Source code for psychopy.clock

#!/usr/bin/env python
# -*- coding: utf-8 -*-

"""
Created on Tue Apr 23 11:28:32 2013

Provides the high resolution timebase used by psychopy, and defines some time
related utility Classes.

Moved functionality from core.py so a common code
base could be used in core.py and logging.py; vs. duplicating the getTime and
Clock logic.

@author: Sol
@author: Jon
"""

# Part of the PsychoPy library
# Copyright (C) 2002-2018 Jonathan Peirce (C) 2019-2024 Open Science Tools Ltd.
# Distributed under the terms of the GNU General Public License (GPL).
import logging
import time
import sys
from datetime import datetime

from packaging.version import Version

try:
    import pyglet
except ImportError:
    pass  # pyglet is not installed

from psychopy.constants import STARTED, NOT_STARTED, FINISHED
import psychopy.logging  # Absolute import to work around circularity

# set the default timing mechanism
getTime = None

# Select the timer to use as the psychopy high resolution time base. Selection
# is based on OS and Python version.
#
# Three requirements exist for the psychopy time base implementation:
#     A) The Python interpreter does not apply an offset to the times returned
#        based on when the timer module being used was loaded or when the
#        timer function first called was first called.
#     B) The timer implementation used must be monotonic and report elapsed
#        time between calls, 'not' system or CPU usage time.
#     C) The timer implementation must provide a resolution of 50 usec or
#        better.
#
# Given the above requirements, psychopy selects a timer implementation as
# follows:
#     1) On Windows, the Windows Query Performance Counter API is used using
#        ctypes access.
#     2) On other OS's, if the Python version being used is 2.6 or lower,
#        time.time is used. For Python 2.7 and above, the timeit.default_timer
#        function is used.
try:
    import psychtoolbox
    havePTB = True
except ImportError:
    havePTB = False

if havePTB:
    # def getTime():
    #     secs, wallTime, error = psychtoolbox.GetSecs('allclocks')
    #     return wallTime
    getTime = psychtoolbox.GetSecs
elif sys.platform == 'win32':
    from ctypes import byref, c_int64, windll
    _fcounter = c_int64()
    _qpfreq = c_int64()
    windll.Kernel32.QueryPerformanceFrequency(byref(_qpfreq))
    _qpfreq = float(_qpfreq.value)
    _winQPC = windll.Kernel32.QueryPerformanceCounter

    def getTime():
        _winQPC(byref(_fcounter))
        return _fcounter.value / _qpfreq

elif sys.platform == "darwin":
    # Monotonic getTime with absolute origin. Suggested by @aforren1, and
    # copied from github.com/aforren1/toon/blob/master/toon/input/mac_clock.py 
    import ctypes
    _libc = ctypes.CDLL('/usr/lib/libc.dylib', use_errno=True)

    # create helper class to store data
    class mach_timebase_info_data_t(ctypes.Structure):
        _fields_ = (('numer', ctypes.c_uint32),
                    ('denom', ctypes.c_uint32))

    # get function and set response type
    _mach_absolute_time = _libc.mach_absolute_time
    _mach_absolute_time.restype = ctypes.c_uint64

    # calculate timebase
    _timebase = mach_timebase_info_data_t()
    _libc.mach_timebase_info(ctypes.byref(_timebase))
    _ticks_per_second = _timebase.numer / _timebase.denom * 1.0e9

    # scaling factor so that timing works correctly on Intal and Apple Silicon
    _scaling_factor = _timebase.numer / _timebase.denom

    # then define getTime func
    def getTime():
        return (_mach_absolute_time() * _scaling_factor) / 1.0e9

else:
    import timeit
    getTime = timeit.default_timer


class Timestamp(float):
    """
    Object to represent a timestamp, which can return itself in a variety of formats.

    Parameters
    ----------
    value : float or str
        Current time, as either:
        - float : Seconds since arbitrary start time (if only using as a duration)
        - float : Seconds since epoch (for an absolute time)
        - str : Time string in the format specified by the parameter `format`
    format : str or class
        Time format string (as in time.strftime) indicated how to convert this timestamp to a string, and how to
        interpret its value if given as a string. Use `float` (default) to always print timestamp as a float, or `str`
        as
    lastReset : float
        Epoch time at last clock reset. Will be added to raw value if printing to string.

    """
    def __new__(cls, value, format=float, lastReset=0.0):
        # if given a string, attempt to parse it using the given format
        if isinstance(value, str):
            # substitute nonspecified str format for ISO 8601
            if format in (str, "str"):
                format = "%Y-%m-%d_%H:%M:%S.%f%z"
            # try to parse
            try:
                value = datetime.strptime(value, format)
            except ValueError as err:
                # if parsing fails, try again without %z (as this is excluded in GMT)
                if format.endswith("%z"):
                    value = datetime.strptime(value, format[:-2])
            # convert to timestamp
            value = datetime.timestamp(value) - lastReset

        return float.__new__(cls, value)

    def __init__(self, value, format=float, lastReset=0.0):
        self.lastReset = lastReset
        self.format = format
        # create self as float representing the time
        float.__init__(value)

    def __str__(self):
        # use strftime to return with own format
        return self.strftime(format=self.format)

    def __format__(self, format_spec):
        if self.format in (float, "float"):
            # format as normal if float is requested
            return float.__format__(self, format_spec)
        else:
            # otherwise just stringify
            return str(self)

    def resolve(self, format=None):
        """
        Get the value of this timestamp as a simple value, either str or float.

        Parameters
        ----------
        format : str, class or None
            Time format string, as in time.strftime, or `float` to return as a float. Defaults (None) to using the
            format given when this timestamp was initialised.

        Returns
        -------
        str, float
            The value of this timestamp in the requested format.
        """
        # if format is unspecified, use own default
        if format is None:
            format = self.format
        # if format is float, return as simple (non-timestamp) float
        if format in (float, "float"):
            return float(self)
        # otherwise, format to string in requested format
        return self.strftime(format=format)

    def strftime(self, format="%Y-%m-%d_%H:%M:%S.%f%z"):
        """
        Format this timestamp into a string with the given format.

        Parameters
        ----------
        format : str, class or None
            Time format string, as in time.strftime, or `float` to print as a float. Defaults (None) to using the
            format given when this timestamp was initialised.

        Returns
        -------
        str
            This timestamp as a string
        """
        # if format is unspecified, use own default
        if format in (None, "None"):
            format = self.format
        # if format is float, print using base method
        if format in (float, "float"):
            return float.__str__(self)
        # substitute nonspecified str format for ISO 8601
        if format in (str, "str"):
            format = "%Y-%m-%d_%H:%M:%S.%f%z"
        # convert to datetime
        now = datetime.fromtimestamp(self + self.lastReset)
        # format
        return now.strftime(format)


[docs]class MonotonicClock: """A convenient class to keep track of time in your experiments using a sub-millisecond timer. Unlike the :class:`~psychopy.core.Clock` this cannot be reset to arbitrary times. For this clock t=0 always represents the time that the clock was created. Don't confuse this `class` with `core.monotonicClock` which is an `instance` of it that got created when PsychoPy.core was imported. That clock instance is deliberately designed always to return the time since the start of the study. Version Notes: This class was added in PsychoPy 1.77.00 """ def __init__(self, start_time=None, format=float): super(MonotonicClock, self).__init__() if start_time is None: # this is sub-millisecond timer in python self._timeAtLastReset = getTime() else: self._timeAtLastReset = start_time self._epochTimeAtLastReset = time.time() # store default format self.format = format
[docs] def getTime(self, applyZero=True, format=None): """ Returns the current time on this clock in secs (sub-ms precision). Parameters ---------- applyZero : bool If applying zero then this will be the time since the clock was created (typically the beginning of the script). If not applying zero then it is whatever the underlying clock uses as its base time but that is system dependent. e.g. can be time since reboot, time since Unix Epoch etc. Only applies when format is `float`. format : type, str or None Format in which to show timestamp when converting to a string. Can be either: - time format codes: Time will return as a string in that format, as in time.strftime - `str`: Time will return as a string in ISO 8601 (YYYY-MM-DD_HH:MM:SS.mmmmmmZZZZ) - `None`: Will use this clock's `format` attribute Returns ------- Timestamp Time with format requested. """ # substitute no format for default if format in (None, "None"): format = self.format # substitute nonspecified str format for ISO 8601 if format in (str, "str"): format = "%Y-%m-%d_%H:%M:%S.%f%z" # get time since last reset t = getTime() - self._timeAtLastReset # get last reset time from epoch lastReset = self._epochTimeAtLastReset if not applyZero: # if not applying zero, add epoch start time to t rather than supplying it t += self._epochTimeAtLastReset lastReset = 0 return Timestamp(t, format, lastReset=lastReset)
[docs] def getLastResetTime(self): """ Returns the current offset being applied to the high resolution timebase used by Clock. """ return self._timeAtLastReset
monotonicClock = MonotonicClock()
[docs]class Clock(MonotonicClock): """A convenient class to keep track of time in your experiments. You can have as many independent clocks as you like (e.g. one to time responses, one to keep track of stimuli ...) This clock is identical to the :class:`~psychopy.core.MonotonicClock` except that it can also be reset to 0 or another value at any point. """ def __init__(self, format=float): super(Clock, self).__init__(format=format)
[docs] def reset(self, newT=0.0): """Reset the time on the clock. With no args time will be set to zero. If a float is received this will be the new time on the clock """ self._timeAtLastReset = getTime() + newT self._epochTimeAtLastReset = time.time()
[docs] def addTime(self, t): """Add more time to the Clock/Timer e.g.:: timer = core.Clock() timer.addTime(5) while timer.getTime() > 0: # do something """ self._timeAtLastReset -= t self._epochTimeAtLastReset -= t
[docs] def add(self, t): """ DEPRECATED: use .addTime() instead This function adds time TO THE BASE (t0) which, counterintuitively, reduces the apparent time on the clock """ psychopy.logging.log(msg=( "Clock.add() is deprecated in favor of .addTime() due to the counterintuitive design " "(it added time to the baseline, which reduced the values returned from getTime() )" ), level=psychopy.logging.DEPRECATION ) self._timeAtLastReset += t self._epochTimeAtLastReset += t
[docs]class CountdownTimer(Clock): """Similar to a :class:`~psychopy.core.Clock` except that time counts down from the time of last reset. Parameters ---------- start : float or int Starting time in seconds to countdown on. Examples -------- Create a countdown clock with a 5 second duration:: timer = core.CountdownTimer(5) while timer.getTime() > 0: # after 5s will become negative # do stuff """ def __init__(self, start=0): super(CountdownTimer, self).__init__() self._countdown_duration = start if start: self.reset()
[docs] def getTime(self): """Returns the current time left on this timer in seconds with sub-ms precision (`float`). """ return self._timeAtLastReset - getTime()
[docs] def addTime(self, t): """Add more time to the CountdownTimer e.g.: countdownTimer = core.CountdownTimer() countdownTimer.addTime(1) while countdownTimer.getTime() > 0: # do something """ self._timeAtLastReset += t self._epochTimeAtLastReset += t
[docs] def reset(self, t=None): """Reset the time on the clock. Parameters ---------- t : float, int or None With no args (`None`), time will be set to the time used for last reset (or start time if no previous resets). If a number is received, this will be the new time on the clock. """ if t is not None: self._countdown_duration = t Clock.reset(self, self._countdown_duration)
[docs]class StaticPeriod: """A class to help insert a timing period that includes code to be run. Parameters ---------- screenHz : int or None the frame rate of the monitor (leave as None if you don't want this accounted for) win : :class:`~psychopy.visual.Window` If a :class:`~psychopy.visual.Window` is given then :class:`StaticPeriod` will also pause/restart frame interval recording. name : str Give this StaticPeriod a name for more informative logging messages. Examples -------- Typical usage for the static period:: fixation.draw() win.flip() ISI = StaticPeriod(screenHz=60) ISI.start(0.5) # start a period of 0.5s stim.image = 'largeFile.bmp' # could take some time ISI.complete() # finish the 0.5s, taking into account one 60Hz frame stim.draw() win.flip() # the period takes into account the next frame flip # time should now be at exactly 0.5s later than when ISI.start() # was called """ def __init__(self, screenHz=None, win=None, name='StaticPeriod'): self.status = NOT_STARTED self.countdown = CountdownTimer() self.name = name self.win = win if screenHz is None: self.frameTime = 0 else: self.frameTime = 1.0 / screenHz self._winWasRecordingIntervals = False
[docs] def start(self, duration): """Start the period. If this is called a second time, the timer will be reset and starts again Parameters ---------- duration : float or int The duration of the period, in seconds. """ self.status = STARTED self.countdown.reset(duration - self.frameTime) # turn off recording of frame intervals throughout static period if self.win: self._winWasRecordingIntervals = self.win.recordFrameIntervals self.win.recordFrameIntervals = False
[docs] def complete(self): """Completes the period, using up whatever time is remaining with a call to `wait()`. Returns ------- float `1` for success, `0` for fail (the period overran). """ self.status = FINISHED timeRemaining = self.countdown.getTime() if self.win: self.win.recordFrameIntervals = self._winWasRecordingIntervals if timeRemaining < 0: msg = ('We overshot the intended duration of %s by %.4fs. The ' 'intervening code took too long to execute.') vals = self.name, abs(timeRemaining) psychopy.logging.warn(msg % vals) return 0 wait(timeRemaining) return 1
def _dispatchWindowEvents(): """Helper function for :func:`~.psychopy.core.wait`. Handles window event if needed or returns otherwise. """ from . import core if not (core.havePyglet and core.checkPygletDuringWait): return # nop # let's see if pyglet collected any event in meantime try: # this takes focus away from command line terminal window: if Version(pyglet.version) < Version('1.2'): # events for sounds/video should run independently of wait() pyglet.media.dispatch_events() except AttributeError: # see http://www.pyglet.org/doc/api/pyglet.media-module.html#dispatch_events # Deprecated: Since pyglet 1.1, Player objects schedule themselves # on the default clock automatically. Applications should not call # pyglet.media.dispatch_events(). pass for winWeakRef in core.openWindows: win = winWeakRef() if (win.winType == "pyglet" and hasattr(win.winHandle, "dispatch_events")): win.winHandle.dispatch_events() # pump events
[docs]def wait(secs, hogCPUperiod=0.2): """Wait for a given time period. This function halts execution of the program for the specified duration. Precision of this function is usually within 1 millisecond of the specified time, this may vary depending on factors such as system load and the Python version in use. Window events are periodically dispatched during the wait to keep the application responsive, to avoid the OS complaining that the process is unresponsive. If `secs=10` and `hogCPU=0.2` then for 9.8s Python's `time.sleep` function will be used, which is not especially precise, but allows the cpu to perform housekeeping. In the final `hogCPUperiod` the more precise method of constantly polling the clock is used for greater precision. If you want to obtain key-presses during the wait, be sure to use pyglet and then call :func:`psychopy.event.getKeys()` after calling :func:`~.psychopy.core.wait()` If you want to suppress checking for pyglet events during the wait, do this once:: core.checkPygletDuringWait = False and from then on you can do:: core.wait(sec) This will preserve terminal-window focus during command line usage. Parameters ---------- secs : float or int Number of seconds to wait before continuing the program. hogCPUperiod : float or int Number of seconds to hog the CPU. This causes the thread to enter a 'tight' loop when the remaining wait time is less than the specified interval. This is set to 200ms (0.2s) by default. It is recommended that this interval is kept short to avoid stalling the processor for too long which may result in poorer timing. """ # Calculate the relaxed period which we periodically suspend the thread, # this puts less load on the CPU during long wait intervals. relaxedPeriod = secs - hogCPUperiod # wait loop, suspends the thread periodically and consumes CPU resources t0 = getTime() while True: elapsed = getTime() - t0 if elapsed > secs: # no more time left, break the loop break if elapsed > relaxedPeriod: # hog period sleepDur = 0.00001 # 0.1ms else: relaxedTimeLeft = relaxedPeriod - elapsed sleepDur = 0.01 if relaxedTimeLeft > 0.01 else relaxedTimeLeft time.sleep(sleepDur) _dispatchWindowEvents()
[docs]def getAbsTime(): """Get the absolute time. This uses the same clock-base as the other timing features, like `getTime()`. The time (in seconds) ignores the time-zone (like `time.time()` on linux). To take the timezone into account, use `int(time.mktime(time.gmtime()))`. Absolute times in seconds are especially useful to add to generated file names for being unique, informative (= a meaningful time stamp), and because the resulting files will always sort as expected when sorted in chronological, alphabetical, or numerical order, regardless of locale and so on. Version Notes: This method was added in PsychoPy 1.77.00 Returns ------- float Absolute Unix time (i.e., whole seconds elapsed since Jan 1, 1970). """ return int(time.mktime(time.localtime()))

Back to top