.callback, 0.5)

    def callback(self, dt):
        print('In callback')

# A Foo object is created and the method start is called.
# Because no reference is kept to the instance returned from Foo(),
# the object will be collected by the Python Garbage Collector and
# your callback will be never called.
Foo().start()

# So you should do the following and keep a reference to the instance
# of foo until you don't need it anymore!
foo = Foo()
foo.start()

Schedule before frame¶

New in version 1.0.5.

Sometimes you need to schedule a callback BEFORE the next frame. Starting from 1.0.5, you can use a timeout of -1:

Clock.schedule_once(my_callback, 0) # call after the next frame
Clock.schedule_once(my_callback, -1) # call before the next frame

The Clock will execute all the callbacks with a timeout of -1 before the next frame even if you add a new callback with -1 from a running callback. However, Clock has an iteration limit for these callbacks: it defaults to 10.

If you schedule a callback that schedules a callback that schedules a … etc more than 10 times, it will leave the loop and send a warning to the console, then continue after the next frame. This is implemented to prevent bugs from hanging or crashing the application.

If you need to increase the limit, set the max_iteration property:

from kivy.clock import Clock
Clock.max_iteration = 20

Triggered Events¶

New in version 1.0.5.

A triggered event is a way to defer a callback. It functions exactly like schedule_once() and schedule_interval() except that it doesn’t immediately schedule the callback. Instead, one schedules the callback using the ClockEvent returned by it. This ensures that you can call the event multiple times but it won’t be scheduled more than once. This is not the case with Clock.schedule_once():

# will run the callback twice before the next frame
Clock.schedule_once(my_callback)
Clock.schedule_once(my_callback)

# will run the callback once before the next frame
event = Clock.create_trigger(my_callback)
event()
event()

# will also run the callback only once before the next frame
event = Clock.schedule_once(my_callback)  # now it's already scheduled
event()  # won't be scheduled again
event()

In addition, it is more convenient to create and bind to the triggered event than using Clock.schedule_once() in a function:

from kivy.clock import Clock
from kivy.uix.widget import Widget

class Sample(Widget):
    def __init__(self, **kwargs):
        self._trigger = Clock.create_trigger(self.cb)
        super(Sample, self).__init__(**kwargs)
        self.bind(x=self._trigger, y=self._trigger)

    def cb(self, *largs):
        pass

Even if x and y changes within one frame, the callback is only run once.

CyClockBase.create_trigger() has a timeout parameter that behaves exactly like CyClockBase.schedule_once().

Changed in version 1.10.0: CyClockBase.create_trigger() now has a interval parameter. If False, the default, it’ll create an event similar to CyClockBase.schedule_once(). Otherwise it’ll create an event similar to CyClockBase.schedule_interval().

Unscheduling¶

An event scheduled with CyClockBase.schedule_once(), CyClockBase.schedule_interval(), or with CyClockBase.create_trigger() and then triggered can be unscheduled in multiple ways. E.g:

def my_callback(dt):
    pass

# call my_callback every 0.5 seconds
event = Clock.schedule_interval(my_callback, 0.5)

# call my_callback in 5 seconds
event2 = Clock.schedule_once(my_callback, 5)

event_trig = Clock.create_trigger(my_callback, 5)
event_trig()

# unschedule using cancel
event.cancel()

# unschedule using Clock.unschedule
Clock.unschedule(event2)

# unschedule using Clock.unschedule with the callback
# NOT RECOMMENDED
Clock.unschedule(my_callback)

The best way to unschedule a callback is with ClockEvent.cancel(). CyClockBase.unschedule() is mainly an alias for that for that function. However, if the original callback itself is passed to CyClockBase.unschedule(), it’ll unschedule all instances of that callback (provided all is True, the default, other just the first match is removed).

Calling CyClockBase.unschedule() on the original callback is highly discouraged because it’s significantly slower than when using the event.

Threading and Callback Order¶

Beginning with 1.10.0, all the events scheduled for the same frame, e.g. all the events scheduled in the same frame with a timeout of 0, well be executed in the order they were scheduled.

Also, all the scheduling and canceling methods are fully thread safe and can be safely used from external threads.

As a a consequence, calling CyClockBase.unschedule() with the original callback is now significantly slower and highly discouraged. Instead, the returned events should be used to cancel. As a tradeoff, all the other methods are now significantly faster than before.

Advanced Clock Details¶

The following section goes into the internal kivy clock details as well as the various clock options. It is meant only for advanced users.

Fundamentally, the Kivy clock attempts to execute any scheduled callback rhythmically as determined by the specified fps (frame per second, see maxfps in config). That is, ideally, given e.g. a desired fps of 30, the clock will execute the callbacks at intervals of 1 / 30 seconds, or every 33.33 ms. All the callbacks in a frame are given the same timestamp, i.e. the dt passed to the callback are all the same and it’s the difference in time between the start of this and the previous frame.

Because of inherent indeterminism, the frames do not actually occur exactly at intervals of the fps and dt may be under or over the desired fps. Also, once the timeout is “close enough” to the desired timeout, as determined internally, Kivy will execute the callback in the current frame even when the “actual time” has not elapsed the timeout amount.

Kivy offers now, since 1.10.0, multiple clocks with different behaviors.

Default Clock¶

The default clock (default) behaves as described above. When a callback with a timeout of zero or non-zero is scheduled, they are executed at the frame that is near the timeout, which is a function of the fps. So a timeout of zero would still result in a delay of one frame or about 1 / fps, typically a bit less but sometimes more depending on the CPU usage of the other events scheduled for that frame.

In a test using a fps of 30, a callback with a timeout of 0, 0.001, and 0.05, resulted in a mean callback delay of 0.02487, 0.02488, and 0.05011 seconds, respectively. When tested with a fps of 600 the delay for 0.05 was similar, except the standard deviation was reduced resulting in overall better accuracy.

Interruptible Clock¶

The default clock suffers from the quantization problem, as frames occur only on intervals and any scheduled timeouts will not be able to occur during an interval. For example, with the timeout of 0.05, while the mean was 0.05011, its values ranged between 0.02548 - 0.07348 and a standard deviation of 0.002. Also, there’s the minimum timeout of about 0.02487.

The interruptible clock (interrupt) will execute timeouts even during a frame. So a timeout of zero will execute as quickly as possible and similarly a non-zero timeout will be executed even during the interval.

This clock, and all the clocks described after this have an option, ClockBaseInterruptBehavior.interupt_next_only. When True, any of the behavior new behavior will only apply to the callbacks with a timeout of zero. Non-zero timeouts will behave like in the default clock. E.g. for this clock when True, only zero timeouts will execute during the the interval.

In a test using a fps of 30, a callback with a timeout of 0, 0.001, and 0.05, resulted in a mean callback delay of 0.00013, 0.00013, and 0.04120 seconds, respectively when ClockBaseInterruptBehavior.interupt_next_only was False. Also, compared to the default clock the standard deviation was reduced. When ClockBaseInterruptBehavior.interupt_next_only was True, the values were 0.00010, 0.02414, and 0.05034, respectively.

Free Clock¶

The interruptible clock may not be ideal for all cases because all the events are executed during the intervals and events are not executed anymore rhythmically as multiples of the fps. For example, there may not be any benefit for the graphics to update in a sub-interval, so the additional accuracy wastes CPU.

The Free clock (free_all) solves this by having Clock.xxx_free versions of all the Clock scheduling methods. By free, we mean the event is free from the fps because it’s not fps limited. E.g. CyClockBaseFree.create_trigger_free() corresponds to CyClockBase.create_trigger(). Only when an event scheduled using the Clock.xxx_free methods is present will the clock interrupt and execute the events during the interval. So, if no free event is present the clock behaves like the default clock, otherwise it behaves like the interrupt clock.

In a test using a fps of 30, a callback with a timeout of 0s, 0.001s, and 0.05s, resulted in a mean callback delay of 0.00012s, 0.00017s, and 0.04121s seconds, respectively when it was a free event and 0.02403s, 0.02405s, and 0.04829s, respectively when it wasn’t.

Free Only Clock¶

The Free clock executes all events when a free event was scheduled. This results in normal events also being execute in the middle of the interval when a free event is scheduled. For example, above, when a free event was absent, a normal event with a 0.001s timeout was delayed for 0.02405s. However, if a free event happened to be also scheduled, the normal event was only delayed 0.00014s, which may be undesirable.

The Free only clock (free_only) solves it by only executing free events during the interval and normal events are always executed like with the default clock. For example, in the presence of a free event, a normal event with a timeout of 0.001s still had a delay of 0.02406. So this clock, treats free and normal events independently, with normal events always being fps limited, but never the free events.

Summary¶

The kivy clock type to use can be set with the kivy_clock option the config. If KIVY_CLOCK is present in the environment it overwrites the config selection. Its possible values are as follows:

When kivy_clock is default, the normal clock, ClockBase, which limits callbacks to the maxfps quantization - is used.
When kivy_clock is interrupt, a interruptible clock, ClockBaseInterrupt, which doesn’t limit any callbacks to the maxfps - is used. Callbacks will be executed at any time.
When kivy_clock is free_all, a interruptible clock, ClockBaseFreeInterruptAll, which doesn’t limit any callbacks to the maxfps in the presence of free events, but in their absence it limits events to the fps quantization interval - is used.
When kivy_clock is free_only, a interruptible clock, ClockBaseFreeInterruptAll, which treats free and normal events independently; normal events are fps limited while free events are not - is used.

kivy.clock.Clock = None¶: The kivy Clock instance. See module documentation for details.

class kivy.clock.ClockEvent¶

Bases: builtins.object

A class that describes a callback scheduled with kivy’s Clock. This class is never created by the user; instead, kivy creates and returns an instance of this class when scheduling a callback.

An event can be triggered (scheduled) by calling it. If it’s already scheduled, nothing will happen, otherwise it’ll be scheduled. E.g.:

event = Clock.schedule_once(my_callback, .5)
event()  # nothing will happen since it's already scheduled.
event.cancel()  # cancel it
event()  # now it's scheduled again.

cancel()¶: Cancels the callback if it was scheduled to be called. If not scheduled, nothing happens.

clock¶: The CyClockBase instance associated with the event.

get_callback()¶: Returns the callback associated with the event. Callbacks get stored with a indirect ref so that it doesn’t keep objects alive. If the callback is dead, None is returned.

is_triggered¶: Returns whether the event is scheduled to have its callback executed by the kivy thread.

loop¶: Whether this event repeats at intervals of timeout.

next¶: The next ClockEvent in order they were scheduled.

prev¶: The previous ClockEvent in order they were scheduled.

release()¶: (internal method) Converts the callback into a indirect ref.

tick()¶: (internal method) Processes the event for the kivy thread.

timeout¶: The duration after scheduling when the callback should be executed.

class kivy.clock.FreeClockEvent¶

Bases: kivy._clock.ClockEvent

The event returned by the Clock.XXX_free methods of CyClockBaseFree. It stores whether the event was scheduled as a free event.

free¶: Whether this event was scheduled as a free event.

class kivy.clock.CyClockBase¶

Bases: builtins.object

The base clock object with event support.

clock_resolution¶

If the remaining time until the event timeout is less than clock_resolution, the clock will execute the callback even if it hasn’t exactly timed out.

If -1, the default, the resolution will be computed from config’s maxfps. Otherwise, the provided value is used. Defaults to -1.

create_trigger()¶

Create a Trigger event. Check module documentation for more information.

Returns: A ClockEvent instance. To schedule the callback of this instance, you can call it.

New in version 1.0.5.

Changed in version 1.10.0: interval has been added. If True, it create a event that is called every <timeout> seconds similar to schedule_interval(). Defaults to False.

get_events()¶: Returns the list of ClockEvent instances currently scheduled.

get_min_timeout()¶: Returns the remaining time since the start of the current frame for the event with the smallest timeout.

get_resolution()¶: Returns the minimum resolution the clock has. It’s a function of clock_resolution and maxfps provided at the config.

max_iteration¶: The maximum number of callback iterations at the end of the frame, before the next frame. If more iterations occur, a warning is issued.

on_schedule()¶: Function that is called internally every time an event is triggered for this clock. It takes the event as a parameter.

schedule_del_safe()¶

Schedule a callback. Might be called from GC and cannot be cancelled.

It’s unsafe to call various kinds of code, such as code with a lock, from a __del__ or __dealloc__ methods. Since Kivy’s Clock uses a lock, it’s generally unsafe to call from these methods. Instead, use this method, which is thread safe and __del__ or __dealloc__ safe, to schedule the callback in the kivy thread. It’ll be executed in order after the normal events are processed.

The callback takes no parameters and cannot be canceled.

New in version 1.11.0.

schedule_interval()¶

Schedule an event to be called every <timeout> seconds.

Returns: A ClockEvent instance. As opposed to create_trigger() which only creates the trigger event, this method also schedules it.

schedule_once()¶

Schedule an event in <timeout> seconds. If <timeout> is unspecified or 0, the callback will be called after the next frame is rendered.

Returns: A ClockEvent instance. As opposed to create_trigger() which only creates the trigger event, this method also schedules it.

Changed in version 1.0.5: If the timeout is -1, the callback will be called before the next frame (at tick_draw()).

unschedule()¶

Remove a previously scheduled event.

Parameters

callback: ClockEvent or a callable.

If it’s a ClockEvent instance, then the callback associated with this event will be canceled if it is scheduled.

If it’s a callable, then the callable will be unscheduled if it was scheduled.

Warning

Passing the callback function rather than the returned ClockEvent will result in a significantly slower unscheduling.

all: bool

If True and if callback is a callable, all instances of this callable will be unscheduled (i.e. if this callable was scheduled multiple times). Defaults to True.

Changed in version 1.9.0: The all parameter was added. Before, it behaved as if all was True.

class kivy.clock.CyClockBaseFree¶

Bases: kivy._clock.CyClockBase

A clock class that supports scheduling free events in addition to normal events.

Each of the create_trigger(), schedule_once(), and schedule_interval() methods, which create a normal event, have a corresponding method for creating a free event.

create_trigger()¶

Create a Trigger event. Check module documentation for more information.

Returns: A ClockEvent instance. To schedule the callback of this instance, you can call it.

New in version 1.0.5.

Changed in version 1.10.0: interval has been added. If True, it create a event that is called every <timeout> seconds similar to schedule_interval(). Defaults to False.

create_trigger_free()¶: Similar to create_trigger(), but instead creates a free event.

get_min_free_timeout()¶: Returns the remaining time since the start of the current frame for the free event with the smallest timeout.

schedule_interval()¶

Schedule an event to be called every <timeout> seconds.

Returns: A ClockEvent instance. As opposed to create_trigger() which only creates the trigger event, this method also schedules it.

schedule_interval_free()¶: Similar to schedule_interval(), but instead creates a free event.

schedule_once()¶

Schedule an event in <timeout> seconds. If <timeout> is unspecified or 0, the callback will be called after the next frame is rendered.

Returns: A ClockEvent instance. As opposed to create_trigger() which only creates the trigger event, this method also schedules it.

Changed in version 1.0.5: If the timeout is -1, the callback will be called before the next frame (at tick_draw()).

schedule_once_free()¶: Similar to schedule_once(), but instead creates a free event.

class kivy.clock.ClockBaseBehavior(**kwargs)[source]¶

Bases: builtins.object

The base of the kivy clock.

MIN_SLEEP = 0.005¶: The minimum time to sleep. If the remaining time is less than this, the event loop will continuo

frames¶: Number of internal frames (not necessarily drawed) from the start of the clock.

New in version 1.8.0.

frames_displayed¶: Number of displayed frames from the start of the clock.

frametime¶: Time spent between the last frame and the current frame (in seconds).

New in version 1.8.0.

get_boottime()[source]¶: Get the time in seconds from the application start.

get_fps()[source]¶: Get the current average FPS calculated by the clock.

get_rfps()[source]¶

Get the current “real” FPS calculated by the clock. This counter reflects the real framerate displayed on the screen.

In contrast to get_fps(), this function returns a counter of the number of frames, not the average of frames per second.

get_time()[source]¶: Get the last tick made by the clock.

idle()[source]¶: (internal) waits here until the next frame.

tick()[source]¶: Advance the clock to the next step. Must be called every frame. The default clock has a tick() function called by the core Kivy framework.

tick_draw()[source]¶: Tick the drawing counter.

time = functools.partial(<function _libc_clock_gettime_wrapper.<locals>._time>)¶

usleep(microseconds)[source]¶: Sleeps for the number of microseconds.

class kivy.clock.ClockBaseInterruptBehavior(interupt_next_only=False, **kwargs)[source]¶

Bases: kivy.clock.ClockBaseBehavior

A kivy clock which can be interrupted during a frame to execute events.

idle()[source]¶: (internal) waits here until the next frame.

usleep(microseconds)[source]¶: Sleeps for the number of microseconds.

class kivy.clock.ClockBaseInterruptFreeBehavior(**kwargs)[source]¶

Bases: kivy.clock.ClockBaseInterruptBehavior

A base class for the clock that interrupts the sleep interval for free events.

class kivy.clock.ClockBase(**kwargs)[source]¶

Bases: kivy.clock.ClockBaseBehavior, kivy._clock.CyClockBase

The default kivy clock. See module for details.

usleep(microseconds)[source]¶: Sleeps for the number of microseconds.

class kivy.clock.ClockBaseInterrupt(interupt_next_only=False, **kwargs)[source]¶

Bases: kivy.clock.ClockBaseInterruptBehavior, kivy._clock.CyClockBase

The interrupt kivy clock. See module for details.

class kivy.clock.ClockBaseFreeInterruptAll(**kwargs)[source]¶

Bases: kivy.clock.ClockBaseInterruptFreeBehavior, kivy._clock.CyClockBaseFree

The free_all kivy clock. See module for details.

class kivy.clock.ClockBaseFreeInterruptOnly(**kwargs)[source]¶

Bases: kivy.clock.ClockBaseInterruptFreeBehavior, kivy._clock.CyClockBaseFree

The free_only kivy clock. See module for details.

idle()[source]¶: (internal) waits here until the next frame.

kivy.clock.mainthread(func)[source]¶

Decorator that will schedule the call of the function for the next available frame in the mainthread. It can be useful when you use UrlRequest or when you do Thread programming: you cannot do any OpenGL-related work in a thread.

Please note that this method will return directly and no result can be returned:

@mainthread
def callback(self, *args):
    print('The request succeeded!',
          'This callback is called in the main thread.')


self.req = UrlRequest(url='http://...', on_success=callback)

New in version 1.8.0.

« Cache manager

Compatibility module for Python 2.7 and >= 3.4 »

Version

Quick search

Table Of Contents

Clock object¶

Schedule before frame¶

Triggered Events¶

Unscheduling¶

Threading and Callback Order¶

Advanced Clock Details¶

Default Clock¶

Interruptible Clock¶

Free Clock¶

Free Only Clock¶

Summary¶